Prepublication Copy--Subject to Further Editorial Correction

Signposts in Cyberspace:
The Domain Name System and Internet Navigation

Committee on Internet Navigation and the Domain Name System:
Technical Alternatives and Policy Implications

Computer Science and Telecommunications Board
Division on Engineering and Physical Sciences

Washington, D.C.

Prepublication Copy--Subject to Further Editorial Correction
THE NATIONAL ACADEMIES PRESS 500 Fifth Street, N.W. Washington, D.C. 20001

NOTICE: The project that is the subject of this report was approved by the Governing Board of the
National Research Council, whose members are drawn from the councils of the National Academy of
Sciences, the National Academy of Engineering, and the Institute of Medicine. The members of the
committee responsible for the report were chosen for their special competences and with regard for
appropriate balance.

Support for this project was provided by the U.S. Department of Commerce and the National Science
Foundation under Grant No. ANI-9909852 and by the National Research Council. Any opinions,
findings, conclusions, or recommendations expressed in this publication are those of the authors and do
not necessarily reflect the views of the National Science Foundation or the Commerce Department.

International Standard Book Number

Cover designed by Jennifer M. Bishop.

Copies of this report are available from the National Academies Press, 500 Fifth Street, N.W., Lockbox
285, Washington, D.C. 20055, (800) 624-6242 or (202) 334-3313 in the Washington metropolitan area.

Copyright 2005 by the National Academy of Sciences. All rights reserved.

Printed in the United States of America

Prepublication Copy--Subject to Further Editorial Correction

The National Academy of Sciences is a private, nonprofit, self-perpetuating society of distinguished scholars
engaged in scientific and engineering research, dedicated to the furtherance of science and technology and to their
use for the general welfare. Upon the authority of the charter granted to it by the Congress in 1863, the Academy has
a mandate that requires it to advise the federal government on scientific and technical matters. Dr. Bruce M. Alberts
is president of the National Academy of Sciences.
The National Academy of Engineering was established in 1964, under the charter of the National Academy of
Sciences, as a parallel organization of outstanding engineers. It is autonomous in its administration and in the
selection of its members, sharing with the National Academy of Sciences the responsibility for advising the federal
government. The National Academy of Engineering also sponsors engineering programs aimed at meeting national
needs, encourages education and research, and recognizes the superior achievements of engineers. Dr. Wm. A. Wulf
is president of the National Academy of Engineering.
The Institute of Medicine was established in 1970 by the National Academy of Sciences to secure the services of
eminent members of appropriate professions in the examination of policy matters pertaining to the health of the
public. The Institute acts under the responsibility given to the National Academy of Sciences by its congressional
charter to be an adviser to the federal government and, upon its own initiative, to identify issues of medical care,
research, and education. Dr. Harvey V. Fineberg is president of the Institute of Medicine.
The National Research Council was organized by the National Academy of Sciences in 1916 to associate the broad
community of science and technology with the Academy's purposes of furthering knowledge and advising the
federal government. Functioning in accordance with general policies determined by the Academy, the Council has
become the principal operating agency of both the National Academy of Sciences and the National Academy of
Engineering in providing services to the government, the public, and the scientific and engineering communities.
The Council is administered jointly by both Academies and the Institute of Medicine. Dr. Bruce M. Alberts and Dr.
Wm. A. Wulf are chair and vice chair, respectively, of the National Research Council.

Prepublication Copy--Subject to Further Editorial Correction

Prepublication Copy--Subject to Further Editorial Correction

ROGER LEVIEN, Strategy & Innovation Consulting, Chair
S. ROBERT AUSTEIN, Internet Systems Consortium
STANLEY M. BESEN, Charles River Associates
CHRISTINE L. BORGMAN, University of California, Los Angeles
TIMOTHY CASEY, University of Nevada
HUGH DUBBERLY, Dubberly Design Office
MARYLEE JENKINS, Arent Fox Kintner Plotkin & Kahn, PLLC
JOHN C. KLENSIN, Independent Consultant
MILTON L. MUELLER, Syracuse University
SHARON NELSON, Washington State Attorney General's Office
HAL R. VARIAN, University of California, Berkeley


ALAN S. INOUYE, Study Director
MARJORY S. BLUMENTHAL, Director (through June 2003)
MARGARET MARSH HUYNH, Senior Program Assistant
KRISTEN BATCH, Research Associate


Prepublication Copy--Subject to Further Editorial Correction

DAVID E. LIDDLE, U.S. Venture Partners, Co-Chair
JEANNETTE M. WING, Carnegie Mellon University, Co-Chair
ERIC BENHAMOU, Benhamou Global Ventures, LLC
DAVID D. CLARK, Massachusetts Institute of Technology, CSTB Chair Emeritus
WILLIAM DALLY, Stanford University
MARK E. DEAN, IBM Almaden Research Center
DEBORAH ESTRIN, University of California, Los Angeles
JOAN FEIGENBAUM, Yale University
HECTOR GARCIA-MOLINA, Stanford University
KEVIN KAHN, Intel Corporation
JAMES KAJIYA, Microsoft Corporation
MICHAEL KATZ, University of California, Berkeley
RANDY H. KATZ, University of California, Berkeley
WENDY A. KELLOGG, IBM T.J. Watson Research Center
SARA KIESLER, Carnegie Mellon University
BUTLER W. LAMPSON, Microsoft Corporation, CSTB Member Emeritus
TERESA H. MENG, Stanford University
TOM M. MITCHELL, Carnegie Mellon University
DANIEL PIKE, GCI Cable and Entertainment
FRED B. SCHNEIDER, Cornell University
WILLIAM STEAD, Vanderbilt University

KRISTEN BATCH, Research Associate
JENNIFER M. BISHOP, Program Associate
JANET BRISCOE, Manager, Program Operations
JON EISENBERG, Senior Program Officer
RENEE HAWKINS, Financial Associate
MARGARET MARSH HUYNH, Senior Program Assistant
HERBERT S. LIN, Senior Scientist
LYNETTE I. MILLETT, Senior Program Officer
JANICE SABUDA, Senior Program Assistant
GLORIA WESTBROOK, Senior Program Assistant

For more information on CSTB, see its Web site at <>, write to CSTB,
National Research Council, 500 Fifth Street, N.W., Washington, DC 20001, call (202) 334-2605,
or e-mail the CSTB at


Prepublication Copy--Subject to Further Editorial Correction

The Domain Name System (DNS), which was developed in the early 1980s, provides a way of
associating alphanumeric names, which are easier for humans to use, with the numerical addresses that
designate every location on the Internet. The system of DNS servers distributed across the Internet
invisibly converts the names--serving as signposts in cyberspace--into the numerical addresses required
by network routers to reach the signposted locations.
The mnemonic quality of domain names became a practical necessity when the rapid increase in
the use of e-mail and the World Wide Web (Web) caused the number of Internet users and uses to
increase tremendously. Web sites often became known to their visitors by their distinctive domain
names--for example, or Carefully chosen domain names often
enabled a searcher to navigate to a site simply by guessing (e.g., Consequently, those
signposts gained economic, social, cultural, and political value and they became the objects of pride,
competition, and dispute. It was fitting, therefore, that the DNS also provided the name--the "Dot Com
Era"--for the period of the 1990s when "gold rush fever" drove frenzied efforts to stake out and exploit
virtually every potentially valuable site on the Web. Inevitably, such efforts led to intense conflicts,
especially disputes involving trademarks, which provided the impetus for the 1998 congressional mandate
to initiate this study (see Box P.1). However, the passage of time, the rapid evolution of the Internet and
the DNS, the additional and differing interests of the funding agencies, and the logic of the committee's
charter have resulted in a report whose scope differs in some respects from the original congressional
request, but is as a result more responsive to the current interests of the report's sponsors and audience.


Although the initial feverish period of Internet exploitation appears to have passed, in its third
decade the DNS will face new challenges arising from continued growth in the size and scope of the
Internet and from its increasing integration into almost every aspect of human activity almost everywhere
on the globe. The Internet will need more signposts, in more languages, to satisfy more uses and users.
And the DNS will have to be carefully developed and managed to ensure that it can meet those needs
while continuing to provide reliable, efficient, and secure service.
Furthermore, even if the DNS successfully adapts and grows, users of the Internet will confront
new challenges in reaching the resources that they are seeking on the Internet, whether they are
educational, social, political, cultural, commercial, or recreational. The challenges will arise not from the
absence of resources or of signposts for them, but from their presence in such volume and variety that
navigating through the maze to find the right ones may become too arduous or too complex for most
users. Reciprocally, those who put resources on the Internet will want to be easily found by their
prospective users in the cluttered bazaar of competing or confusing resources and signposts on the
Internet. Thus, the larger issue of the third decade of the DNS is that of navigation through the Internet--
the need for its users to find their way quickly and confidently to the resources they desire and for its
resources to be easily and reliably found by the users they seek.
This study builds on CSTB's prior work related to the Internet, most notably on The Internet's
Coming of Age and The Digital Dilemma.1 One of the important lessons from this prior work is that
contentious issues in information technology policy (e.g., the domain name trademark issues as described

1 See Computer Science and Telecommunications Board (CSTB), National Research Council (NRC), 2001, The
Internet's Coming of Age
, Washington, D.C.: National Academy Press; and CSTB, NRC, 2000, The Digital
Dilemma: Intellectual Property in the Information Age
, Washington, D.C.: National Academy Press.

Prepublication Copy--Subject to Further Editorial Correction

Excerpt from Public Law 105-305


(b) Matters To Be Assessed in Study.--The study shall assess and, as appropriate, make recommendations for policy,
practice, or legislative changes relating to­
(1) the short-term and long-term effects on the protection of trademark rights and consumer interests of increasing or
decreasing the number of generic top-level domains;
(2) trademark rights clearance processes for domain names, including­
(A) whether domain name databases should be readily searchable through a common interface to facilitate the
clearing of trademark rights and proposed domain names across a range of generic top-level domains;
(B) the identification of what information from domain name databases should be accessible for the clearing of
trademark rights; and
(C) whether generic top-level domain registrants should be required to provide certain information;
(3) domain name trademark rights dispute resolution mechanisms, including how to--
(A) reduce trademark rights conflicts associated with the addition of any new generic top-level domains; and
(B) reduce trademark rights conflicts through new technical approaches to Internet addressing;
(4) choice of law or jurisdiction for resolution of trademark rights disputes relating to domain names, including
which jurisdictions should be available for trademark rights owners to file suit to protect such trademark rights;
(5) trademark rights infringement liability for registrars, registries, or technical management bodies;
(6) short-term and long-term technical and policy options for Internet addressing schemes and the impact of such
options on current trademark rights issues; and
(7) public comments on the interim report and on any reports that are issued by intergovernmental bodies.

in Public Law 105-305) are often much more complex and require analysis in a much larger context than
a popular characterization of "us versus them" would suggest. In the interval between the enactment of
Public Law 105-305 and the initiation of this study, CSTB was able to conduct preliminary background
work to develop a statement of task (see Box P.2) that addresses the congressional mandate but also
ensures that the necessary larger context is included explicitly. Moreover, the larger context was
necessary to respond appropriately to the interests of the National Science Foundation, which joined with
the U.S. Department of Commerce as co-sponsors of this study.


The CSTB convened a cross-disciplinary study committee comprising computer scientists and
engineers, information science/retrieval and navigation experts, lawyers, public policy analysts, a graphic
designer, economists, and business strategists. Many but not all of the members were directly engaged
with the DNS or with Internet navigation (see Appendix A for the biographies of committee members).
The committee members brought different and complementary perspectives to the examination of the
DNS and Internet navigation. In some cases, they also held views that strongly conflicted with those of
other committee members. The conclusions reached and the recommendations developed by the
committee are thus the products of a multidimensional examination of the issues and a careful negotiation
of agreements among members holding contrasting opinions. The sharp discussions and e-mail threads
fueled by the committee's diversity of experience and opinion helped it to avoid overly simple
conclusions or recommendations reflecting just one perspective. Information gathering, discussion,
argument, negotiation, and compromise were the stages the committee passed through in addressing most
of the topics.

Prepublication Copy--Subject to Further Editorial Correction

Statement of Task

This project will examine the future of Internet navigation and the Domain Name System (DNS)
in light of the evolution and interaction of Internet usage, information technology, the economy, and
society. The original purpose of the DNS was to provide identifiers for network objects that are more
easily remembered and enduring than the numerical addresses and port numbers used by the network
infrastructure. However, domain names are now often used for purposes for which they were not
originally intended, such as searching, corporate identification, and marketing. And certain domain
names, especially those in the .com top-level domain, have acquired substantial economic value, leading
to conflict and competition over their ownership and a perceived scarcity of desirable names.
The continuing increase in the number of Internet users and sites, the deepening integration of the
Internet into the economy and social processes, the growth in embedded computing devices, and the
possible introduction of permanent personal and object identifiers--among other factors--pose
challenges to the continued viability and usefulness of the DNS, as currently constituted. This project
will describe and evaluate emerging technologies and identify how they might affect the ability of users to
find what they are seeking on the Internet and the role of the DNS. Some of the topics to be considered
include extension of the DNS through the addition of generic top-level domains and multilingual domain
names; introduction of new name assignment and indexing schemes (including alternate root servers);
adoption of new directory structures or services for locating information resources, services, or sites of
interest; and deployment of improved user interfaces.

The technologies that support finding information on the Internet are deployed within a complex
and contentious national and international policy context. The "right" to use a particular domain name,
like any name, can often be disputed. These disputes include conflicts among commercial claimants as
well as conflicts between non-commercial and commercial claimants. Effective solutions must consider
the potentially competing interests of domain name registrants and trademark holders; the different
interests of stakeholders including businesses, from small firms to multinational corporations;
educational, arts, and research institutions; not-for-profit charitable and service organizations; government
entities at all levels from town to nation; nation-states and international organizations; and individuals
(i.e., the general public); as well as public interests such as freedom of speech and personal privacy.
The project's report will examine the degree to which the options offered by new technology or
new uses of existing technology can mitigate concerns regarding commercial and public interests (which
will include a discussion of trademark-related issues), facilitate or impede further evolution of the
Internet, and affect steps being taken to enhance competition among domain name registrars, the
portability of Internet names, and the stability of the Internet. For each of the prospective technologies,
the final report is expected to characterize the institutions, governance structures, policies, and procedures
that should be put in place to complement it and will specify the research (if any) required to design,
develop, and implement the technology successfully. Also identified will be the options foregone or
created by particular technologies and the difficulties associated with each technological alternative.

The committee did its work through its own deliberations and by soliciting input from a number
of other experts (see Appendix B for a list of those who briefed the committee) and from the international
public through an open invitation published on the Web.2 It first met in April 2001 and six times
subsequently in plenary session. Additional information was derived from reviewing the published
literature, monitoring selected listservs and Web sites, and obtaining informal input at various
conferences and other meetings. Committee members and National Research Council (NRC) staff made

2 See The Future Of Internet Navigation And The Domain Name System, "An Invitation to Individuals Worldwide
to Provide Input to a Study Conducted by the U.S. National Academy of Sciences." Available at:

Prepublication Copy--Subject to Further Editorial Correction
several site visits, which included participation in meetings of the Internet Corporation for Assigned
Names and Numbers (ICANN), the Internet Engineering Task Force, and the World Summit on the
Information Society. Significant input was also derived from committee members during the course of
their professional activities outside of the committee's work. During the editorial phase of the study, facts
were checked for accuracy with either published sources or subject experts.
At the outset of the study, some conflict and controversy were expected, given the intense debate
about the DNS and its associated institutions such as the ICANN and the rapidly growing interest in the
use of commercially sponsored navigation services. We were not disappointed. However, the committee
was able to achieve consensus in a number of areas as described in the main text. Moreover, the
committee believes that this report represents a contribution to future discussions related to the DNS by
serving as a reference document containing much of the basic, relevant technical and institutional
background material and many of the policy alternatives in as clear and objective a manner as possible.
A number of committee members withdrew from the committee for various reasons. In a few
instances, new employment or professional opportunities raised conflict-of-interest concerns. Several
committee members were simply unable to participate in the committee's work because of increased
professional or personal obligations.

Although the report refers to several companies, products, and services by name, such reference
does not constitute an endorsement by the committee or the National Academies.


The committee appreciates the support and guidance of its sponsors. The committee's initial
contacts at the U.S. Department of Commerce were J. Beckwith Burr, Amy Page, and Karen Rose; in the
latter portion of the study, Cathy Handley and Robin Layton represented the Department. Aubrey Bush
and George Strawn were the committee's initial contacts at the National Science Foundation, with
Darleen Fisher assuming this role during the final months of the project. The committee also appreciates
the financial support of CSTB's core sponsors: the Air Force Office of Scientific Research, Cisco
Systems, the Defense Advanced Research Projects Agency, Department of Energy, Intel Corporation,
Microsoft Research, National Aeronautics and Space Administration, National Institute of Standards and
Technology, National Library of Medicine, National Science Foundation, and Office of Naval Research.
Additional financial support was provided by the National Academies.

In addition, we would like to thank those individuals who provided valuable inputs into the
committee's deliberations. Those who briefed the committee at one of our plenary meetings are listed in
Appendix B. Others who provided us with important inputs include Ronald Andruff (RNA Partners,
Inc.), Carl Bildt (AG Global Solutions and ICANN At-Large Study Committee), Mason Cole
(SnapNames), Shari Garmise (Cleveland State University), Carolyn T. Hoover (dotCoop), Cary Karp
(MuseDoma), Kalpana Shankar (University of California, Los Angeles), Paul Twomey (Internet
Corporation for Assigned Names and Numbers), Anastasia Zhadina (Robin, Blecker & Daley), and
Matthew Zook (University of Kentucky). We would also like to acknowledge those organizations that
hosted committee meetings: AOL Time Warner, Inc.; University of California, Berkeley; University of
California, Los Angeles; Harvard University; and VeriSign, Inc.

The committee appreciates the thoughtful comments received from the reviewers of this report
and the efforts of the NRC's report review coordinator and monitor. The review draft stimulated a
comparatively large volume of comments, each of which was taken into account during revision of the
draft. Many of the comments provided additional reference material, relevant anecdotes, and additional
observations to bolster or counter the committee's earlier thinking. In sum, the comments were
instrumental in helping the committee to sharpen and improve the report. However, the reviewers' are not
responsible for the report's conclusions or recommendations, with which some of them may disagree, or
for its structure and specific content. Those are solely the committee's responsibility.

Prepublication Copy--Subject to Further Editorial Correction

Finally, the committee would like to acknowledge the staff of the NRC for their work. Alan
Inouye served as the study director with overall staff responsibility for the conduct of the study and the
development of this final report. Margaret Marsh Huynh had primary responsibility for the administrative
aspects of the project such as organizing meeting logistics. Marjory Blumenthal, as director of CSTB,
and her successor, Charles Brownstein, provided the committee with valuable administrative and
technical guidance. Cynthia Patterson and Kristen Batch provided research and writing support for
various stages of the report drafting and revising process. The committee acknowledges the valuable
contribution of Susan Maurizi of the NRC's editorial staff. The committee would also like to thank Janet
Briscoe and Renee Hawkins of the CSTB staff, Liz Panos of the staff of the Division on Engineering and
Physical Sciences, and Janice Mehler of the Report Review Committee for their support of the
committee's work.

Roger Levien, Chair
Committee on Internet Navigation and the
Domain Name System

Prepublication Copy--Subject to Further Editorial Correction
Acknowledgment of Reviewers

This report has been reviewed in draft form by individuals chosen for their diverse perspectives
and technical expertise, in accordance with procedures approved by the National Research Council's
Report Review Committee. The purpose of this independent review is to provide candid and critical
comments that will assist the institution in making its published report as sound as possible and to ensure
that the report meets institutional standards for objectivity, evidence, and responsiveness to the study
charge. The review comments and draft manuscript remain confidential to protect the integrity of the
deliberative process. We wish to thank the following individuals for their review of this report:

Aristotle Balogh, VeriSign, Inc.
Timothy Bray, Textuality
J. Beckwith Burr, Wilmer, Cutler & Pickering
kc claffy, Cooperative Association for Internet Data Analysis
David Clark, Massachusetts Institute of Technology
Steve Crocker, Shinkuro, Inc.
Bruce Croft, University of Massachusetts, Amherst
Leslie Daigle, VeriSign, Inc.
Graeme Dinwoodie, Chicago-Kent College of Law
Joseph Farrell, University of California, Berkeley
Michael Froomkin, University of Miami
Hector Garcia-Molina, Stanford University
Marti Hearst, University of California, Berkeley
Randy Katz, University of California, Berkeley
Butler Lampson, Microsoft Corporation
F. Thomson Leighton, Akamai Technologies and the Massachusetts Institute of Technology
Michael Lesk, Rutgers University
Lars-Johan Liman, Autonomica
Clifford Lynch, Coalition for Networked Information
M. Stuart Lynn, Independent Consultant
Tom Mitchell, Carnegie Mellon University
Ivan Png, National University of Singapore
Fred Schneider, Cornell University
Paul Vixie,, Inc.
Tan Tin Wee, National University of Singapore

Although the reviewers listed above have provided many constructive comments and suggestions,
they were not asked to endorse the conclusions or recommendations, nor did they see the final draft of the
report before its release. The review of this report was overseen by Alexander H. Flax, independent
consultant, and Joseph Bannister, University of Southern California. Appointed by the National Research
Council, they were responsible for making certain that an independent examination of this report was
carried out in accordance with institutional procedures and that all review comments were carefully
considered. Responsibility for the final content of this report rests entirely with the authoring committee
and the institution.


Prepublication Copy--Subject to Further Editorial Correction



1.2 The Domain Name System

1.4 The Dynamics of Change
1.4.1 Increasing

1.5 Internet Naming and Navigation

1.6 Objectives of This Report

1.7 Roadmap for This Report


2.1 Origin of the Domain Name System

2.2 Designing the Domain Name System

2.2.1 Simple, Mnemonic, and Deeply Hierarchical Names

2.3 Deploying the Domain Name System

2.3.3 Convergence in Electronic Mail Systems

2.3.5 The DNS as a Production System

2.4 Continuing Growth and Evolution of the Internet As a Technical Infrastructure

2.5 Economic and Social Value of Domain Names

2.5.1 Demand for Domain Names and Emergence of a Market

2.5.2 The Rise of Conflicts Over Domain Names




2.7 Administration of Domain Names


3.1 Operation of the Domain Name System

3.1.1 A New, Remote Query

3.2 Architecture of the Domain Name System

Prepublication Copy--Subject to Further Editorial Correction

3.2.3 Programs: BIND and Others



Maintenance of DNS Standards--The Internet Engineering Task Force

Providing Root Name Server Software--Internet Software Consortium, Inc. and
Other Software Providers

3.3 Implementation--The Domain Name System Root Zone

3.3.1 Characteristics of the Root Zone




3.3.2 Technical System of the Root Zone



3.3.3 Institutional Framework of the Root Zone
Approving the Root Zone File-- U.S. Department of Commerce and ICANN
Maintaining the Root Zone File--VeriSign
Selecting the Root Name Server Operators--Self Selection
Operating the Root Name Servers--The Root Name Server Operators

3.4 Implementation--The Top-Level Domains

3.4.1 Characteristics of the TLDS
Recharacterizing TLDs

3.4.2 Technical System of the TLDs

3.4.3 Institutional Framework of the TLDs
Selecting new TLDs
Selecting the Organizations Responsible for the TLDs
Selecting the TLD Registry Operators
Operating the TLD Registries

3.5 Implementation--The Second- and Third-Level Domains

3.5.1 Technical System of the Second- and Third-Level Domains
the Second- and Third-Level Domains
Selecting the Organizations to Register Domains
Registering Domain Names
Resolving Domain Name Conflicts


4.2 Improving Security of the Domain Name System--DNSSEC

4.3 Linking the Telephone and Internet Naming System--ENUM

Prepublication Copy--Subject to Further Editorial Correction

4.3.1 Mechanics and Operations of ENUM

4.3.2 Technical and Public Policy Issues

4.4.1 Internationalizing Domain Names in Applications
Client-Side Support

4.4.3 Chinese, Japanese, and Korean Scripts

4.5 Responding to Domain Name Errors--Aggregation and Site Finder

4.5.2 Site Finder by VeriSign
Technical Issues


5.1 Governance of the Domain Name System

5.1.1 Relationship to Governance of the Internet

5.1.2 Where Should Stewardship of the DNS Reside?
Alternative A: Existing Intergovernmental Organization--International
Telecommunication Union
Alternative B: International Treaty Organization
Alternative C: Private Organization with International Participation

5.2 Management of the Domain Name System

5.2.1 Scope of ICANN's Authority

5.2.2 Composition of the ICANN Board

5.2.3 Nature of ICANN's Management Processes
Alternative A: Markle Foundation Proposal (2002)
Alternative B: Non-governmental Organization and Academic ICANN Study
Proposal (2001)
Alternative C: ICANN as Registry for the Root (2004)
Alternative D: Proposal--ICANN as a Private Trade Association (2002)
Alternative E: Center for Democracy and Technology Proposal--Narrowed Scope
with Broad Participation (2004)
Alternative F: Reformed ICANN--Narrowed Scope with Broad Participation
Summary of the Alternatives

5.3 Oversight and Operation of Root Name Servers

5.3.1 Current Situation: Diverse Autonomy


Alternative A: Funding and Regulation
Alternative B: Competitive Market
Alternative C: Distributed Root Zone File
Alternative D: DOC Relaxes MoU Requirement
Summary of the Alternatives

5.3.3 Conclusions and Recommendations

Prepublication Copy--Subject to Further Editorial Correction

5.4 Regulation of Generic Top-Level Domains

5.4.1 Should New gTLDs Be Added? If So, How Many New gTLDs and How Fast?
Technical and Operational Performance Issues
User Needs and Economic Issues

5.4.2 If New gTLDs Are to Be Added, What Types Should They Be and How Should
They and Their Operators Be Selected?
Which Types of gTLDs Should Be Added?
How Should the Operators of gTLDS Be Selected?
What Selection Process Should Be Used?

5.5 Oversight of Country Code Top-Level Domains
Alternative A: "Thick" ICANN
Alternative B: "Thin" ICANN
Alternative C: International Oversight
Alternative D: Self-governing Root Management Organization
Comparison of the Four Alternatives

5.6 Resolution of Conflicts Over Domain Names

5.6.1 Assessment of the UDRP

5.6.2 Proposed Improvements to the UDRP

5.6.3 Disputes Concerning Internationalized Domain Names

5.7 Provision and Protection of Whois Data

5.7.1 Assessment of Whois Data Issues


5.7.2 Whois and Internationalized Domain Names
5.7.3 Conclusions and Recommendations


6.1 The Nature of Internet Navigation

6.1.1 Vast and Varied Resources for Multiple Purposes

6.1.3 Complexity and Diversity of Uses, Users, and Providers

6.1.4 Lack of Human Intermediaries

6.1.5 Democratization of Information Access and Provision

6.1.6 Lack of Context or Lack of Skill

6.1.9 The Sum of the Differences

6.2 Internet Navigation Aids and Services--History

6.2.1 Aiding Navigation via the Internet

6.2.2 Aiding Navigation Through the World Wide Web

6.3 Addendum--Searching the Web Versus Searching Libraries


7.1 Navigation Aids and Services

7.1.1 Direct Access via a Uniform Resource Locator or Domain Name

7.1.2 Direct Access via Hyperlinks

Prepublication Copy--Subject to Further Editorial Correction

7.1.3 Direct Access via Bookmarks

7.1.4 Direct Access via KEYWORDS

7.1.5 Direct Access via Metadata

7.1.6 Navigation via Directory Systems

7.1.7 Navigation via Search Engines
Algorithmic Search
Monetized Search
Search Engine Marketing and Optimization
The Deep, Dark, or Invisible Web
Metasearch Engines
7.1.8 Use of Navigation Aids

7.2 Internet Navigation--Institutional Framework

7.2.1 The Commercial Providers of Navigation Services

7.2.2 The Business of Internet Navigation

7.2.3 The Navigation Services Market


8.1.1 Navigation Service Algorithms and Operations

8.1.3 Navigation to Audio and Visual Materials

8.1.4 Making Greater Use of Contextual Information



A Biographies of Committee Members, A-1
B Speakers and Participants at Meetings and Site Visits, B-1

Prepublication Copy--Subject to Further Editorial Correction

Executive Summary

Most of the Internet's users rely on the Domain Name System (DNS) and navigation aids or
services to find the resources they seek or to attract users to the resources they provide. Yet, although they
perform well, both the DNS and Internet navigation services face challenges arising from technological
change and from institutions with a wide variety of commercial, cultural, social, and political agendas.
Individually, or together, those pressures could force operational changes that would significantly reduce
access to Internet-linked resources by segments of the user community, reducing the Internet's value as a
global resource.
This document reports the conclusions of an assessment of the current state and the future
prospects of the DNS and its interactions with Internet navigation, including its uses as a means of
navigation itself and as an infrastructure for navigation by other means. The assessment is the result of the
deliberations of a committee that encompasses a wide range of disciplines, experience, and viewpoints.
The report is addressed to the technologists, policy makers, and others whose decisions will affect the
future of the DNS and Internet navigation aids and services. The specific conclusions and
recommendations of the Committee on Internet Navigation and the Domain Name System appear
throughout this summary in boldface type.


Domain names are commonly used to designate services and devices on the Internet, as a more
memorable and more permanent alternative to the numerical addresses employed by its routing
computers. They are the valued, often valuable, and often user-friendly names on the signposts that
designate many things connected to the Internet. Consequently, which names are available, who controls
their allocation, what is charged for their use, how their uses are managed, and the answers to many
related questions are important to virtually everyone who uses the Internet, whether as information seeker
or provider.

Overall, the DNS's technical system and institutional framework have performed reliably
and effectively during the two decades of the DNS's existence. The DNS has coped with the extremely
rapid expansion of Internet usage driven by the wide deployment of the World Wide Web in the 1990s
and the widespread adoption of e-mail. The hierarchical, distributed structure of the DNS technical
system, operated collaboratively by a group of mostly autonomous organizations, has proven to be
scalable, reliable, secure, and efficient.

The DNS technical system can continue to meet the needs of an expanding Internet. Early in
the committee's assessment it became apparent that it would not be fruitful to consider alternate naming
systems. As noted, the DNS operates quite well for its intended purpose and has demonstrated its ability
to scale with the growth of the Internet and to operate robustly in an open environment. Moreover,
significantly increased functionality can be achieved though applications such as navigation
systems built on the DNS, or offered independently, rather than through changing the DNS directly.
Hence, the need did not seem to be to replace the DNS, but rather to maintain and incrementally improve
it. Furthermore, given the rapidly increasing installed base and the corresponding heavy investments in
the technical system and the institutional framework, the financial cost and operational disruption of
replacing the DNS would be extremely high, if even possible at all.

However, the continued successful operation of the DNS is not assured; many forces, driven
by a variety of factors, are challenging the DNS's future. Required and desirable technologies to

Prepublication Copy--Subject to Further Editorial Correction
increase security and enable the use of non-Roman scripts for domain names are not being incorporated
into the technical system as quickly as many would like. There are persistent and substantial controversies
concerning the structure and policies of the DNS's institutional framework. Moreover, there have been
many efforts to use the DNS, because it exists and is so widely deployed, for many purposes for which it
may not be appropriate. In addition, national legislation and court decisions are addressing Internet and
domain name issues with potentially conflicting and disruptive consequences for the operation of the

Security Challenges

Like all public networked systems, the system of public domain name servers is threatened
by a variety of purposeful attacks, both malicious and mischievous, by individuals or groups that
aim to disable or divert their operations.
The operators of the DNS are responding to these threats, but
not all the desirable steps to ensure security have yet been implemented.

Denial-of-Service Attacks

Denial-of-service attacks attempt to overwhelm key name servers and their links to the Internet
with so much traffic that they are incapable of responding to legitimate queries. The root name servers
have the capacity and capability to respond to many times the normal number of queries they receive, and
have alternate connections to the network if some are blocked. Their ability to respond to attacks has been
improved by some operators' recent addition of multiple distributed copies (called "anycast" servers) of
the base name servers, increasing both capacity and connectivity. In anticipation of future denial-of-
service attacks and normal growth in demand, and to improve service globally, anycast server
deployment should be expanded.

Physical Vulnerability

Notwithstanding the deployment of anycast servers and installation of backup servers at remote
locations, there remain concentrations of base root name servers in the Washington, D.C., area and, to a
lesser extent, in the Los Angeles area. This is a system vulnerability, since anycast servers depend on the
13 critical root name servers to obtain root zone information. There should be a further physical
diversification of the location of root name servers, ideally into hardened locations.

Message Alteration

In response to the threat of alteration of messages being transmitted among name servers, the
technical community has developed DNS Security Extensions (DNSSEC), which uses digital signatures
to verify that the content of a message to or from a name server arrives unaltered and that its origin is as
stated. DNSSEC only gives assurance that what was sent was not changed during transmission; it cannot
and is not intended to assert that the message is factually correct. For example, DNSSEC has no
capability to guarantee that it is communicating the address for a given domain name. The security of the
DNS would be significantly improved if DNSSEC were widely deployed among name servers for
the root zone and top-level domains in particular, and throughout the DNS in general.


Prepublication Copy--Subject to Further Editorial Correction
Performance Monitoring

Although some steps have been taken, more could be done to continuously monitor the
performance and traffic flows of the DNS so as to enable rapid detection of and response to attacks or

Governance Challenges

The DNS works through the voluntary cooperation of its autonomous component entities.
That cooperation, in turn, depends on their tacit agreement on two principles that together enable
the Internet and the DNS to evolve and remain effective:

Universal open standards. The first principle is that the protocols and standards defining
operation of the Internet and the DNS will be open and established by the Internet Engineering Task
Force (IETF), an international voluntary organization of technical specialists. This technical framework
enables every device on the Internet to connect to and communicate with every other, and it has been
critical to the success of the Internet and the DNS. Because changes in Internet and DNS protocols,
standards, and practices are matters of consequence beyond specific Internet services, alterations to the
functions of or modifications to established standards and practices have traditionally been vetted by the
IETF before being implemented.
Innovation at the edges. The second principle is that applications should be offered by
devices on the edges of the Internet, rather than at the Internet's internal nodes or on its links. In general,
applications located at the edges have little effect on the stability of the Internet, so there is no need to
regulate them. The DNS is not, strictly speaking, internal to the Internet (the translation service is
performed by computers at the edges), but functions as though it were. It can thus be thought of as a core
service, which although not absolutely necessary, is extremely useful in giving a relatively user-friendly
face to Internet resources, and for enabling access to those resources even when their Internet addresses
change. Moreover, it is a deeply embedded and ubiquitous service that enables other services and
functions, including most aids to Internet navigation.

This tacit agreement governs the basic behavior of the many autonomous operators of the DNS,
but there is also a need for an authority to make decisions about the allocation of limited resources central
to DNS operations. The most critical of these decisions are the determination of which top-level domains
(TLDs) shall appear in the root zone file of the DNS, which organizations shall be designated as
responsible for their operation, and the terms under which those organizations shall operate.

The principal organizations that constitute this authority are, currently, the U.S. Department of
Commerce (DOC) and the Internet Corporation for Assigned Names and Numbers (ICANN), although
national bodies have considerable influence over the operations of the associated country-code top-level
domains (ccTLDs). Both the DOC and ICANN face significant challenges to their authority and
legitimacy in management of the DNS.

Stewardship of the DNS

As the Internet has become an increasingly important component of the international
infrastructure, there has been growing pressure to introduce some form of international political control
over the DNS. This pressure comes both from existing international organizations seeking authority over
the Internet or the DNS, and from individual countries that would like to end the stewardship role of the
United States.

Prepublication Copy--Subject to Further Editorial Correction

Governance of the DNS is part, but not all, of governing the Internet. Efforts to leverage it
to influence broader Internet policy are, therefore, likely to be ineffective and could also be
detrimental to the DNS.
Many of the governance issues that concern governments control of spam and
uses of the Internet for illegal purposes; resolving the disparities between developed and developing
countries in Internet usage; protection of privacy, freedom of expression, and intellectual property other
than domain names; and the facilitation and regulation of e-commerce have little or nothing to do with
the DNS per se. The DNS would not be an effective vehicle for addressing such issues. Attempts to
change the DNS or extend its management and administrative processes to do so could interfere with
reaching agreements on the already contentious issues concerning the DNS itself.

Governance of the DNS is not an appropriate venue for the playing out of national political
interests. One valued and essential quality of the DNS institutional framework has been its relative
freedom from direct pressures arising from conflicts among competing national interests and policy
agendas (apart from sovereignty-associated issues such as ccTLD delegations and redelegations).
International disputes arising in other contexts have largely been kept away from the DNS as they
should be.

For that reason: The committee does not support efforts to put the DNS directly under the
control of governments or intergovernmental agencies. In practical terms, the U.S. government, which
must agree, has not supported turning DNS stewardship over to other governments or an international
organization, although that could change. Although the 2005 U.N.-sponsored World Summit on the
Information Society (WSIS) may produce proposals for a non-governmental agent an internationally
negotiated convention or multi-stakeholder organization with oversight or other influence over the
DNS, no proposal that can be evaluated for either practicality or feasibility has yet (in December 2004)
been made.

One way to respond to concerns about the U.S. government's role as steward of the DNS is
for it to transfer its stewardship role to a non-governmental body specifically, ICANN. In the
September 2003 revision of its agreement with ICANN, the DOC stated its intent to transfer its
stewardship to ICANN if within 3 years ICANN is able to fulfill a mutually agreed set of tasks.

If ICANN does not fulfill the agreed tasks, and a proposal for creation of a non-governmental
organization having Internet governance responsibilities results from the WSIS process before the transfer
date, the DOC could consider transferring the stewardship role to the proposed organization. That would
entail comparing a not-yet-existing organization to one with 8 years of experience and evolution.
Life without the stewardship of the U.S. government will open ICANN to political and
commercial pressures. A free-standing ICANN would lack the oversight and, importantly, the protection
provided by the U.S. government's stewardship. If ICANN becomes steward of the DNS, legitimacy
based on the "consent of the governed" would be the principal basis for its continued authority and its
ability to resist inappropriate pressure from governments and other powerful interests. Final responsibility
for satisfying the needs of its constituencies in an equitable, open, and efficient manner would lie with its
Before completing the transfer of its stewardship to ICANN (or any other organization), the
Department of Commerce should seek ways to protect that organization from undue commercial or
governmental pressures
and to provide some form of oversight of performance.

Legitimacy of ICANN

ICANN is a work in progress; its long-term success is not assured. After a troubled start, it has
introduced several innovations to the institutional framework of the DNS, including competition among
registrars and an arbitral process for resolving disputes over domain names, the Uniform Domain Name
Dispute Resolution Policy (UDRP). In 2003 it had to undertake a major reform of its own organization,
stimulated by dissatisfaction with its operation under its initial structure. It is working on the revision of

Prepublication Copy--Subject to Further Editorial Correction
key decision processes in response to complaints about their lack of transparency and fairness.
Furthermore, ICANN has been unable to conclude formal agreements with many of the organizations
critical to its responsibilities, notably the root name server operators and the vast majority of the ccTLD
registries. Nevertheless, through its responsibility for recommending changes in the root zone file, which
defines which TLDs are in the DNS and where their operators are located on the Internet, ICANN has
been able to exercise authority over the coherence and stability of the DNS.
Since its beginning, ICANN has been the subject of controversy and contention flowing from the
many diverse constituencies that have been attracted to it and their correspondingly diverse views. The
critics' concerns have been with ICANN's scope, its organizational structure, and its management
processes. The concern about scope has been principally the extent to which ICANN has exceeded its
technical-administrative responsibilities, for example, to regulate TLD registry operations; but others
have been disappointed by its unwillingness to take on broader issues. The structural concerns have
included perceptions of imbalance in the historical composition of ICANN's board, of failings in the
board selection processes, and of inadequate representation of certain constituency groups. The process
concerns have been the perceived lack of transparency, effectiveness, accountability, and recourse in
ICANN's electoral and decision processes.
ICANN is more likely to improve its perceived legitimacy by narrowing its scope and by
improving its processes rather than by seeking an ideally representative composition of its board.
No composition of its board is likely by itself to confer the perception of legitimacy on ICANN
among its constituency groups.
A narrowing of scope and improvement of processes are elements of the
path that ICANN claims to be following in carrying out its 2003 reform. However successful its reform,
ICANN faces the challenge of reaching an effective modus operandi with three critical sets of participants
in the DNS's institutional framework: the root name server operators, the generic TLD registries, and the
ccTLD registries.

Root Name Server Operators

No greater oversight of the root name server operators will be necessary so long as they
continue to operate effectively and reliably and to improve the DNS's security, stability, and
The effective daily operation of the root, and therefore the DNS, lies in the hands of the
operators of the 13 critical root name servers. They have provided reliable and efficient service as the
Internet has undergone rapid growth in the numbers of its users and providers. Although the DOC has
assigned ICANN responsibility for the stability and security of the root name server system, ICANN's
authority has not been sufficient for it to manage or regulate the root name server operators directly, nor is
it clear that doing so is desirable or necessary. The real challenge to ICANN is to identify how it can best
ensure the stability and security of the root name server system, given the long-standing autonomy of the
operators and the effectiveness of their operations.
More formal coordination of the root name server operators is desirable in the longer term.
ICANN is currently the most appropriate organization to assume the coordination role. Although
direct management or oversight may be neither necessary nor feasible, with continued growth in the
Internet and demands on the DNS, a more formal process of coordination of the root name server
operators with ICANN's facilitation will become desirable so as to ensure rapid response to persistent
security needs and to other challenges.
The present independent funding arrangements for the root name servers are advantageous
and should continue, because the multiplicity of sources contributes to the resilience, autonomy,
and diversity of the root name server system.
The root name server operators do not receive direct
compensation for the service they perform. While running a root server may only add an incremental cost
in the range of tens of thousands of dollars for an organization already operating a secure Internet site,
fully loaded costs have been estimated at up to $1 million or more depending on numerous factors
including the number of locations, bandwidth requirements, and staffing levels. The costs are covered by

Prepublication Copy--Subject to Further Editorial Correction
each organization as part of other operations. Although a central source of funds to compensate all the
root name server operators for their services might appear desirable, it is likely to be accompanied by an
unacceptable regulatory or control role for the funding organization and would reduce the robustness of
the current arrangement.
ICANN should work with the root name server operators to establish a formal process for
replacing operators. that directly engages the remaining root name server operators. Under the
process, ICANN would be responsible for the final decision on the basis of recommendations from
the root name server operators.
One or more of the current root name server operators may withdraw
for organizational or performance reasons, and it would be reasonable to have in place an agreed process
to deal with such eventualities.

Generic Top-Level Domain Registries

A major challenge to ICANN since its founding has been deciding whether, when, and how to
add generic top-level domains (gTLDs) and, if any are added, how many. It has faced strong pressures
both to add gTLDs and to stop or, at least moderate, the pace of such additions. The committee addressed
the issue of gTLD addition broadly in terms of both effects and constituencies affected, but for simplicity
the multidimensional arguments for and against new gTLDs were clustered into two groups: technical and
operational performance, and user needs and economic benefits.
Considering technical and operational performance alone, the addition of tens of gTLDs per
year for several years poses minimal risk to the stability of the root. However, an abrupt increase
(significantly beyond this rate) in the number of gTLDs could have technical, operational, economic, and
service consequences that could affect domain name registrants, registries, registrars, and Internet users.
From the standpoint of user needs and economic benefits, neither the arguments in favor of
nor those against additional gTLDs are conclusive. Thus, the decision to add gTLDs is one requiring
judgment and cannot be determined by formal analysis alone.
If new gTLDs are added, they should be added on a regular schedule that establishes the
maximum number of gTLDs (on the order of tens per year) that could be added each time and the
interval between additions.
Addition of gTLDs should be carried out cautiously and predictably, so that
on the one hand, the stability and reliability of the system can be protected, and on the other hand, those
considering acquiring a gTLD can do so with a realistic view of future prospects.
A mechanism to suspend the addition of gTLDs in the event that severe technical or
operational problems arise should accompany a schedule of additions. It should explicitly specify
who has the authority to suspend additions and under what conditions.

A neutral, disinterested party should conduct an evaluation of new gTLDs approximately 1
or 2 years after each set of new gTLDs is operational to make recommendations for improving the
process for selecting and adding gTLDs.

If new gTLDs are to be created, the currently employed comparative hearing or expert
evaluation processes should not be assumed to be the only processes for selecting their operators. In
its addition of gTLDs in 2000, ICANN used a comparative hearing process to select 7 from the 44
applicants. In its 2004 addition of sponsored gTLDs, ICANN used a non-competitive process that
replaces subjective judgments by its staff and board with judgments by expert groups that are insulated
from lobbying, but whose decision-making processes are not transparent. By doing so, it has reduced a
few of the potential sources of dissatisfaction with the resultant selections compared with the process used
in 2000. However, the question remains as to whether it is necessary for ICANN to qualify new gTLDs,
as this process does, on such matters as sponsorship by a community, business and financial plans, and
addition of new value to the name space.
For creation of new gTLDs, ICANN should consider alternate processes that are less reliant
on expert, staff, or board judgments. One such approach would be pre-qualification of applicants only
on technical capability, basic financial viability, and adherence to registrant protection standards and

Prepublication Copy--Subject to Further Editorial Correction
ICANN policies. (ICANN should establish requirements to minimize the dangers of domain name
registrants losing their service and the value invested in their domains if a registry fails and should
carefully consider possible side effects.) If the number of qualified applicants turns out to be less than the
number of available slots, all would be chosen; if not, a market-based selection process an
auction could be used to select among them. Because of the wide range of intents and corresponding
designs of such processes, they must be carefully designed, drawing on the wide range of previous
experience in the design of auctions.

Country-Code Top-Level Domain Registries

Resolution of ICANN's role vis-á-vis the ccTLDs is one of the critical challenges to
establishing an ICANN that is viewed as a legitimate and appropriate steward for the DNS.
Although the ccTLDs represent 243 of the 258 TLDs, ICANN had formal agreements with only a dozen
of the ccTLD operators as of December 2004.

The ccTLDs as a group now operate only partially under the oversight of any higher authority,
ICANN or government. A number of ccTLDs are overseen by their national governments; some have
established non-governmental bodies to represent the local Internet community and exercise varying
degrees of oversight; some are completely autonomous non-profit bodies that operate voluntarily to meet
local Internet community interests; and some are commercial bodies with some contractual linkage to the
national government.

The only body that currently has an opportunity to exercise oversight over all the ccTLDs is
ICANN. The principal way in which it exercises that authority is through recommendations to the DOC
about which organization should be delegated responsibility for a specific ccTLD. Yet this issue arises
only when the present delegatee resigns or is challenged or a new ccTLD is established.

The relationship between ccTLDs and ICANN has been difficult from the beginning of ICANN.
First, a large number of the ccTLDs felt no need to contribute to ICANN's budget, since they did not
think that they received any corresponding benefits. Second, many ccTLDs resented ICANN's major role
in deciding on delegations and redelegations--essentially a policy role that they felt would be better
performed locally. They also believed that their position as one constituency within ICANN's initial
Domain Names Supporting Organization, whose other constituencies primarily addressed gTLD issues,
did not adequately reflect their importance.

Under its 2003 reorganization, ICANN attempted to respond to their concerns by replacing the
Domain Names Supporting Organization with two organizations, the Generic Names Supporting
Organization (GNSO) and the Country-Code Names Supporting Organization (ccNSO). ICANN intends
thereby to draw the ccTLDs more actively into its operations and build a stronger basis for their support.
Furthermore, ICANN's Governmental Advisory Committee was (in August 2004) working on a revision
of its Principles for the Delegation and Administration of Country Code Top Level Domains to address
many of the concerns expressed about them.

If the creation of the ccNSO does not result in increased participation by the ccTLDs in
ICANN policy making, then ICANN may find itself subject to increasing pressures to constrain its
role to that of gTLD management and root zone file record keeping and to turn ccTLD oversight
over to some other organization.
The success of the ccNSO will depend on its ability to attract an
increasing number of members, both from the large ccTLDs that are needed for financial and other
support of ICANN and the smaller ccTLDs that can benefit from the support that ICANN could offer
them. Even more critical is the refinement of the principles and processes for delegation and redelegation
of ccTLD registries and their acceptance by most of the ccTLDs.


Prepublication Copy--Subject to Further Editorial Correction

Commercial Challenges

Perhaps the most subtle, but still significant, challenge that the DNS faces is that arising from the
imperative faced by commercial operators of DNS elements they must strive to increase their revenues
and profits in the face of competition. On the Internet, increasing revenues generally means increasing
traffic to one's service, sometimes by diverting it from another operator's service. This imperative raises
the temptation to seek traffic and revenue by breaking or bending the fabric of tacit agreements that
underlies the success of the Internet and the DNS.

ICANN should strengthen its contracts with TLD operators (especially the largest ones) to
ensure that it has the authority to review proposed changes in their services that could have a
detrimental effect on the DNS or on other services that depend on the DNS. It should establish an
open, transparent, and speedy process of review for such changes that solicits contributions from
the technical community, other DNS operators, other affected Internet operations, and end users.
recent case in point is the unanticipated and unannounced introduction by VeriSign, a commercial
registry, of a service, called Site Finder, that altered the conventional response to erroneous queries to the
.com and .net TLDs, by returning pointers to its own search page, rather than sending back an error
message. After being called on by ICANN to suspend the service, VeriSign did so under protest and is
currently seeking relief in the courts.

TLDs and other DNS operators that do not have agreements with ICANN should
voluntarily agree to adhere to published technical standards and to consult the technical
community and conduct public review processes before introducing new services that could have a
detrimental effect on the DNS or on other services that depend on the DNS.

Dispute Resolution Challenges

Arbitral domain name dispute resolution processes, rather than national courts, should
continue to be encouraged as the initial and primary vehicle for resolving most disputes associated
with the rights to domain names.
The Uniform Domain Name Dispute Resolution Policy was
implemented by ICANN in December 1999 and has been adopted by all registrars in nine of the generic
top-level domains, as well as voluntarily by managers of several ccTLDs. In addition, managers of other
ccTLDs have adopted their own policies based on modified versions of the UDRP.
The UDRP has generally satisfied the need for an effective and cost-efficient means of
resolving disputes concerning domain names; however, it has weaknesses that should be addressed.
The UDRP has both positive and negative aspects, which differ, however, depending on whether they are
being considered from the perspective of the complainants or of the respondents. Although many
observers believe that the UDRP has enabled speedy and fair resolution of domain disputes, others
believe that the current system is biased toward the interests of trademark holders. Notwithstanding its
perceived disadvantages, more than 9000 decisions concerning over 15,000 domain names have been
rendered under the UDRP.
The feasibility and desirability of five specific UDRP improvements should be further
considered by ICANN:

Improving consistent use of arbitral precedents to enable similar issues to be addressed in
a more consistent manner that also supports case-by-case knowledge building;
Establishing an internal appeals process that would review the small number of decisions
that are clearly faulty or that cover a situation or issue for which competing bodies of precedent exist;

Prepublication Copy--Subject to Further Editorial Correction
Using three-member panels. Some analyses of UDRP proceedings indicated a significant
difference in outcomes depending on whether they were heard by one-member or three-member panels:
three-member panels found for the complainant in a smaller percentage of the cases;
Improving panelist knowledge about the technology underlying the DNS, the uses of
domain names (beyond Web sites), and the application of the policies and rules applicable to domain
name disputes; and
Improving the nature and structure of incentives in the process. Under the current
funding structure, the revenue for panelists depends on the volume of cases, creating incentives either for
haste or for marketing strategies and tactics to attract cases by defining lucrative niches.

Internationalization Challenges

Continuing and increased attention to internationalized domain names (IDNs) is necessary.
Efforts to coordinate work across different countries, regions, and language groups should be
undertaken to prevent the balkanization of the Internet.
Of particular interest in many countries is
access to the Internet and the DNS using home-country languages and scripts. Unfortunately, the design
of the DNS, as well as the general nature of multiscript environments, presents formidable technical and
linguistic challenges for the accommodation of languages that use non-Roman characters, which require
compromises for their solution.
Some experts have argued for a major overhaul of the Internet's infrastructure to incorporate
IDNs. However, pressure to act quickly reduced support for solutions that would require extensive
changes in architectures or standards; the result was an effort led by the IETF that culminated in the
Internationalizing Domain Names in Applications (IDNA) mechanism.1 The central goal of the IDNA
scheme is to enable end-user viewing of IDNs without altering the DNS protocols themselves, using a
client-side set of procedures, implemented at the edge of the DNS.
However, the IDNA mechanism solved only part of the internationalization problem. Remaining
to be addressed are the questions of potential consumer confusion; conflict avoidance or resolution for
similar-appearing names; differences in interpretations for different languages; restrictions on
registrations on a per-domain basis; implications for the UDRP and the Whois database (of information
about domain name registrants); security issues raised by IDN; and the implications of (and alternatives
to) multilingual top-level domains.


In contrast to the unique role played by the DNS, navigation through the Internet is not supported
by a unique integrated technical system. Among the many ways to navigate the Internet, only two involve
dedicated technical systems--search engines and directories. Moreover, the institutional framework of
those technical systems is an open market, with many, generally commercial, competitors offering
navigation services, and specialized non-commercial services focused on non-profit resource providers
and seekers.
Finding and accessing a desired resource via the Internet poses challenges that are
substantially different from the challenges in navigating to resources in non-digital, non-networked

A wide range of navigation aids and services now permit large segments of the Internet,
particularly the World Wide Web, to be traversed rapidly and efficiently in ways previously
unimaginable. They offer users across the globe convenient access to much human knowledge and

1 IDNA is described in: Patrik Fältström, Paul Hoffman, and Adam M. Costello. 2003. "Internationalizing Domain
Names in Applications (IDNA)." RFC 3490. March. Available at <>.

Prepublication Copy--Subject to Further Editorial Correction
experience and open an international audience to purveyors of content and services, no matter
where they may be located.

Use of Navigation Aids and Services

Surveys indicate a high level of satisfaction with navigation aids and services at present.
An analysis of navigation behavior, based on survey data from March 2003,2 indicates that
Internet users tend to use preferred sites and services consistently, visiting them repeatedly, using their
bookmarks or remembered Uniform Resource Locators (URLs). Search engines produced only 13 percent
of site referrals, navigation through entry of a known or guessed URL or use of a bookmark produced 66
percent of referrals, and flow along hyperlinks produced 21 percent.
According to a recent survey,3 residents of the United States conducted 3.9 billion searches in
June 2004, an average of 33 searches per user. Search engines have been used by 84 percent of U.S.
residents who use the Internet--more than 107 million people; on an average day, about 38 million of the
64 million U.S. residents who are online use a search engine. Using search engines is second only to
using e-mail as the most popular Internet activity, except when major news stories are breaking. A vast
majority of searchers say that they find the information they want most of the time, and more than two-
thirds consider search engines a fair and unbiased source of information. But only a third of searchers say
they could not live without search engines; about half say that, although they like using search engines,
they could go back to other ways of finding information.
As the material accessible through the Internet continues its rapid increase in volume and
variety and as its societal importance grows, Internet navigation aids and services are likely to be
challenged to deliver more precise responses, in more convenient forms, to more diverse questions,
from more users with widely varying skills.
Efforts to improve the basic algorithms and operations of
Internet navigation services will continue because of competitive pressures, evolving user requirements,
and technological advances. Among the specific areas where improvements are needed are query
interfaces and results displays for desktop, portable, and collaborative devices; navigation of audio and
visual materials; management of the navigation process; use of contextual information (while protecting
privacy); and understanding the wide range of navigation behaviors of the highly diverse users who now
seek resources on the Internet.
As the Internet has become the sole or most accessible location of many valuable resources, the
importance has grown of ensuring that they will persist indefinitely at the same URL (or in an archive on
the same site) or, alternatively, that they will be preserved at another site where they can be readily found.
Ensuring persistence is primarily the responsibility of resource providers, while third parties national
libraries or private organizations such as the Internet Archive--are undertaking some preservation efforts.

Although commercial services can be expected to support substantial research and development
on these topics, academic research and development activities have provided the innovative basic
technologies for many successful navigation aids and services. Public support of such academic
research and development efforts should be continued.

2 The data were collected on March 6, 2003, by WebSideStory's StatMarket from about 12 million visitors to
125,000 sites using its proprietary analytical platform and compared with figures from the previous year. Reported
in: Brian Morrissey. 2003. "Search Guiding More Web Activity." CyberAtlas. March 13. Available at
3 See Deborah Fallows, Lee Rainie, and Graham Mudd. 2004. "The Popularity and Importance of Search Engines."
Data Memo. Pew Internet & American Life Project. August. Available at
<>. The results came both from a telephone
survey of 1399 Internet users and from tracking of Internet use by comScore Media Metrix.

Prepublication Copy--Subject to Further Editorial Correction
Commercial Navigation Services

The Internet navigation services industry has financed the development and evolution of
services that meet many of the needs of a wide range of searchers at little or no cost to them,
especially when they are seeking commercial material. At the same time, it has provided advertisers
with an efficient, cost-effective means to gain access to potential customers at the time that they are
most interested in the advertiser's product or service.
The primary source of income for commercial
Internet navigation services is selling advertising linked to search queries. Consequently, as for many
broadcast media, it is the content and service providers that are subsidizing users' access to navigation
services so that they can present advertisements to them at the time of their expressed interest in a topic.

The major search services currently identify the results whose presentation in response to
specified search terms is paid for by advertisers (so-called sponsored links or sponsored search listings)
and set them off from the direct results of more neutral search algorithms. As long as the distinction is
clear and users are aware of it, sponsored search listings should present few problems while providing the
great benefit of free search services to the user.

The potential for abuse exists, however. It would be possible, for example, for a search
service to accept payment for assured placement in the "top 10" of what would appear to be a
neutral listing.
Should abuses grow, search services could find themselves under increased public
pressure for government scrutiny or facing more disputes and criticism concerning such activities
from other commercial entities.
None of the navigation services have been accused of accepting
payment for highly ranked inclusion of particular responses to queries, but some have accepted payment
to ensure inclusion, but not ranking, in the otherwise neutral listing. Furthermore, the distinct placement
and typography of the sponsored listing could be weakened to the point that a casual user would not be
aware of its difference from the neutral algorithmic search results. Thus far, competition among services,
third-party evaluations, and the perceived value to the user of search transparency have served as
important forces constraining misbehavior of these kinds.

Although competition and the desire to be seen as useful by searchers are incentives for fair
and open behavior, appropriate regulatory agencies of the U.S. federal government and of other
governments should pay careful and continuing attention to the result ranking and display
practices of Internet navigation services and their advertisers to ensure that information can flow
freely and that those critical practices are fully disclosed.
The behavior of commercial navigation
services can have a substantial influence on the kind, quality, and appropriateness of the information that
Internet users receive. Although there is no evidence that abuse has yet occurred, the potential for abuse is
inherent in the navigation services' ability to affect users' access to information for commercial or other
In the future, competition among general navigation services is more likely to take the form
of rivalry among a small number of established large players rather than competition with a large
number of small newcomers.
Over the past 4 years, there has been considerable consolidation in the
general search services market, which reflects the increasing importance of economies of scale--the
considerable hardware and software costs of developing and operating a search engine are independent of
the number of users, whereas revenues from advertising are directly dependent on them. The result is that
the barriers to entry are high, and only a company with substantial financial resources and technical skills,
such as Microsoft or IBM, is in a position to introduce its own competitive general navigation service, as
Microsoft began to do in 2004.

Innovation, Competition, and Regulation

The importance of the Internet as the infrastructure linking a growing worldwide audience

with an expanding array of resources means that improving Internet navigation will remain a
profitable goal for commercial developers and a challenging and socially valuable objective for


Prepublication Copy--Subject to Further Editorial Correction
academic researchers. Consolidation of navigation services makes it difficult for innovative services to
start small and build volume over time unless they have a very large amount of patient investment capital.
But, so long as no single service becomes dominant, each of the major competitors will face continuing
pressure to improve its offerings either through internal innovation or through the acquisition of
innovative small companies, paths they are currently actively pursuing.

Since competition in the market for Internet navigation services promotes innovation,
supports consumer choice, and prevents undue control over the location of and access to the diverse
resources available via the Internet, public policies should support the competitive marketplace
that has emerged and avoid actions that damage it.

Potential rulings in some jurisdictions could substantially reduce the ability of search
engines to sell keywords using the current automated methods. As with the Domain Name System,
the most contentious intellectual property issues affecting navigation services concern trademarks,
specifically the sale of trademarked terms to advertisers as keywords whose use will bring up their
advertisements. Since there is no arbitral process, such as the UDRP, by which such disputes could be
resolved outside the courts and with worldwide effect, it seems likely that conflicting court decisions in
different jurisdictions, worldwide, will establish the potentially conflicting rules by which navigation
services will have to abide.


The preservation of a stable, reliable, and effective Domain Name System will remain
crucial both to effective Internet navigation and to the operation of the Internet and most of the
applications that it supports.

Despite the differences in the way in which they developed, the relationship between the DNS
technical system and Internet navigation aids and services is strong and fundamental the DNS has
served as the stable core on which the incremental evolution of the different navigation aids and services
has depended. The development of navigation services is likely to continue to relieve some of the
commercial pressures on the DNS as users become increasingly comfortable with using them as their
primary means to navigate the Internet, but both the Domain Name System and Internet navigation aids
and services will be significant elements of the Internet for the foreseeable future.
The demonstrated success of the DNS and navigation aids and services in meeting the basic
needs of all Internet users should not be jeopardized by efforts to constrain or direct their evolution
outside the open architecture of the Internet, or to use them to enable control of the free flow of
information across the Internet.

The governance and administration of the DNS should not become a vehicle for addressing
political, legal, or economic issues beyond those of the DNS itself.


Prepublication Copy--Subject to Further Editorial Correction

The Internet is rapidly becoming everybody's neighborhood. Just a few keystrokes take us to an
online bookstore; several mouse clicks deliver us to an online newsstand; only a bit more effort connects
us with distant friends or family. For many of us, seeking out a Web site or an e-mail address is almost as
important as finding the way to the library, a theater, the nearest mall, the bookstore, or the neighborhood
playground. And for those who use the Internet to deliver products or services, their clientele's ability to
find them is essential to their success. Navigating the virtual neighborhood has become a life skill for
those needing something on the Internet and a life-or-death matter for businesses with something to offer
on the Internet.
To navigate--to follow a course to a goal--across any space, a method is needed for designating
locations in that space. On a topographic map, each location is designated by a combination of a latitude
and a longitude. In the telephone system, a telephone number designates each location. On a street map,
locations are designated by street addresses. Just like a physical neighborhood, the virtual neighborhood
has addresses--32 or 128 bit numbers, called Internet Protocol (IP) addresses--that define the specific
location of every device on the Internet. And also like the physical world, the virtual world has names--
called domain names, which are generally more easily remembered and informative than the addresses
that are attached to most devices--that serve as unchanging identifiers of those devices even when their
specific addresses are changed. The use of domain names on the Internet relies on a system of servers--
called name servers--that translate the user-friendly domain names into the corresponding IP addresses.
This system of addresses and names linked by name servers establishes the signposts in cyberspace and
serves as the basic infrastructure supporting navigation across the Internet. It is called the Domain Name
System (DNS).
This report is concerned with the Domain Name System and its interactions with Internet
navigation, including its uses as a means of navigation itself and as an infrastructure for navigation by
other means. Since the World Wide Web is the application running on the Internet that contains the
greatest number of locations to which most users want to navigate, this report often draws examples from
the Web. However, there are other applications that use the Internet, not least e-mail, and others that are
being developed for it. The DNS supports most of them. Unless otherwise specified, the information in
this report, its conclusions, and its recommendations apply to the DNS in its role as a basic infrastructure
element of the entire Internet, not just of the World Wide Web.
The report's specific objectives and how it is organized to address them are spelled out in this
chapter, which begins with an introduction to the Internet, the Domain Name System, and Internet
navigation, and with an examination of the forces affecting them. Four basic concepts that are used
throughout this report--names, navigation, technical system, and institutional framework--are defined
and briefly described in Box 1.1.

The Internet, according to the National Research Council, is "a diverse set of independent networks,
interlinked to provide its users with the appearance of a single, uniform network . . . . The networks that
compose the Internet share a common architecture (how the components of the networks interrelate) and
software protocols (standards governing the interchange of data) that enable communication within and
among the constituent networks." 1

1Computer Science and Telecommunications Board, National Research Council. 2001. The Internet's Coming of
. Washington, D.C.: National Academy Press, p. 29.

Prepublication Copy--Subject to Further Editorial Correction
Internally, the Internet comprises two types of elements: communication links, channels over
which data travel from point to point; and routers, computers at the network's nodes that direct data
arriving along incoming links to outgoing links that will take them toward their destinations. Altogether,
the Internet is a complex network of routers and links, the latter varying in transmission medium
(telephone lines, cable lines, optical fiber cable, satellite, wireless); servers and other hosts; and access
equipment. Links in the network may be characterized by their transmission capacity (low-capacity local
lines to very high capacity "backbone" cables) and by their latency (short-latency local fiber links to long-
travel-time satellite links). Data travel along the Internet in packets adhering to the standard Transmission
Control Protocol/Internet Protocol (TCP/IP) that defines the packets' format and header information.

BOX 1.1
Four Basic Concepts

Names and Naming Systems
Names and naming systems are everywhere in society. A license plate on a car, a serial number
on a product, and a stock market symbol for a company are a few examples of names that are used within
formal naming systems. Each of these examples is a unique identifier, created according to the
specifications of a naming system, which is associated through strict naming rules with a single
automobile, product, or company. Equally important are a host of informal or less strict naming systems,
such as the naming of people by families, the naming of streets or locations, or the naming of files and
directories on a personal computer. In these processes, there is no guarantee of unique names for objects.
More generally, naming is used to distinguish individual objects within a broad class, the object
space. The set of allowable names for objects in that class is called the name space. A name is then a
member of that set used to differentiate one member of the object class from another. A naming system is
the combination of an object space, the name space that is applied to it, the rules governing the
assignment of names to objects, the files recording the assignments, and the administrative processes (if
any) applying the rules and maintaining the files.

Navigation, Navigation Aids, and Navigation Services
Navigation is the process of following a course from one place to another. In the narrow sense,
the term is used to refer to a person or entity (e.g., a vehicle) being directed along a course from an origin
to a specified destination. In the broader sense, navigation refers to following a course on, across, or
through (e.g., navigate a stream) or making one's way somewhere (e.g., Lewis and Clark navigating to the
"western passage" or Dr. Livingstone navigating to the source of the Nile). Navigation involves a set of
skills (e.g., reading a compass, using a search engine). The place to which one wishes to navigate may be
known explicitly (e.g., latitude and longitude, a street address, a Web site address) or only in general
terms (e.g., source of the Nile, sites with information about veteran's benefits). In the former case,
navigation requires only the two steps of laying out a route to the known location and following it. In the
latter case, however, there is a prior step of identifying the desired location (or locations) through a search
process of some form.
A navigation aid is anything that assists navigation, such as a map or a compass. In the physical
world, navigation aids include a sextant and precision clock, and a compass and topographical map. In
document-oriented environments, navigation aids include printed directories for the telephone system and
online or card catalogs for library collections. Human intermediaries also can serve as navigation aids in
these environments, such as directory assistance operators in the telephone system and reference
librarians for library services. In the Internet, navigation aids include bookmarks and lists of favorites,
hyperlinks, and restricted keyword systems, such as AOL keywords.
A navigation service is a navigation aid that is based on a complex technical system (see below).
In the physical world, the Global Positioning System is a navigation service, as is an inertial navigation
system. Automated directory assistance and online white pages are navigation services for the telephone

Prepublication Copy--Subject to Further Editorial Correction
network, and online card catalogs are navigation services for libraries. In the Internet, directories and
search engines are navigation services.
Internet navigation, navigation aids, and navigation services are discussed in greater detail in
Chapters 6, 7, and 8.

Technical Systems
A technical system is an integrated set of engineered elements (components and practices) that
delivers a specific service to users. Some familiar examples of technical systems are the telephone
system, the air transport system, the electric power system, and the Domain Name System. In the case of
the telephone system, the engineered elements are organized in a complex network that includes the
switching facilities (both hardware and software) that set up the circuits linking telephones for a call, the
transmission lines that carry the calls, and the telephone instruments that originate and receive calls; the
service it delivers is telephone connectivity; and the users are people who want to communicate with
others by voice, data, or facsimile.
The single technical system that is the Domain Name System is described and discussed in
Chapters 2, 3, and 4; the technical systems that support Internet navigation services are characterized in
Chapters 6, 7, and 8.

Institutional Framework
An institutional framework is a collection of organizations and policies whose decisions and
actions enable a technical system to be constructed, operated, controlled, regulated, and improved. An
institutional framework and its technical system are complementary to each other. Each of the examples
of technical systems described above has a complementary institutional framework. The telephone
system, for example, depends for its effective and efficient development and operation upon a
complementary framework comprising equipment suppliers, operating companies, local, state, national,
and international regulatory bodies, international standards organizations, and the technical community.
The institutional framework of the DNS is discussed in Chapters 2, 3, and 5; that of Internet
navigation services, in Chapters 6, 7, and 8,

Each router uses the origin and destination IP addresses in each arriving packet to determine which link to
direct it along. A message from a sender to a receiver may be broken into multiple packets, each of which
may follow a different path through the Internet. Information in the packets' headers enables the message
to be restored to its proper order at its destination.
The origins and destinations of data transiting the Internet are computers (or other digital devices)
located at its "edges." They are, typically, connected to the Internet through an Internet service provider
(ISP) that handles the necessary technical and financial arrangements. A distinctive feature of the Internet
is that all the user services (such as e-mail or the World Wide Web) accessible through it are provided by
applications running on computers located at its edges. The "center" of the Internet--its links and
routers--provides the critical connectivity among them. As a consequence of this architecture, most of
the service innovation takes place at the edges, completely independently of the network itself. It is an
embodiment of the end-to-end argument in systems design2 that says that "the network should provide a
very basic level of service--data transport--and that the intelligence--the information processing needed
to provide applications--should be located close in or close to the devices attached to the edge of the

2 See J.H. Saltzer, D.P. Reed, and D.D. Clark. 1984. "End-to-End Arguments in System Design," ACM Transactions
on Computer Systems
3 CSTB, NRC, The Internet's Coming of Age, 2001, p. 36.

Prepublication Copy--Subject to Further Editorial Correction
A consequence of this architecture is that innovation at the edges is eased and facilitated,
requiring no coordination with network architects or operators, as long as the basic protocols are adhered
to. Conversely, innovation at the center of the network is difficult and often slow, since it requires the
cooperation of many providers and users. Because of the higher potential for inadvertent disruption as a
side effect of a change at the center of the system, difficult and time-consuming effort must be devoted to
testing and validating each proposed change. All such changes have been subject to cooperative
collaboration and agreement by the Internet engineering community since the earliest days of research
and implementation. (See "Maintenance of DNS Standards" in Section 3.2.5.) That principle has been a
major factor in the successful design, development, and implementation of the technology.
The Internet's architecture has enabled it to respond very successfully to the challenges of growth
in the number of its users and in the capacity of its links and the complexity of their connectivity, as well
as to provide a robust base for the growth of services such as e-mail and the World Wide Web.


The DNS was put into place by technologists in the early 1980s when the Internet provided basic
non-commercial services to a small community of specialists.4 The World Wide Web had not yet been
invented. The DNS's designers intended it to be a simple and stable way for users and applications to
identify computers on the Internet. They gave it a hierarchical structure so that the responsibility for
maintaining the necessary information tying domain names to IP addresses (and other data) could be
distributed to the organizations actually managing the relevant networks and groups of hosts across the
edges of the network.5 They designed it as an inverted tree with the expectation that most domain names
would lie several branches down, requiring relatively few names in the upper part of the tree. Figure 1.1
illustrates the Domain Name System's role in support of navigation across the Internet.6 Complete
domain names7 incorporate the names of the nodes in the tree above them. So in Figure 1.1, the domain
name designates the www leaf lying on the .cstb branch, which lies on the .nas
branch, which lies on the .edu branch.
A number of factors, including the introduction of the World Wide Web in the early 1990s, have
transformed the Internet community from a small town into a great and rapidly expanding metropolis with
an extremely large, highly diverse body of users, relatively few of whom are computer specialists. These
users employ the Internet as the communications backbone for a vast range of commercial and non-
commercial purposes. As a result, the Internet has expanded both in scale and in the scope of its
applications. Because of the elegance of its technical design, the DNS has, thus far, been able to adjust to
the expanded scale of the Internet, evolving to meet the increased operational demand adequately.
However, as a consequence of the growth in the scope of the Internet, the DNS is now used in ways that
were not anticipated when it was designed. These unanticipated uses have led, in turn, to a substantial
increase in the number and complexity of the institutions responsible for its operation and management
and to less use than was originally expected of deep naming hierarchies and distributed, but localized,
management of names. This growth in institutional complexity has been driven primarily by the fact that
domain names acquired increased value, which required mechanisms to deal with their allocation and
ownership. Their acquisition of increased value followed from four developments:

4 Chapter 2 describes the development of the Domain Name System from the early 1980s to the present.
5 Chapter 3 describes the design and operation of the Domain Name System.
6 This depiction of the DNS is highly simplified. More detailed descriptions of the DNS are provided in Chapters 2
and 3.
7 Formally, these are called "fully qualified domain names" to distinguish them from partial domain names that
describe the path only from some node below the root.

Prepublication Copy--Subject to Further Editorial Correction
Preference for short and memorable names. The first development was a preference among users
for short and memorable domain names, which led to an unexpectedly unbalanced distribution of domain
names. More devices were named at the second or third level of the inverted tree, thereby widening it,
rather than deepening it to the fourth and lower levels as had been anticipated. And although the
implementation of the DNS offered a range of generic top-level domains--such as .com, .net,
.org, .gov, and .edu--for a variety of reasons, .com became the preferred choice, further
unbalancing the tree. Even with the addition of new generic top-level domains after the year 2000,
memorable domain names within the .com domain remain the preferred choice for most businesses,
many non-profit organizations, and numerous individuals.8
Navigation role of .com. What led most directly to the growth in importance of the .com
domain was the second development--its self-fulfilling role in navigation. For example, as commercial
uses grew, users seeking IBM's site on the World Wide Web could guess with an
expectation of success. That common behavior naturally led organizations and individuals to seek
registration of domain names that users might be able to guess to find their site, which in turn improved
the users' chances of navigating by guessing. That users were inclined to use domain names to search for
content of interest increased the desirability of domain names corresponding to generic words, such as
"business," "jobs," or "sex." And the importance of having those names in the .com domain was
increased even more by the design of Web browsers. Recognizing user inclinations, designers made the
default behavior of many Web browsers when confronted by an incomplete domain name the automatic
addition of .com to the end of it and www as the default prefix.9 Thus, domain names became not only
the way of designating locations on the Internet, but also a principal means of navigating to them.10
Valuable second level domain names. The third development was the recognition that certain
domain names within the top-level domains--second-level domain names --are more valuable than most
others. The result was an aftermarket for domain names, generally in the .com top-level domain, in
which some have been resold for prices far greater than the nominal registration fee paid by the original
registrant. Furthermore, in an effort to protect their rights and prevent others from abusing them,
trademark holders have sought to acquire many of the domain names incorporating their trademarks and,
given the likelihood of entry errors, words that are typographically close to them in all of the relevant top-
level domains. This effort has, in turn, led to competition among trademark holders with the same mark
(though in different industries or regions) for the small number of memorable domain names
incorporating their marks. (Individuals or groups with other legitimate claims to a name--such as those
with the surname McDonald11--have also asserted their rights to domain names incorporating
trademarks.) It has also attracted speculators who rush to acquire potentially desirable domain names
(both trademarks and generic words) in order to resell them to those for whom the value would be
substantially greater than the registration fee.12
Marketing function. The value of domain names has been further enhanced by their widespread
use in marketing materials as a secondary, or even primary (e.g.,, identifier of an

8 In mid-2004 there were almost 27 million .com registrations compared with 4.4 million for .net and 2.8 million
for .org. See Table 3.3.
9 Browsers in 2005 no longer make this assumption. Instead, they commonly assume that the entry is a search term.
10 The unique role is elaborated on in Chapter 2.
11 The new top-level domain, .name, for registration by individuals was intended to meet that need, although by
mid-2004 the companies involved with registrations in that domain had not found business models capable of
supporting those operations.
12 A case in point is the name "," which changed hands for $150,000 in 1997 and was resold for
$7.5 million in 1999. See " The $7.5 Million Domain," ZDNet News, Jennifer Mack, December 1,
1999, available at <>. However, in the aftermath of the bust, the prices realized in the aftermarket for domain names have also subsided substantially, although at
the end of 2003 the name "" was sold for $1.3 million by a person who paid $15,000 for it in 1997. See
"Domain Names Once Again Fetch Top Dollar," Associated Press, Anick Jesdanun, December 25, 2003.

Prepublication Copy--Subject to Further Editorial Correction
organization. In that role they appear on stationery, in newspaper ads, on billboards, and on the sides of
buses. This marketing function of domain names is the fourth unanticipated development.

Because of these developments, the Domain Name System--originally a modest technical system
introduced to provide easy-to-remember and portable names for locations on the Internet--has become a
critical tool facilitating global communication by designating sources of information, products, and
services as well as the e-mail boxes of people and organizations throughout cyberspace. As a
consequence, its simple original institutional framework, managed essentially by one person,13 has been
replaced with a complex network of institutions comprising numerous public and private, commercial and
non-commercial organizations that register domain names and operate name servers; and one non-
governmental organization with international scope that, with the authority and oversight of the U.S.
government, provides technical coordination and establishes some elements of global policy--the Internet
Corporation for Assigned Names and Numbers (ICANN).

1. What is the IP address of the
domain name

2. The IP address of is
3. Retrieve the home page at
4. Here is the
home page at

FIGURE 1.1 The Domain Name System and Internet navigation for the Web--navigating to The Web site and IP address used are fictional.


With the growth of the size, complexity, and variety of applications using the Internet, and
especially the rapid growth of the World Wide Web, a range of aids to the navigation process (especially
on the Web) have appeared.14 The DNS is a single technical system providing a single service, which is
operated and controlled in a complex institutional framework. In contrast, among aids to navigation on

13 Jon Postel held this responsibility for many years. For further information, see <>.
14 If the location of a desired Web site is known, then navigation can be direct--the DNS determines the IP address
of the site. If the location is unknown, then some navigation aid must first be used to determine the location. See
Section 71 for a detailed discussion of "direct" and "indirect" navigation.

Prepublication Copy--Subject to Further Editorial Correction
the Web there are numerous specialized navigation services, each operated by different providers that
compete openly without any comprehensive institutional framework for their operation and control.
Principal among these navigation services are search engines, which at the possible cost of a few more
keystrokes open up a far wider range of possibilities on the Web than simple domain name guessing; and
directories, which provide a "yellow pages" or "white pages" guide to locations, principally on the Web.
As search engines have improved in user-perceived quality and ease of use, they have become a principal
means of navigation to new destinations for many users.15 In place of guessing, intrepid Web travelers
enter a descriptive word or phrase in the search engine and use the resultant list to direct their journeys.
In June 2004, nearly 4 billion searches were conducted each month by almost 110 million people in the
United States, an average of 33 searches per person per month.16 It appears that a consequence of the
growing use of search engines for navigation across the Web may be a reduction, though by no means
elimination, of the direct use of the DNS to support navigation by guessing domain names.
Because locations that offer search or directory capabilities are accessed so often, a number have
evolved into portals, which are Web sites offering directed links to popular categories of services, such as
My Yahoo!. Portals such as MSN and AOL have also evolved from provision of Internet service. Some
users find the portals more desirable than search engines alone and begin their navigation from them. In
doing so, they are relying on the editorial judgment and commercial or other arrangements of the portal to
get started. However, once a World Wide Web site is reached--no matter how--subsequent navigation
often flows along the network of links from one site to others. And it is likely that most experienced
users deploy a combination of navigation aids and services: employing search engines, portals, and direct
entry of destinations into browsers at various times.17 Other navigation aids are also in use. In some Web
portals run by ISPs (such as AOL) or through extensions to browsers, a specified vocabulary of key
words can be used to reach specific destinations.
The cumulative effect of these aids to navigation has, thus far, been positive. They enable users
to find sources of products, services, information, and contacts that they would not have been able to
identify previously. Complementarily, they enable providers to reach audiences that might not otherwise
have known of their existence. Unlike the development of the top levels of the DNS, which has been
under the technical control of the Internet engineering community and the governance of ICANN,
national governments, and the operational organizations, these navigation services have for the most part
been provided by private organizations. Despite their benefits, however, navigation aids are also
beginning to raise policy concerns. Most search engines and directories now accept payment from
advertisers for placement of an ad on the pages of responses to queries with specific search terms. If not
clearly identified such ads might give searchers a false sense of an advertiser's importance or relevance
and reduce the chances that a non-advertiser will be located. Concerns about the practices of the providers
of navigation services are likely to grow as Internet users rely increasingly on these services as a principal
means of navigation.

15 According to data from WebSideStory, both direct navigation using a known domain name and the use of Web
search engines increased substantially from 2002 to 2003. In March 2003, 13.6 percent of Web site accesses were
generated by search engine listings, while 66 percent were the result of bookmarks or direct entry of a known
address. See <>, press release, March 12, 2003.
16 Deborah Fallows, Lee Rainie, and Graham Mudd. 2004. "The Popularity and Importance of Search Engines."
Data Memo. Pew Internet & American Life Project. August. Available at
<>. The results came both from a telephone
survey of 1399 Internet users and from tracking of Internet use by comScore Media Metrix.
17 These destinations might be derived from guesses about domain names as discussed above or references provided
by others (e.g., in an e-mail) that are copied and pasted into browsers, as well as by the use of bookmarks for
destinations that are accessed frequently.

Prepublication Copy--Subject to Further Editorial Correction

Since the early 1980s, when the DNS was adopted, five forces have inexorably driven the
transformation of the Internet from its origins as a small, primarily North American research network,
which was run by a tight-knit group of specialists for use within their research and industrial
communities, into its current state as a diffusely managed and increasingly critical part of the global
information and communication infrastructure. These driving forces are increasing scale, technological
progress, increasing economic value, increasing social value, and internationalization. In addition to
having a profound impact on the Internet (and the World Wide Web) as a whole, these forces have
simultaneously transformed the DNS and Internet navigation to subjects of substantial commercial, legal,
political, and social importance. The likelihood of the continued influence of these forces raises
important questions about the future viability, operability, and governability of the DNS and Internet
navigation--the subjects of this study.

1.4.1 Increasing Scale

When the DNS was developed, the Internet comprised on the order of a 1000 sites and perhaps
10,000 users. In two decades it has grown to more than 30 million sites and over 600 million users.18
Though its designers did not fully anticipate the rapid growth in users and uses stimulated by the World
Wide Web, the DNS has technically scaled quite well to the current size. In addition, new navigation
tools have been deployed to assist users in searching the vastly larger Internet. The Internet continues to
grow in number of users, number of addresses, and number and diversity of attached devices. By 2010, at
current growth rates, the Internet could have more than 60 million sites and well over a billion users

1.4.2 Technological Progress

When the DNS was developed, most of the hosts were workstations, minicomputers, or
mainframe computers. Personal computers had just begun their penetration of the business and home
markets in North America, Europe, and Japan. Internetworking communication took place over
backbones that had 56 kbps (kilobits per second) speeds--about 1/180,000th the speeds of backbones in
2004, which run at 10 Gbps (gigabits per second). As the capacities of computers and communications
networks have soared, the DNS and navigation systems have taken advantage of the increased
computational capability and bandwidth to meet the challenges of scaling. Continuing technological
advances in computing and communications offer the possibility of strengthening the DNS and increasing
the capacities of navigation services, while at the same time further empowering those who would attack
the services or attempt to misdirect them for their own benefit.

1.4.3 Increasing Economic Value

When the DNS was developed, there was probably little or no economic value associated with
possession of a particular domain name, which could be obtained at no cost, although having a

18 These numbers reflect estimates made in May 2003 by CyberAtlas; see <>.
19 To serve them, the basic IP address--currently 32 bits--is being enlarged to 128 bits, enabling addressing of a
wide range of devices from computers and cell phones to home digital media centers and home appliances. The
current IP address space is called the IPv4 address space; the new version is called the IPv6 address space. IPv6 is
slowly being adopted, working in parallel with IPv4.

Prepublication Copy--Subject to Further Editorial Correction
hierarchical naming system was judged to be valuable. A distinctive Internet culture had developed well
before this time, led by the relatively small and homogeneous community of engineers and scientists who
were its primary users. It placed high value on voluntary service, free access within the community, and
consensus decision making. However, the growth of applications on the Internet for commerce,
information, art, and entertainment attracted commercial, legal, governmental, and other communities
whose values and processes differ from those of the early Internet culture. Their arrival led to the
development of a vigorous market for domain names and of a variety of mechanisms to deal with fair
allocation of the now economically valuable domain names. Not surprisingly, throughout these
developments there has been a continuing tension between the technical community and the public
interest community about the proper goals and mechanisms for the allocation of domain names and the
management of the DNS.
As domain names have gained economic value, so, too, has their owners' desire grown for
opportunities to publicize those names (as part of Web site and e-mail addresses) to potential users of the
corresponding Internet locations. Consequently, many search engines and other navigational services,
which originally provided a single listing of search results in the order of estimated relevance to the user's
query, now also give prominent placement to those willing to pay for it. As noted above, the search
engine industry faces a continuing challenge in finding the proper balance between the interests of the
users of search engines and the advertisers on them, against the backdrop of the ever present possibility of
government intervention.

1.4.4 Increasing Social Value

When the DNS was developed, there was a modest level of social, political, or cultural value
associated with specific domain names. As the Internet grew in size and evolved in use, it became a
primary medium for communication, commerce, information, art, and entertainment; accordingly, domain
names assumed greater social, political, and cultural significance as the memorable designators of the
Internet locations of political groups, cultural resources, and social activities. But as a result, the DNS
became entangled in issues of privacy versus accountability, freedom of expression versus national legal
restrictions, and the rights of producers of intellectual property versus those of its users.
In the future, the Internet can be expected to be even more widely used for interpersonal
communication, for the public expression of ideas, for access to information, for the development of
virtual communities around common interests, and for the production and distribution of art and
entertainment. It will be a major portion of the global social fabric, facilitating and/or controlling the flow
of information, expression, art, and entertainment. Until or unless the DNS is replaced, the signs
designating the location of information, art, entertainment, viewpoints, and services will continue to
depend on domain names.
For that reason, it will be essential to sustain the DNS as the reliable signposting infrastructure of
the Internet, facilitating the Internet's use as a medium of free and open communication reaching all
corners of the globe, while balancing privacy, freedom of expression, property rights, cultural mores, and
national laws. As a result of the Internet's increased social value, the desire to navigate freely across it can
also be expected to encounter legal, commercial, cultural and political challenges that will pose issues
about the appropriate balance among rights, freedoms, mores, and laws.

1.4.5 Internationalization

When the DNS was developed, the Internet's geographic scope was limited primarily to North
America, parts of Western Europe, and a few countries on the Pacific Rim. And it was operated by a
loose confederation of bodies and individuals, primarily in the United States, most of whom had received
substantial support from the U.S. government. As use of the Internet has spread beyond its initial sites to

Prepublication Copy--Subject to Further Editorial Correction
encompass every continent and region and almost all nations, the network has responded successfully.
But internationalization has posed two specific challenges for the DNS.
First, until recently domain names have been limited to strings of Roman letters, Arabic numbers,
and the hyphen, a subset of the ASCII20 character set. However, the native languages of an increasing
number of Internet users employ different character sets. Recently, following years of work, a means of
enabling presentation of internationalized domain names (domain names encoding other character sets
into ASCII characters) has been adopted. It should become an important facilitator of Internet access and
use for those communities.21
And, second, although ICANN has international participation, its authority rests on a contract
from the U.S. Department of Commerce, which is perceived by some as undercutting its legitimacy as a
representative of the international community. That concern may increase as the economic and social
value of the DNS as the critical signposting infrastructure of the Internet continues to grow.22
Although the DNS is only now moving toward presentation of non-ASCII scripts in domain
names, Internet content in most important applications, including e-mail and the Web, has been
internationalized for well over a decade. Most Internet navigation services have incorporated the
capability to search in multiple languages. For example, in November 2004 the Google search engine
supported searches in over 100 languages and dialects and provided a customized version of the search
interface for 103 different nations.23 At the same time, the Yahoo! directory and search service offered
portals customized for 32 national and/or language groups.24 Since the navigation services are provided
by a variety of organizations in an open forum, they are less subject to concerns about the
internationalization of their governance. However, as their importance as the principal means of access to
the Internet grows, they may well come under pressure from those who believe that in one aspect of their
service or another, they do not adequately take into account the concerns or interests of certain nations,
ethnic groups, or linguistic communities.


Owing to the five forces outlined above, Internet naming and navigation have become matters of
broad concern throughout the world. Those concerns are given voice by the large number of competing
interest groups that now take a vigorous interest in the DNS and, to a somewhat lesser degree, Internet
Product and service providers compete for named locations on the Internet and have a strong
interest in the means for setting up new regions and allocating named locations in them. Internet users
have a complementary interest in being able to find the information or service they want wherever it may
be located, even as the Internet continues to grow in size and in diversity. All cultures have an interest in
being able to name locations and access and navigate the Internet in their native languages. Trademark
holders have an interest in protecting their rights in names from being infringed. Nations and their
citizens want assurance that their interests will be treated fairly and their needs supported by the
institutional frameworks that affect the Internet's naming and navigation infrastructures. The Internet

20 The American Standard Code for Information Interchange (ASCII) was originally developed for use with
teletype. It was extended by IBM to represent 256 characters and has become a de facto standard.
21 Internationalized domain names (IDNs) have recently been approved by ICANN for use by registries with which
it has agreements. See "Standards for ICANN Authorization of Internationalized Domain Name Registrations in
Registries with Agreements," posted March 13, 2003, on the ICANN Web site, <>. See
Section 4.4 for a more complete discussion of this subject and more extensive references.
22 Changes in ICANN's organizational structure and decision processes responded to this concern, although debate
continued into 2004 about the effectiveness of those changes. See Sections 5.1 and 5.2 for an extended discussion
of this issue.
23 For a listing see <>.
24 For a listing see <>.

Prepublication Copy--Subject to Further Editorial Correction
technical community wants to ensure that everything is done to improve and nothing is done to
compromise the reliability, security, and stability of the Internet itself. Individuals also want to be certain
that neither their access nor their rights will be unduly affected by the actions of the other groups. The
interaction among these various interests in naming and navigating the Internet--a global infrastructure
that is undergoing rapid growth in scale while absorbing continual technological change--raises
important issues that lie at the intersection of technology, economics, public policy, law, and user
This report addresses those issues from a specific perspective, that of the Domain Name System.
Navigation across the vast and multifaceted complex of human activity connected through the Internet is
a subject that warrants a major report in its own right. It is too large, in its full richness, to fit within a
report that was initiated to address significant questions about the future of the DNS. Yet, at the same
time, navigation is so intertwined with the present and future of the DNS that it cannot be completely
absent. Consequently, this report concentrates on navigation over the Internet primarily in its
relationships with the DNS. Even under that constraint, however, it is necessary to introduce fundamental
issues of Internet navigation to provide a background for the more circumscribed examination of its
interrelationships with the DNS.
The DNS interrelates with navigation across the Internet in five ways.

First, the DNS plays a direct navigational role by providing the IP address of a World Wide
Web site, an e-mail server, or another network host or resource whose domain name is known, or can be
Second, the DNS serves as an enhancer of navigation because many navigation services
return locators incorporating the domain names of relevant Web sites. These names usually provide more
(although not necessarily reliable) information to the user about the provider whose location has been
returned than just the IP address or a blank link would.
Third, navigation services complement the DNS by, for example, enabling navigation to Web
sites whose domain names are not known by the user or by enabling searches within sites that have been
reached by use of their domain names.
Fourth, navigation services relieve some of the pressure on the Domain Name System by
reducing the need for a site to have a short memorable name in order to be found. It appears that efforts
and funds spent in previous years to obtain desirable domain names are now being diverted to some
degree to efforts and expenditures to ensure a presence and high ranking in the results of search engines
or directories.
Fifth, navigation services could, in the extreme, substitute entirely for the Domain Name
System on the Web because they could directly return IP addresses. However, as noted above, this
approach would deprive the user of any information about the provider contained in the domain name. It
would also deprive the provider of the marketing value of the domain name. And it would eliminate the
use of domain names as stable identifiers of Internet resources whose IP addresses change, which was one
of the original motivations for the creation of the DNS.

Of these five roles, it is the third and fourth--navigation as a complement to and a relief for the
DNS--that are the focus of this report's examination of Internet navigation.


This report is addressed to those who are or will be concerned with policies and practices that
affect the operation and evolution of the DNS and Internet navigation. That is a large audience. It
includes the technologists who research, design, implement, and operate the DNS and navigation systems;

25 See CSTB, NRC, The Internet's Coming of Age, 2001.

Prepublication Copy--Subject to Further Editorial Correction
the governmental policy makers and their staffs who establish, oversee, and operate the framework of
institutions and laws that govern or regulate those systems; the commercial and non-commercial
organizations that operate, manage, and use those systems; and the users and providers who depend on
those systems for access to and the accessibility of Internet locations.
During the time that this report has been in preparation, the DNS and Internet navigation have
seen many technical and institutional changes, some substantial, others modest; some controversial,
others agreeable; some likely to last, others temporary expedients that will eventually be replaced. Some
of the changes made have addressed the issues that gave rise to the initial request for this report. Clearly,
this study was not the proper vehicle to address those specific issues. However, it is equally clear that
many issues of similar character are or soon will arrive on the agendas of the policy, technical, provider,
and user communities. Yet those called upon to deal with policy and practices affecting the DNS and
Internet navigation often have little or no knowledge of the full complexity of those arenas. Those who
are engaged with the technology of these worlds do not always appreciate the nuances of the policy,
economic, and legal issues, while those experienced with the legal, economic, and policy aspects often are
largely unaware of the intricacies of the technology. The result is often a clash of cultures that inhibits
both effective policy making and desirable technological change. Both groups would benefit from having
a reliable source of information about the technologies and the institutions that control them, upon which
they can base reasonable and effective policies. And, where appropriate, they might also benefit from the
conclusions and recommendations of a broadly knowledgeable committee that has spent several years
reviewing the two worlds.
Therefore, this report, which is the result of extensive and collaborative work by a committee
whose members are drawn from both the technology and the policy communities, is intended to serve five

1. To provide a thorough and objective description and assessment of the Domain Name
System--both its technology and the institutional framework within which that technology operates;
2. To describe and analyze alternative approaches to the principal technology prospects and
institutional issues that are likely to affect the future of the DNS;
3. To provide a thorough and objective description and assessment of Internet navigation, with
sufficient background information to provide context;
4. To describe and analyze alternative approaches to some of the technology prospects and
institutional issues that are likely to affect the future of Internet navigation; and
5. To present conclusions and make recommendations where it was possible for the committee
to reach agreement--in any case, to characterize the range of alternative views.

This report has been structured to respond to those objectives.


This report is divided into three parts. The DNS is the subject of the first, consisting of Chapters
2, 3, 4, and 5. Internet navigation in its relationship to the DNS is the subject of the second, consisting of
Chapters 6, 7, and 8. Chapter 9 summarizes the interaction between the DNS and Internet navigation.
Because the options for moving forward are partially constrained by the decisions taken along the
path to the present, the first part begins with a careful review of the development of the Domain Name
System. Chapter 2 examines the evolution of the technical design of the DNS and its associated
operational, administrative, and governance mechanisms. It describes the sequence of important technical
decisions and innovations, as well as the new governance and administrative mechanisms that have been
introduced in response to the Internet's rapid growth. Several of the early technical decisions, taken at the
time of restricted use of internetworks by specialized groups, still constrain the DNS.

Prepublication Copy--Subject to Further Editorial Correction

Chapter 3 describes the current state of the DNS, considering both the technical system, which
performs the linkage of domain names with IP addresses and associated data, and the higher-level
institutional framework, which carries out operational, administrative, and policy-setting functions
essential for the DNS to function. It explains and evaluates the operation of the DNS technical system
and identifies and assesses each of the functions carried out by the highest levels of the institutional
Chapter 4 describes the prospective technologies that can respond to the challenges the DNS
faces from malicious attacks, the growing intersection of the telephone system and the Internet, the need
to internationalize the name system, and the need to regulate the introduction of potentially disruptive
new services. Chapter 5 deals with the key institutional issues facing the DNS: governance of the DNS
itself, oversight of root operations, governance of the top-level domains, improvement of the dispute
resolution process, and improvement of the DNS's information service (called the Whois service).
The distinctive characteristics and historical development of Internet navigation, as it relates to
the DNS, are described in Chapter 6. The current state of navigation aids and services and the framework
of commercial institutions within which they operate are presented in Chapter 7. Chapter 8 addresses
some prospective technologies whose introduction, and a number of the institutional issues whose
resolution, can have a major influence on the future development of Internet navigation and its
relationship to the DNS.
Finally, Chapter 9 sums up the interaction between the DNS and Internet navigation.
Throughout the chapters on the DNS and Internet navigation, the committee's conclusions and
recommendations are incorporated into the text where appropriate.
The goal of this report is to clarify the sometimes controversial, often arcane, and frequently
uncertain issues concerning the signposting and navigational infrastructure of the Internet. The committee
hopes that by providing such clarification, this report will itself serve as a navigational aid to the policy
and technology communities as they find their way to decisions that will enable the Internet to remain an
efficient and reliable channel of global communication and commerce.

Prepublication Copy--Subject to Further Editorial Correction
The Emergence and Evolution of the Domain Name System

The Domain Name System (DNS) was designed and deployed in the 1980s to overcome
technical and operational constraints of its predecessor, the HOSTS.TXT system. Some of the
initial design decisions have proven to be extraordinarily flexible in accommodating major
changes in the scale and scope of the DNS. Other initial design decisions constrain technical and
policy choices to the present day. Thus, an understanding of the system architecture and the
rationale for the design characteristics of the DNS provides the base for understanding how the
DNS has evolved to the present and for evaluating possibilities for its future. This chapter
outlines the origin and development of the DNS and describes its key design characteristics,
which include both technological and organizational aspects.1


For the first decade or so of the ARPANET,2 the host3 table file (HOSTS.TXT) served as
its directory. HOSTS.TXT provided the network address for each host on the ARPANET,4 which
could be looked up by using the host's one-word English language name, acronym, or
abbreviation. The Network Information Center (NIC) at the Stanford Research Institute5
managed the registration of hosts and the distribution of the information needed to keep the
HOSTS.TXT file current. The list of host names and their mapping to and from network
addresses was maintained in the frequently updated HOSTS.TXT file, which was copied to and
stored in each computer connected to the ARPANET. Thus, HOSTS.TXT6 was introduced to:

Simplify the identification of computers on the ARPANET. Simple and familiar
names are much easier for humans to remember than lengthy (12-digit) numeric strings; and

1 A general presentation of the history of the Internet is beyond the scope of this report. One source of
documentation on the Internet's history is available at <>.
2 The Internet grew out of the ARPANET project (funded by the Defense Advanced Research Projects
Agency (DARPA), which was known as ARPA for a period of its history); for many years the ARPANET
served as the core of the Internet.
3 A host is the primary or controlling computer in a network.
4 These network addresses could be represented using the Internet Protocol (IP) format or in the equivalent
(now unused) ARPANET Network Control Protocol (NCP) format. The current version (v4) of IP
represents addresses using 32 bits, usually expressed as four integers in the range from 0 to 255, separated
by dots. An example of an IP address is ""
5 Stanford Research Institute became known as SRI International in 1977.
6 For further discussion, see "Host Names On-line," Request for Comments (RFC) 606, L. Peter Deutsch,
December 1973; "Hostnames Server," RFC 811, Ken Harrenstien, Vic White, and Elizabeth Feinler,
March 1982; and "DOD Internet Host Table Specification," RFC 952, Ken Harrenstien, M. Stahl, and
Elizabeth Feinler, October 1985, all available at <>. RFCs are created to
document technical and organizational aspects of the Internet. The Internet Engineering Task Force (IETF)
manages the process for discussing, evaluating, and approving RFCs. For a discussion of the role of the
DNS more generally, see "Role of the Domain Name System," RFC 3467, John C. Klensin, February 2003.


Prepublication Copy--Subject to Further Editorial Correction
Provide stability when addresses changed. Since addresses in the ARPANET are a
function of network topology and routing,7 they often had to be changed when topology or
routing changed. Names in the host table could remain unchanged even as addresses changed.

The HOSTS.TXT file had a very simple format. Each line in HOSTS.TXT included
information about a single host, such as the network address, and when provided, system
manufacturer and model number, operating system, and a listing of the protocols that were
Because a copy of the host table was stored in every computer on the ARPANET, each
time a new computer was added to the network, or a change was made, the entire table had to be
updated at every computer on the network for the new machine or location to be recognized.8 As
increasing numbers of computers joined the ARPANET, the updating task became more and
more burdensome and subject to error and failure, and, as a consequence, several major problems
developed from the use of the HOSTS.TXT file:

Failure to scale. As the ARPANET started to grow rapidly, it became clear that the
centralized HOSTS.TXT file failed to scale in two ways. First, the volume of updates threatened
to overwhelm the NIC staff maintaining HOSTS.TXT. Second, because every system needed to
have an up-to-date copy of HOSTS.TXT, announcement of a new copy of HOSTS.TXT meant
that the NIC server where the current version of HOSTS.TXT was stored was inundated with
attempts to download the file. Moreover, the download problem was aggravated because
HOSTS.TXT kept getting bigger. In short, more hosts on the network meant more updates, more
hosts trying to download, and more data to download.
Inadequate timeliness. It often took several days to get a new host listed in
HOSTS.TXT while the NIC staff processed the request to add the host entry. Until it was listed
and communicated, the host was effectively invisible to the rest of the ARPANET. In a
community already becoming accustomed to getting data instantly over the network, this delay
was a source of frustration. Similarly, correcting an error often took a few days, because fixes to
any errors were not generally available until the next HOSTS.TXT file was released--which
caused further frustration. The maintainers of some hosts also did not update their copies of the
table at very frequent intervals, resulting in those hosts having obsolete or incomplete information
even when the master copy of the table was up-to-date.
Susceptibility to failure. The system had multiple ways to fail. Probably the most
famous outage occurred when the NIC released a version of HOSTS.TXT that omitted the entry
for the system where the HOSTS.TXT file was stored. When the subsequent HOSTS.TXT file
was released, most systems could not download it, because they could not look up the relevant
host name! There were also cases where partial tables were inadvertently released. Furthermore,
seemingly innocuous additions to HOSTS.TXT could cause the programs that converted
HOSTS.TXT into local formats to fail.
Name conflicts. The HOSTS.TXT name space was flat, which meant that host names
had to be unique. Popular host names such as FRODO were selected first, and so some people
had to invent alternate names for their systems.
The emergence of these problems caused technologists to develop a new, distributed,
method for managing the mapping of names to addresses.

7 Routing refers to the way data flow on the ARPANET. Data transmitted from point A to point B may
traverse many different paths, or routes, on the ARPANET. Note that the ARPANET, as the original
network to employ the Internet Protocol (IP), is often referred to as "the Internet," although the term later
formally encompassed the aggregate of interconnected IP-based networks.
8 It was the obligation of individual network and host operators to download the latest HOSTS.TXT file to
their machines.


Prepublication Copy--Subject to Further Editorial Correction

In the early 1980s, research on naming systems--systems for associating names with
addresses--was underway and a few prototype naming systems had just been developed, most
notably the Grapevine and Clearinghouse systems at the Xerox Palo Alto Research Center
(PARC).9 Also in progress at this time was preliminary work on other computer network
addressing standards such as X.400.10 Because of the uncertainty as to whether these research
and development efforts would yield in the near term an operational system with the required
functionality and needed scale, Internet researchers elected to develop their own protocols.
In August 1982, Zaw-Sing Su and Jon Postel authored "The Domain Naming Convention
for Internet User Applications," Request for Comments (RFC) 819, which described how Internet
naming should be changed to facilitate a distributed name system. This document envisioned that
Internet names would be organized into logical hierarchies, represented by text components
separated by a period (".") (thus the existing host "ISIF"--host computer "F" at the Information
Sciences Institute (ISI)--would become F.ISI), and that the various parts of the name as assigned
(i.e., the parts delimited with periods) would be managed by different network servers. RFC 819
specified only how names would be represented--the details of how the management of various
parts of assigned names would take place operationally by the different network servers remained
to be determined.
In November 1983, Paul Mockapetris authored "Domain Names--Concepts and
Facilities" (RFC 882) and "Domain Names--Implementation and Specification" (RFC 883),
which specified a set of protocols, called the Domain Name System, that implemented the
hierarchical name space proposed by Su and Postel. Reflecting the discussions of the previous
several months on the electronic mail list Namedroppers, the proposed DNS supported more
sophisticated services and features than simply converting host names to addresses (e.g., the
proposed system would provide a way to map a name to different addresses, depending on the
purpose for which an inquiry was being made). With some modest changes, the proposed
protocols are exactly those in use two decades later.
Conceptually, the DNS is implemented through a distributed and hierarchical series of
tables, linked like the branches of an inverted tree springing from a single, common root. When
an address is sought, the search proceeds successively from the table at the root (or "top") of the
tree to successive branches and leaves, or "lower tables," until the table that holds the desired
address is found. For a particular query, only the last table in the search serves as a "white pages
directory." All of the other tables serve as "directories of directories," each one pointing to lower
level "directories" on a path to the one holding the desired address. Thus, the entries in a table at
any given level of the tree can include pointers to lower-level tables as well as final network
addresses. See Figure 2.1.
When a change is made in the network, only the table directly affected by that change
must be updated and only the local organization (e.g., the system administration function in a
university or corporation) responsible for that table needs to make the update. As a result, the
work of registering changes is distributed among many organizations, thus reducing the burden
each must carry.

9 See Andrew D. Birrell, Roy Levin, Roger M. Needham, and Michael D. Schroeder, 1982, "Grapevine: An
Exercise in Distributed Computing," Communications of the ACM 25(4):260-274.
10 The International Organization for Standardization and the International Telecommunication Union
endorsed X.400 as a standard that describes a messaging service (e.g., electronic mail). The first version of
X.400 was published in 1984 by the Comité Consultatif International Téléphonique et Télégraphique
(CCITT), which is now the International Telecommunication Union­Telecommunication Standardization
Sector (ITU-T)).


Prepublication Copy--Subject to Further Editorial Correction
The DNS naming syntax corresponds to the levels in the hierarchical tree. Each node in
the tree has a name that identifies it relative to the node above it. The highest level, the "root
node," has the null name. In text it is written as a single dot (".") or simply implied (and thus not
shown at all). Each node below the root is the root of another subtree, a domain, that can in turn
be further divided into additional subtrees, called subdomains. Each subdomain is written in text
to include its name and the subdomains above it in the applicable hierarchy. In
Figure 2.1, .com, .org, and .edu are top-level domains (TLDs) and,,
and are subdomains of the TLDs, often called second-level domains. The third-level
domain,, is a subdomain within the second-level domain.
The DNS name of a computer is the name of its node or end point in the Domain Name
System. Thus, would be the computer (or device) named "frodo" that
is located within the subdomain of the second-level domain within
the .edu TLD. On the other hand, (without any further
subdomains) could point directly to a particular computer.
Applications, such as Web browsers and e-mail software, use domain names as part of
the Uniform Resource Identifiers (URIs) or other references that incorporate information about
the protocols required for communication with the desired information source. Examples of URIs
are and In the first
example, "http" refers to the Hypertext Transfer Protocol (HTTP) used for communication with
sites on the World Wide Web. In the second example, a particular user at the host identified by
"" is identified as the addressee for electronic mail.
In terms of information technology, the Domain Name System is implemented through a
series of name servers that are located at each of the nodes in the hierarchy. Each name server
contains a table that indicates the locations of the name servers immediately below it in the
hierarchy and the portion of the hierarchy for which it contains the final (authoritative) network
addresses. Thus, the root name servers (at the top of the hierarchy) contain the locations of each
of the name servers for the top-level domains.11 At any given node, such as .com or,
there are expected to be multiple (physical) name servers at different Internet Protocol (IP)
addresses, each with identical information; the purpose of this redundancy is to share the
workload to ensure adequate system performance.
When a user wants to reach, his or her computer
usually sends a message to a nearby name server (usually local or operated by the user's Internet
service provider), where software (called a resolver), in conjunction with other name servers and
resolvers, performs a series of queries to find the name server that is authoritative for That server is then queried for the corresponding IP
address(es) and returns the resulting address(es) to the user's computer.12

11 Each of these root name servers contains identical information; the purpose of having multiple root name
servers is to distribute the query workload and ensure reliable operation. Specifics concerning the root
name servers are discussed in Chapter 3.
12 The summary provided in this paragraph is quite simplified; there are many discrete technical processes
that are not articulated here. See Chapter 3 for a more detailed explanation.


Prepublication Copy--Subject to Further Editorial Correction

"." root


FIGURE 2.1 The hierarchical Domain Name System inverted tree structure.

2.2.1. Simple, Mnemonic, and Deeply Hierarchical Names

As indicated above, domain names were intended to enable a more convenient and
efficient way of referring to IP addresses and other information, using a simple taxonomy. The
early DNS included eight generic top-level domains (gTLDs): .edu (higher education
institutions--most of which were based in the United States), .gov (U.S. government), .mil
(U.S. military), .com (commerce), .net (network resources), .org (other organizations and
persons13), .int (international treaty organizations) , and .arpa (network infrastructure). In
addition, country-code top-level domains (ccTLDs) were created based on the two-letter code set
(e.g., .gh for Ghana or .au for Australia) in the ISO 3166-1 standard.14
Despite the ability of the protocols and data structures themselves to accommodate any
binary representation, DNS names were historically restricted to a subset of the ASCII character
set.15 Selection of that subset was driven in part by human factors considerations, including a
desire to eliminate possible ambiguities in an international context. Hence, character codes that
had international variations in interpretation were excluded; the underscore character (too much
like a hyphen) and case distinctions (upper versus lower) were eliminated as being confusing
when written or read by people; and so on. These considerations appear to be very similar to

13 Initially, the .org TLD was intended as the category for organizations and individuals that did not fall
into any of the other categories. Through time, many individuals increasingly viewed .org as
representing the domain name space for non-profit organizations.
14 Thus, the determination of what constitutes a country did not need to be addressed by those who
administer the DNS. See <>.
15 This subset, which derives primarily from the original HOSTS.TXT naming rules, includes the 10 Arabic
digits, the 26 letters of the English alphabet, and the hyphen.


Prepublication Copy--Subject to Further Editorial Correction
those that resulted in similarly restricted character sets being used as protocol elements in many
International Telecommunication Union (ITU) and International Organization for Standardization
(ISO) protocols.
Another initial assumption behind the design of the DNS was that there would be
relatively many physical hosts for each second-level domain name and, more generally, that the
system would be deeply hierarchical, with most systems (and names) at the third level or below.
Some domains--those of most universities and some large corporations and the country code for
the United States (.us)--follow this model, at least in its original design, but most do not16 (see
Chapter 3 for discussion). However, experience through the year 2004 has shown that the DNS is
robust enough--given contemporary machines as servers and current bandwidth norms--to
operate reasonably well even though the design assumption of a deep hierarchy is not satisfied.
Nonetheless, it is still useful to remember that the system could have been designed to work with
a flat structure (e.g., the huge, flat structure under .com comprising tens of millions of names)
rather than a deeply hierarchical one. For example, based on an assumption of a flat structure at
the TLD level, one would probably not wish to assign specific operational responsibility by TLD
(as is the case currently). Instead, it might have made more sense to design the system as one
database that is replicated on a limited number of servers (to share the workload and coordinate
updates in a manageable way).
2.2.2. Experimental Features

The DNS specification included a number of experimental features, intended to enhance
the services that the DNS could provide beyond simple name-to-address lookup. Several of these
features were intended to facilitate improved support of electronic mail. Several resource
records17 were intended to improve e-mail routing, helping to ensure that e-mail sent to a
particular host took a reliable route to that host. The DNS also included features intended to
support e-mail lists and aliases. The idea was to make it easier to maintain mailing lists and to
forward mail when someone's e-mail address changed. In addition, the DNS contained a feature
to track "well-known services." The purpose of this feature was to provide a list of services (e-
mail, File Transfer Protocol, Web) that are available from a host. Most of the experimental
features have not been adopted for general use. Indeed, the original set of e-mail-related record
types were deprecated in favor of a newer model (see Section 2.3.3) and the "well-known
services" record was determined to be unworkable.


Whereas the design of the DNS looked reasonable on paper, several limitations of the
new system, as with any new system, did not become apparent until initial deployment began.
Addressing these limitations caused a delay in the full implementation of the DNS. The plan
called for a switchover to the DNS in September 1984, but full conversion did not take place until

16 The .us country code TLD was designed originally to use geographical and political jurisdictions as
subdomains. As one moves to the left, each subdomain represents a subset of the area represented by the
immediately preceding name. For example, in the name, "va" represents
the state of Virginia within the United States, "reston" represents a city within Virginia, and "cnri"
represents an organization in the city of Reston.
17 Each table within the domain name tree hierarchy contains resource records, which are composed of
fields such as the type (i.e., does this record correspond to a host address, an authoritative name server, or
something else) and time to live (i.e., for what period of time may this record be cached before the source
of the information should be consulted again?). See Box 3.2 for detailed discussion of resource records.


Prepublication Copy--Subject to Further Editorial Correction
1987. Some of the delay was attributable to reconciling naming conflicts.18 A large part of the
delay derived from a far longer than expected period to implement and debug the DNS, of which
a significant portion derived from simple procrastination--just not getting around to installing
and implementing the DNS. Another delay included the difficulty of retrofitting the DNS into old
operating systems that were no longer actively maintained.
2.3.1. Caching

The design of the DNS allows for the existence of caches. These are local data storage
and/or memory that can significantly reduce the amount of network traffic associated with
repeated successful queries for the same data by providing access to the data in servers closer to
the end user than the authoritative name server.19 The data in these caches need to be refreshed at
regular intervals20 to ensure that the cached data are valid. In the initial version of the DNS
specification, several timing parameters had time-to-live limits of approximately 18 hours. It
quickly became apparent, however, that in many cases data changed slowly, and so updating
caches every 18 hours or so was unnecessary. As a consequence, the protocol specification was
changed to increase the allowed range of these timing parameters; several other protocol
parameters were also given expanded ranges, based on the theory that one incompatible protocol
change early on would be better than a series of such changes. This happened early enough that
there was no serious difficulty in deploying upgraded software.
In its original design, the DNS did not have a corresponding mechanism for reducing the
network traffic associated with repeated unsuccessful queries (i.e., queries for which no entry in
the relevant authoritative table is found). Within a few years of the initial implementation of the
DNS, it became apparent that such a mechanism would be beneficial, given the number of
identical queries that are unsuccessful. A proposed mechanism for negative response caching
was developed, and the data necessary to support it were added to the protocol in a way that did
not affect software based on earlier versions of the protocol, but the full deployment of the new
mechanism was slow. The name server side of the new mechanism was very simple and was
deployed fairly quickly, but initial support for the client (user) side of the negative caching
mechanism was limited to a few implementations and was not adopted more generally until much
later. The lack of widespread and correct client-side support for negative caching is a problem
that still persists.21
2.3.2. Lookup Timeouts

The biggest single difficulty in the transition from HOSTS.TXT to the DNS, however,
was not due to any specific shortcoming of the DNS. Rather, it was attributable to the
fundamental change in the nature of the lookup mechanism. In the HOSTS.TXT world, any
particular host lookup operation would either succeed or fail immediately--the HOSTS.TXT file

18 Most or all of these conflicts were internal ones--for example, subunits of a university trying to obtain
the same domain name as the university.
19 An additional potential benefit from the use of caches is an improvement in user response time.
20 As defined in the time-to-live field in the resource records. See Chapter 3.
21 Users derive benefits from the implementation of negative caching, namely faster response times. The
larger system also derives benefits through the reduced load of invalid queries. However, there are costs
associated with the implementation and maintenance of negative caching. For a given user, if the estimated
benefit deriving from faster response times is deemed to be worth less than the costs associated with
negative caching, then the user is not likely to implement negative caching, even though the total benefits
(which include the reduced load of invalid queries on the larger system) may exceed these costs. This
phenomenon is explained under the rubric of what economists refer to as externalities.


Prepublication Copy--Subject to Further Editorial Correction
is located on the user's system; it is not dependent on Internet connectivity at the moment of
lookup. The DNS added a third possible outcome to any lookup operation: a timeout attributable
to any of a number of possible temporary failure conditions (e.g., the required name server is
down, so one does not know whether the particular name is indeed in the table or not). The
occurrence of a timeout indicates neither success nor failure; it is the equivalent of asking a yes or
no question and being told "ask again later." Many of the network programs that predated the
DNS simply could not handle this third possibility and had to be rewritten. While this was
something of a problem for programs intended to be run directly by a user (e.g., one then-popular
e-mail client checked the host name of every recipient during composition), it was a far more
serious problem for programs that ran unattended, such as mail transfer agents. These programs
had to be rewritten to handle DNS timeout errors in the same way as they would handle any other
form of connection failure. Conceptually, this was simple enough, but it took several years to
actually track down and fix all the places in all the programs that made implicit assumptions
about the host lookup mechanism. Toward the end of this period, the Internet had entered an era
of periodic "congestion collapses" that eventually led to a fundamental improvement in certain
algorithms used in the Internet infrastructure. During each of these congestion collapses, DNS
lookups (along with all other forms of Internet traffic) frequently timed out, which made it much
more obvious which applications still needed to be converted to handle timeouts properly. To
this day, however, correct handling of the possibility of timeouts during a DNS lookup represents
an issue in application design.
2.3.3. Convergence in Electronic Mail Systems

In the mid-1980s, the Internet was one of the major data networks.22 Although data could
not move from one network to the next, e-mail was able to flow--through carefully designed e-
mail gateways--between the networks. Some of the busiest computers on each network were the
machines whose job it was to relay e-mail from one network to the next.23 Unfortunately, the
system of gateways required users to route their e-mail by explicitly using the e-mail address.
For instance, to send e-mail over the Internet to a colleague at Hewlett Packard Laboratories on
the Computer Science Network (CSNET), one had to address the e-mail to This complex syntax says that the Internet should deliver
the e-mail to and then send the message on to the appropriate address on
Thus, some people had one e-mail box yet had business cards listing three or four different ways
to send e-mail to them. There were ample opportunities for confusion and mis-routed e-mail.25
The original DNS specification tried to address this problem by making it possible to
send e-mail to names that were not connected to the Internet. For instance, if the Example
Company was on the UUCP network, but wanted to exchange e-mail over the Internet, it could

22 These major data networks included BITNET, Internet, CSNET, UUCP, and Fidonet. See The Matrix:
Computer Networks and Conferencing Systems Worldwide
(1990), John S. Quarterman, Bedford, Mass.:
Digital Press; and !%@: A Directory of Electronic Mail Addressing and Networks, 1991, Donnalyn Frey,
Buck Adams, and Rick Adams, Sebastopol, Calif.: O'Reilly and Associates.
23 For instance, and, the e-mail gateways between the Internet and the Computer
Science Network, and an important one of those between the Internet and the Unix-to-Unix network,
respectively, were typically the top two hosts (in terms of traffic sent or received) on the ARPANET in the
24 In some instances, the messages were even messier; someone on the Unix-to-Unix network (UUCP),
might have to write an address such as <ihnp4!seismo!> to send an e-
25 For instance, at Princeton University there was a weekly tape swap between the operators of
princeton.bitnet and princeton.csnet--two machines on different networks that routinely got mail
accidentally intended for the other.


Prepublication Copy--Subject to Further Editorial Correction
register and place an entry in the DNS directing that all e-mail to names ending should be forwarded to the UUCP e-mail gateway, which would know to forward
the e-mail to the Example Company's e-mail hub.
Unfortunately, the original DNS scheme for e-mail routing was not up to the task. It did
not handle certain types of e-mail routing well, and, worse, it could cause e-mail errors that
resulted in lost e-mail.
The result was a new scheme using Mail eXchangers (MXs) so that, for any domain
name, the DNS would store a preferentially ordered list of hosts that would handle e-mail for that
name. The rule is to start with the most preferred host in the list and work down the list until a
host is found that will accept the e-mail.26 This simple rule could be combined with the DNS
facility for wildcarding (following rules that state that all names ending in a particular domain, or
that a particular subset of names ending in a name, should all get the same response)27 to create e-
mail routing for almost any desirable situation. In particular, it was possible to address e-mail to
domains that were off the Internet, or domains that were partly on and partly off the Internet.
The development of a dependable and highly flexible mechanism for routing Internet e-
mail had two almost immediate consequences. First, all the other major e-mail networks
converted to using domain names or names that looked like domain names.28 These other
networks all had non-hierarchical (i.e., flat) name spaces, with many of the same scaling
problems that were experienced with HOSTS.TXT, and were looking for a workable hierarchical
name space. Once it was shown that the DNS would work for e-mail, it was simpler for
companies to adopt domain names and, in some cases, adapt the DNS to run on their networks
rather than to devise their own naming scheme. Thus, within a matter of 18 months to 2 years,
the Babel of e-mail addresses was simplified to user@domain-name, almost everywhere in the
world. A second effect was that companies could now change networks without changing their
host names and e-mail addresses, providing incentives for some companies to make the switch to
the Internet. Indeed, by 1990, almost all the networks that had offered services comparable to the
Internet were either gone or going out of business.29 Around the same time, companies began to

26 See "Mail Routing and the Domain Name System," RFC 974, Craig Partridge, January 1986.
27 For example, in the early days of e-mail connectivity to much of Africa, e-mail hubs were set up inside
the countries, serving all users there. These hubs did not have direct Internet connectivity to the rest of the
world but were typically served through occasional dial-up connections that, in turn, usually used non-
TCP/IP connections. To facilitate this arrangement, the DNS was set up so that all traffic for, say, South
Africa (the .ZA ccTLD), regardless of the full domain name, would be routed to a mail-receiving system in
the United States. That system would then open an international dial-up connection at regular intervals and
transfer all mail over it. The mail hub in South Africa would then distribute the mail to other hubs within
the country, often using the originally specified domain as an indication of the appropriate domestic server.
This model had the added advantage that, when permanent connectivity became available, user and
institutional e-mail addresses and domain names did not need to change--users just saw a dramatic
improvement in service and turnaround time. More information on the history of this strategy may be
found at <> and in John C. Klensin and Randy Bush, "Expanding International E-mail
Connectivity: Another Look," Connexions--The Interoperability Report 7(8), August 1993, pp. 25-29;
available at <>.
28 For convenience and as a transition strategy, many sites chose to treat, for example, "BITNET" and
"UUCP" as if they were top-level domain names, mapping those names through the DNS or other facilities
into gateway paths. So a generation of users believed that, for example, smith@mitvmb.bitnet was an
Internet domain name when, in fact, it was mapped to, where the
latter was a gateway between the Internet and BITNET. The full use of MX records, so that the same user
could be addressed as, came along only somewhat later.
29 In economics, network effects (or, alternatively, positive network externalities) explain the rationale for
the convergence to Internet-based e-mail: The value of a network to a user increases as more users join the
same network, or other networks that are compatible with it. See, for example: Jean Tirole. 1988. The
Theory of Industrial Organization
. Cambridge, Mass., and London: MIT Press, pp. 404-409.


Prepublication Copy--Subject to Further Editorial Correction
encourage their employees to put e-mail addresses on business cards (it had often been
discouraged because, as previously noted, e-mail addresses were so complicated). Domain names
thereafter became a (small) part of everyday business.
There were also some longer-term effects. First, it was very clear that the DNS could
provide names for things that were not hosts. For instance, almost every organization soon made
it possible to mail to, even though there was no actual machine
named, but rather a collection of servers (e.g.,,, and so on) that handled e-mail for the domain. The
DNS began to be viewed as a general naming system. Second, because almost all naming
systems had been designed primarily to support e-mail, and domain names had won the battle for
how e-mail was done, the other naming systems diminished in importance and use, leaving the
DNS as the only widely available naming system. The result was that the DNS was viewed as a
general service, albeit an imperfect one: but even if imperfect, it was the only naming system that
was widely available, and thus it became the one of choice.
2.3.4. The Whois Database

The Whois database was developed in the 1970s to track authorized ARPANET users
and, in particular, those users that could request addresses on the network. For each host or
domain name, the information in the Whois database was supposed to include the contact
information (such as the contact person's name, organization, street address, electronic mail
address, and phone number) of those with responsibility for the host or domain name; additional
information could also be stored and accessed. From a technical design and operational
viewpoint, the Whois database is independent of either the HOSTS.TXT file or the DNS.30 For a
while, the Whois database, maintained by the Network Information Center (NIC), served as a de
facto white pages directory of ARPANET users. Beyond the online database, the NIC printed a
phone book of everyone listed in the Whois database about once a year until 1982.
Around the time of the last NIC phone book, the Whois database was rapidly losing its
value as a white pages directory because many new Internet users were not being included in the
database. However, at the same time, the Whois database was becoming increasingly important
for network operations because the NIC (which at the time also managed the allocation of IP
addresses) would not give out an IP address, a host name, or a domain name to anyone who did
not have a Whois entry. Furthermore, the NIC put all address and name registrations into the
Whois database. So, given a host name or address, any user on the network could query the
database to learn who had control of that host name or address. Thus, if a network operator
noticed (or had a user complain) that a domain name suddenly could not be looked up, or that a
particular network appeared to be unreachable, the operator could query the Whois database and
find out whom to call about the issue.
By the late 1980s, problems began to develop with the Whois database. The first
problem, which proved easy to solve, was that in many cases the formal institutional contact for
the name or address was a corporate or university officer or administrator and was not the
network operations person who actually operated the network or domain name server. The NIC
resolved this problem by updating the database to keep track of both the administrative and
operational contact for each address and domain. The second problem, which was not so easy to

30 "Whois" represents both the name of the implemented system/database as well as the name of the
underlying protocol. This caused, and continues to cause, some confusion since several universities and
enterprises maintained local "white pages" and similar services, which had nothing to do with the central
databases, that were accessed using the Whois protocol.


Prepublication Copy--Subject to Further Editorial Correction
solve, was trying to keep the Whois data current--a problem that existed even before the
explosive growth of the Internet and demands on the DNS in the 1990s.31
2.3.5. The DNS as a Production System

By 1990, the DNS was a production system and deeply ingrained in the Internet and its
culture.32 The use of HOSTS.TXT was declining rapidly. But the move to a production system
was not easy: Deploying the DNS in the 1980s required several years of debugging and resolving
various issues. Timeouts and negative caching remain, to some extent, open issues in 2005.
Several lessons are apparent from the process of developing and deploying the DNS. A
good new design that solves important problems can catch on, but it will take time for solid
implementations to be developed. And even if a new design offers significant advantages,
adoption will take time. Even when the Internet was comparatively small, switches from
HOSTS.TXT to DNS or from e-mail Babel to uniform naming took a significant amount of time.
Given the decentralized nature of the Internet, network service providers, hardware and software
vendors, end users, or others can inhibit worthwhile technical advances from being implemented
through mere procrastination or a deliberate decision that the implementation of a particular
software upgrade is simply not sufficiently beneficial to them. Given the much larger scale and
scope of the DNS and the embedded base of software two decades later, successful
implementation of any proposed new system or major changes to the existing DNS may prove


The increasing popularity of personal computers changed the basic model of computing
in most organizations from a model based on central computing using mainframes or
minicomputers with terminals to one based on personal computers connected in local area
networks, which in turn were connected to central resources (i.e., the client/server model of
computing). The adoption of the personal computer by consumers (which is correlated with the
improving price/performance of computers and, in particular, increasing modem speeds at
affordable prices) provided the household infrastructure for supporting widespread dial-in access
to the Internet by the mid-1990s in the United States.34

31 The history of the Whois database through the 1990s can be found in Section 2.5.3.
32 By the late 1980s, the Internet was in fact an operational network and not only a subject of research, and,
as such, increasingly fell outside DARPA's research mission. At this time, DARPA was working with
other federal agencies, notably the National Science Foundation, to hand off the infrastructure it had
33 See Tirole, The Theory of Industrial Organization, 1988, pp. 406-409, for a brief discussion of the kinds
of coordination and strategic issues that can arise in a network like the DNS.
34 According to the Current Population Survey (conducted by the U.S. Census Bureau), personal computer
adoption in the United States continued to increase throughout the 1990s and demonstrated a five-fold
increase from 1984, the first year data was collected on computer ownership to the year 2000. By the year
2000, 51 percent, or 54 million households, had access to at least one computer at home, up from 36.6
percent in 1997. Households with Internet access more than doubled between these years, from 18 percent
in 1997 to 41.5 percent, or 42 million households by the year 2000. Computer access and Internet access
were becoming synonomous: more than four in five households with computer access also had Internet
access. For the full report, see Eric C. Newburger, September 2001, "Home Computers and Internet Use in
the United States: August 2000," Current Population Reports, Washington, D.C.: U.S. Department of
Commerce, U.S. Census Bureau. Available at < >.


Prepublication Copy--Subject to Further Editorial Correction
To function on the Internet, a computer needs to have some basic information, such as its
IP address, the IP address of at least one router,35 and the IP addresses of a few critical services.36
In the world of a relatively small number of large mainframes or minicomputers, such
information was entered manually on each new computer when installed and, once configured,
rarely changed. In such a world, IP addresses functioned as de facto stable identifiers, with the
DNS (or its HOSTS.TXT predecessor) representing a convenience, not a necessity.37
However, as the number of computers increased sharply, such a custom approach became
increasingly impractical. Thus, a mechanism to automate this startup process was developed.
One approach is contained within the Bootstrap Protocol (BOOTP), a very simple protocol that
enabled a computer to ask a local central server for and receive a number of critical parameters.
BOOTP and other protocols of a similar type shared one important characteristic: Each protocol
had a mechanism to allocate IP addresses to computers, but did not have any mechanism to
reclaim IP addresses when they were no longer needed. In the 1980s, this was not a problem
because IP addresses were plentiful. However, by the early 1990s IP addresses, which had once
seemed to be a nearly inexhaustible resource, were starting to look like a scarce resource that
required conservation, a consequence of the tremendous growth of the Internet. Protocols to
support the "leasing" or temporary assignment of IP addresses were developed,38 such as the
Dynamic Host Configuration Protocol (DHCP)--a direct successor of BOOTP--or the Point-to-
Point Protocol (PPP).39 An important reason for the development of these protocols was to
support system and local area network (LAN) management and auto-configuration, but the timing
was fortuitous inasmuch as these protocols could also help with the conservation of IP addresses.
The spread of network address translators (NATs)--in part, a response to the increased
difficulty of obtaining large blocks of IP addresses in the latter half of the 1990s--further
degraded the usability of IP addresses as stable identifiers. The basic function of a NAT is to
rewrite IP addresses in the data that it forwards. NATs map the set of IP addresses for external
traffic (i.e., the IP addresses that are visible to the world) to a set of IP addresses for internal
traffic (e.g., an organization's LAN); thus, an organization can have many more internal IP
addresses than external ones. The use of NATs distorts the one-to-one mapping between Internet
hosts and IP addresses that many applications assumed in their design--thus, any application that
depends on IP addresses is at risk when its traffic goes through a NAT.40
As a result of these changes, IP addresses have become much less useful as stable
identifiers than they once were. In the case of most application protocols, the "obvious" answer
has been to replace the use of IP addresses with DNS names wherever possible. Thus, over the

35 A router is a device that determines the next Internet Protocol (IP) network point to which a data packet
should be forwarded toward its destination. The router is connected to at least two networks and
determines which way to send each packet based on its current understanding of the state of the networks to
which it is connected. Routers create or maintain a table of the available routes and use this information to
determine the best route for a given data packet.
36 Examples include the address of an e-mail server (because most computers do not operate their own mail
server) and the address of a DNS resolver (explained in Chapter 3).
37 Indeed, the IP addresses of certain important servers were well known to system administrators.
38 After the lease expires, ownership of the address reverts back to the server that issued the address. The
protocol includes mechanisms for lease renewal, and lease times can be quite long at the discretion of the
DHCP server administrator. These "leases" did not include a financial component--"temporary
assignment" is perhaps a more accurate characterization.
39 PPP supports address assignment for dial-up networking by assigning IP addresses to ports on access
servers. Users connect to the access server and are allocated to a port, which has an IP address assigned to
it. Thus, users "lease" the assigned IP address for the duration of their session.
40 In particular, peer-to-peer applications and security protocols that require different public addresses for
each host become much more difficult to deploy.


Prepublication Copy--Subject to Further Editorial Correction
last decade, applications have come to rely on DNS names very heavily as stable identifiers in
place of IP addresses.
Another departure from transparent architecture41 came with the introduction of packet-
filtering routers, one of the simplest kinds of firewalls.42 A number of organizations introduced
such firewalls beginning in the late 1980s with the intent to defend their sites against various real
and perceived threats. The much-publicized Morris worm43 further raised the profile of network
security and provided network administrators with an additional motivation to install firewalls
(thereby further inhibiting transparency in the network architecture).44
As network security attracted increasing attention, some focus was directed to the DNS
itself. DNS security emerged as an issue in the form of a proposed addition of a cryptographic
signature mechanism to the DNS data.45 Such a mechanism would help ensure the integrity of the
DNS data communicated to the end user. The original DNS design did not include a mechanism
to ensure that a name lookup was an accurate representation of the information provided by the
entity responsible for the information. DNS information was assumed to be accurate as the result
of general notions of network cooperation and interoperation (i.e., based on the presumption that
nobody would deliberately attempt to tamper with DNS information). Development work on
DNS security extensions (DNSSEC) started in the early 1990s and continues a decade later.46
See Chapter 4 for further discussion of DNSSEC and DNS security in general.


The nature of the growth in the Internet during the 1990s was qualitatively different from
the growth in the 1980s. Most of the new Internet users in the 1990s were non-technical people
who were not associated with academic institutions or the computer and communications
industry. Instead, these new users represented a cross section of society that typically accessed
the Internet from their places of employment, through dial-up connections from their homes, and
by gaining access through libraries, schools, and community organizations.

41 In this context, a transparent network is one that does not interfere with arbitrary communication between
end points.
42 Packet-filtering routers attempt to block certain types of data from entering or leaving a network.
43 In 1988, a student at Cornell University, Robert T. Morris, wrote a program that would connect to
another computer, find and use one of several vulnerabilities to copy itself to that second computer, and
begin to run the copy of itself at the new location. Both the original code and the copy would then repeat
these actions in a theoretically infinite loop to other computers on the ARPANET. The worm used so many
system resources that the attacked computers could no longer function, and, as a result, 10 percent of the
U.S. computers connected to the ARPANET effectively stopped at about the same time. From "Security of
the Internet," available at <>. Also published in The
Froehlich/Kent Encyclopedia of Telecommunications vol. 15
. Marcel Dekker, New York, 1997, pp. 231-
44 The relative value of firewalls in advancing network security can be debated and such discussions can be
found elsewhere; for example, see Fred B. Schneider, editor, Trust in Cyberspace, 1999, Computer Science
and Telecommunications Board, National Research Council. Washington, D.C.: National Academy Press.
45 Digital signatures do not provide foolproof security, but they can demonstrate that the holder of the
corresponding private cryptographic key (i.e., a secret password) produced the data of interest. This is
more or less like trusting a document that bears a particular seal--one must independently make a
determination that an authorized person had possession of the seal when it was used and that the seal is
legitimate but, if both of those conditions are met, it provides some assurance of the authenticity of the
46 For further information on the historical progression of DNSSEC, see Miek Gieben, "A Short History of
DNSSEC," April 19, 2004, available at <>.


Prepublication Copy--Subject to Further Editorial Correction
These new users and the organizations that supported them (such as Internet service
providers, electronic commerce companies, non-profit information services, and so on.) were
primarily interested in how the Internet in general, and Internet navigation and the Domain Name
System in particular (especially using the World Wide Web and e-mail), could advance and
support personal and business goals--that is, they were not very interested in the technology per
se. Consequently, increasing effort was directed to support these non-technical goals, and thus, it
is not surprising that economic value, social value, and globalization emerged as major forces
influencing the DNS and Internet navigation in the 1990s.
2.5.1. Demand for Domain Names and Emergence of a Market47

The rapid growth of the World Wide Web stimulated interest in and the demand for
domain names because Web addresses (Uniform Resource Locators; URLs [see Box 6.2])48
incorporate domain names at the top of their naming hierarchy. One of the early major uses of
the Web that appealed to a wide range of the new users--and helped to continue attracting
additional new users to the Web--was electronic commerce. The .com generic top-level domain
(gTLD) became a kind of directory service for companies and their products and services. If a
consumer wanted to find the Web site for a company, the consumer would often be able to guess
the URL by entering part or all of the company's name followed by .com in the browser
command line; often, the desired site would be located. This practice was further encouraged by
the use of second-level domain names in advertisements and by the naming of companies by their
second-level domain name (e.g., Even if the user's initial guess(es) did not work,
users would often then try the company's name followed by .net or .org, or variations of the
company's name in combination with one of these gTLDs, such as
It did not take users long to discover that shorter, shallower, URLs were easier to guess,
use, remember, and advertise than longer ones. The shortest URL of all was based solely on a
domain name. Thus, if one wanted to post a distinct set of resources on the Web, or create an
identity for an organization, product, or idea, it often made sense to register a separate domain
name for it rather than create a new directory under a single domain name. Hypothesizing that
customers would look for products and services by guessing at a similar domain name, companies
like the Procter & Gamble Company, for example, registered and used that as a
URL (namely, <>), which also had the advantage of being much easier
to communicate to users, and for users to remember, than, say,
<>. These different domain names would be used
even if all the information resided on a single computer. In short, domain names began to refer to
products or services rather than just network resources (e.g., host names).50
Before the rise of the Web, the largest concentration of domain name registrations was
under the .edu TLD (as of March 1993). The Internet's rapid growth after 1993, however,

47 A significant portion of this subsection was derived from Mueller, Milton L. 2002. Ruling the Root:
Internet Governance and the Taming of Cyberspace
. Cambridge, Mass. and London: MIT Press.
48 Examples of URLs include <> or <>.
49 Sometime in 1996 or 1997, browser manufacturers made .com the default value for any names typed
directly into the browser command line. That is, whenever a user typed <name> without a top-level
domain into the command line, the browser automatically directed the user to www.<name>.com. Making
.com the default value for all browser entries reinforced the value of .com registrations relative to other
TLDs. In effect, a .com domain name functioned as a global keyword, and the possession of a common,
simple word in the .com space was sure to generate significant traffic from Web browsers. This explains,
to some degree, why some domain names sold for hundreds of thousands or even millions of dollars. As
noted in Section 1.2, footnote 9, browsers no longer operate in this way.
50 Generic words were also registered (e.g., was registered by Vicks).


Prepublication Copy--Subject to Further Editorial Correction
radically altered the distribution of domain names across TLDs. Until at least 1997, .com
attracted the large majority of new domain name registrations (see Table 2.1).51 Most of the users
rushing to take advantage of the Web were businesses, and .com was the only explicitly
commercial top-level domain. Furthermore, the U.S.-based InterNIC operated the only
unrestricted, large-scale registry (supporting .com and other gTLDs). Most country-code
registries at this time were slow, or expensive, or followed restrictive policies and considered a
domain name a privilege rather than a commercial service.52 Indeed, in some of the countries
with restrictive country-code registries, such as Japan and France, more businesses registered in
.com than under their own ccTLD. The available statistics provide the basis for estimating that
roughly 75 percent of the world's domain name registrations resided in .com at the end of
1996.53 Thus, the .com TLD became the dominant place for domain name registration
worldwide in the mid-1990s, which by the late 1990s became reflected in popular culture through
phrases such as a "dot com company" (or simply a "dot com") or "dot com economy."

TABLE 2.1 Number of Second-Level Domain Names Registered in the .com, .edu, .org,
and .net Top-Level Domains, 1993-1997

March January January February January
11,000 30,000 232,000 796,000
1,400 2,000 2,000
1,100 3,000 17,000
400 2,000 11,000
13,900 37,000 262,000 896,000

NOTE: It is worth noting that the number of second-level domain names does not correlate with the
number of host computers. For example, the .edu TLD has maintained a relatively vertical structure, and
so even though a university may be adding many hosts to its network, there will be minimal or no impact
on the number of second-level domain names.
SOURCE: A.M. Rutkowski, "Internet DNS Historical Timeline: Official Actions and Statistics," presents
a timeline-style chronology, official actions, and statistics, plus other meetings and events occurring
between the DNS's inception in 1985 and today. Copyright A.M. Rutkowski. July 30, 1997. Available at

51 There were a number of initiatives to use the Internet for commercial purposes (including research and
development activities within companies) prior to the rise of the Web. Such initiatives caused the number
of .com registrations to increase. However, the wide use of the Web caused the registrations of multiple
domain names to single companies, a practice that had been discouraged in the past.
52 In February 1996, when the Internic had about a quarter of a million second-level registrations, Germany
(.de) had only 9000 total registrations, and Great Britain (.uk) had only 4000. Japan, Canada, Australia,
and other major leading participants in the Internet had numbers comparable to the United Kingdom's.
However, some countries (e.g., the United Kingdom) have restrictive policies with respect to registering in
the second-level domain so that most entities actually have to register in the third-level domain (e.g., that would instead be a second-level domain registration in other TLDs (e.g.,
53 InterNIC gTLD registrations accounted for an estimated 85 percent of all domain name registrations
worldwide, and .com accounted for 88.6 percent of all Internic gTLD registrations. (About 62 percent of
all registered domains worldwide resided in .com in 2002.)


Prepublication Copy--Subject to Further Editorial Correction

Interest in domain names extended beyond the for-profit sector. The visibility of
governmental entities and non-profit organizations also became increasingly tied to domain
names as the Web became a key mechanism for providing information and services to the public
and their constituencies. Moreover, individuals also wanted their own domain name as the
identifier for their personal information posted on the Web or for a myriad of other purposes (e.g.,
establishing fan sites54). Opportunistic companies capitalized on (and perhaps helped to create)
this demand by developing services so that users could register a domain name and obtain support
for establishing and maintaining a Web page as an integrated service for a monthly or annual fee.
Domain names also became involved in electoral politics and social commentary.
Political campaigns established Web sites with descriptive domain names in the URLs such as
<>, <>, or
<>55 to provide access to information about their respective candidates,
to organize volunteers, and to solicit contributions. Web sites were also created to critique or
parody virtually anything, from the practices of certain companies or their products or services to
various social and political causes. A common practice was to register a domain name that
included the name of interest followed by "sucks," or something similar, and to associate that
domain name with a Web site that criticized the entity in question. In addition to motivating legal
actions to try to prevent the use of domain names in this fashion, this practice caused many
companies to pre-emptively register these types of domain names for themselves.56
Therefore, for various reasons, the demand for domain names increased tremendously
during the 1990s.57 Further fueling the demand was aggressive marketing by companies that
register domain names and provide related services, efforts by IT companies more generally that
played up domain names (especially .com names) in their larger marketing campaigns, and the
popular and technical press, which devoted a lot of attention to anything related to domain names.
The increasing demand for domain names was attributed to interest in facilitating Internet
navigation as well as to the value of domain names irrespective of their functional utility on the
Internet (e.g., placing a domain name on posters in a subway station as a part of a marketing
campaign). Thus, the real value of certain domain names in the rapidly growing and
commercializing Web and Internet was far greater than the price of setting up a domain name
(which was on the order of $50 at the retail level).58 The predictable consequence was the
development of an aftermarket for certain domain names. In 1996, sold for $15,000 and
in 1997, changed hands for $150,000.59 Not surprisingly, speculation in the
registration of domain names took place: An individual or firm would register domain names
(often very many) with the intent of reselling them to others for a premium. Such speculators
would not only register generic or descriptive names (e.g., "business," "fever," and so on) with
the hope of appealing to multiple prospective purchasers, but would also register domain names
incorporating the trademarks of third parties with the hope that the corresponding trademark

54 See, for example, <> as a fan site and tribute to the actress Julia Roberts that is
based in Germany. Accessed on March 27, 2004.
55 Note that campaign information was not available at the Al Gore site as of March 25, 2004.
56 However, this was a difficult proposition, considering the nearly limitless variations of less-flattering
names that can be devised. See further discussion in the next section.
57 The growth in the registration of domain names was phenomenal. For example, in September 1995,
there were approximately 120,000 registered domain names. By May 1998, 2 million domain names were
registered. "Fact Sheet: NSF and Domain Names," National Science Foundation, Arlington, Virginia,
accessed on March 27, 2004, available at <>.
58 Domain names are registered through and maintained by registrars; see Chapter 3 for an extended
59 was resold for $7.5 million in 1999. With some irony, the committee observes that links to "The Business Search Engine" (as of March 27, 2004), a "comprehensive
business directory."


Prepublication Copy--Subject to Further Editorial Correction

BOX 2.1
The Value of Domain Names

Semantic Value
Economic value often arises when names have some semantic distinction--a meaning--and are
visible in a public arena. The value of meaningful names will differ from nil to very high
depending on the meaning and the potential application. Examples include and

Mnemonic Value
In many commercial contexts, it can be important for a name or identifier to be easily
remembered. If users cannot remember the name, or it is too long or complicated to reproduce,
the object will not be found. Memorability, and in some cases guessability, facilitates more
incoming traffic and more business and, therefore, gives rise to economic value. One example is (i.e., a second-level domain name for the General Motors Corporation).
Personal Value
Even when there is no apparent commercial consequence, the human desire to make a statement
and assert an identity can give economic value to identifiers in a public name. Someone with a
high regard for himself might want the domain name

Stability Value

Users can accumulate equity in a particular identifier, which becomes closely associated with
them and expensive to change. Changing a telephone number or e-mail address that has been
used for many years can be difficult because of the large number of personal contacts and records
that contain the number. Thus, equity in an identifier raises switching costs for consumers,
making them more likely to stay with the provider of that identifier.
Pure Scarcity Value
Meaningless identifiers, such as bank account numbers, function economically as an
undifferentiated resource pool. They may possess economic value by virtue of their scarcity, but
no one cares which particular identifier he or she gets. In the DNS, there are plenty of possible
names (accepting that random strings of letters and digits can produce domain names (e.g.,; there is scarcity of some desired names, with
desirability defined by one of the reasons above (e.g., only a small subset of all possible domain
names have semantic value).

owner would purchase the domain name from the speculator as well. See Box 2.1 for further
discussion on the value of domain names.
An industry emerged to provide services related to the transfer and assignment of domain
names and related services.61 The pressure for new TLDs led to the conversion of some country
codes to quasi-generic TLDs. For example, the marketing of .cc (a country-code TLD created
to represent the Cocos Islands but later marketed as a de facto gTLD) further increased the variety

60 As of March 27, 2004, this domain name was not registered.
61 Registrars, and the industry surrounding registrars and allied services, are discussed in detail in Chapters
3 and 4.


Prepublication Copy--Subject to Further Editorial Correction
and value of certain domain names. However, true additional gTLDs did not materialize in the
1990s (despite the intense arguments and efforts made by some individuals and organizations),
which helped to solidify the dominance of the then-extant TLDs, especially of .com.62
2.5.2. The Rise of Conflicts Over Domain Names

As the number of domain name registrations exploded, conflicts developed over the right
to register particular names at the second level of many of the TLDs.63 The basis for most of
these conflicts derived from the unique naming associated with the DNS. The DNS does not
have the capability to incorporate context into domain names, so each domain name must be
unique worldwide and then in turn, a Web site at that domain name can then be accessed
throughout the world.64 The consequence is that it is significantly harder to pick a domain name
that is both unused and memorable.
Claims to rights to domain names can be based on a number of different legal, political,
economic, ethical, or cultural criteria. A common problem raised by all such claims is that any
reasonable attempt at resolution must balance the value of avoiding confusion or preventing
illegal or otherwise undesired appropriation of identity against the value of free expression, open
communication, and fair use. Whenever semantics and economics enter the picture, however,
society and all of its conflicts come along with them. The resultant conflicts over who has rights
to use particular domain names has enmeshed and continues to enmesh apparently technical
naming processes in economic, public policy, and legal issues.65
Trademark Conflicts

In the commercial world, trademark law provides one of the oldest, most widely
recognized and well-developed regimes for recognizing and protecting exclusivities in the use of
source identifiers of goods or services within specific fields and territories. Trademark laws
developed, in part, to protect consumers against various forms of fraud, deception, or confusion
that might result from the ways in which products and services are identified. By giving
producers an exclusive right in an identifier, trademarks reduce consumer search costs and
channel the benefits of developing a good reputation to the producer responsible for the
reputation. Trademark protection is not, however, intended to give firms ownership of common

62 Discussion of the gTLDs added in the early 2000s (e.g., .info) and general discussion of the issues
involved with adding new gTLDs can be found in Chapters 3 and 4.
63 "Rights to names" refers to claims to exclusive or privileged use of an identifier based on the meaning or
economic value of the name.
64 The mythical (or perhaps real) Joe's Pizza illustrates the point. The DNS can support only one worldwide, but in the physical world, people can usually distinguish among the
(presumably) multiple Joe's Pizza restaurants around the world. If a person is located in the center of a city
and asks a taxi driver to take her to Joe's Pizza, it is presumed that she wishes to travel to a restaurant
within a few miles, not a Joe's Pizza located in a city 2000 miles away. Although the requestor does not
state this context explicitly, it is assumed in the conversation. As of February 4, 2005,
is registered by and offered for resale at a minimum price of $1488.
65 In his paper "External Issues in DNS Scalability," Paul Vixie argues for eliminating the disputes
associated with domain names by eliminating all meaningful top-level domains and replacing them with
meaningless alphanumeric identifiers. Conference paper, November 11, 1995, available at
<>. Whereas the number of conflicts would surely
decline, so too would the benefits to those who prefer specific domain names that have meaning or some
other characteristic. As with phone numbers, however, competition for certain numbers will ensue unless
they are assigned randomly and a secondary market (in either names or entities that control names) is


Prepublication Copy--Subject to Further Editorial Correction
words alone, to prevent non-commercial or fair use, or to inhibit public discussion of companies,
products, and services involving direct references to the mark.
The great emphasis placed on second-level domain names created a problem for some
trademark holders, especially for those that held trademarks that were well known in the United
States or worldwide. Trademark rights (which are traditionally accorded on a country by country
basis) generally arise through use of a trademarkable name, logo, color, sound, or other feature, in
association with the marketing and sale of particular types or classes of goods and services in
commerce. Some countries will allow someone to reserve or even register a trademark without
using it, but most countries (including the United States) require use to claim protection under
trademark law.66 Similar to trademarks, domain names can act as source-identifiers suggesting
the identity, quality, or source of a good or service. Accordingly, domain names may conflict
with trademarks because of their similarity in function.
If a trademark holder allows someone else to use the trademark or mark either in the
same class of goods/services or in some other way that could create confusion in the marketplace,
the other party may begin to acquire rights as a result of that use, and those rights reduce the
value of the mark to its original owner. Within the United States, the problem is further
complicated by the existence of federal and state antidilution laws that permit owners of famous
marks to enjoin others from commercial uses of such marks, even where such commercial uses
are in classes of goods or services different from that of the famous mark.67 Designed to prevent
the whittling away of the identification value of the plaintiff's mark, antidilution laws differ from
more traditional trademark laws, which are designed to prevent consumer confusion within the
same class of goods or services.68 In order to determine whether a mark is subject to federal
antidilution protection (i.e., whether it is considered to be a famous mark), the law provides
several non-exclusive factors that may be considered.69 Some states within the United States
have more generous definitions of "famous marks," while other countries have similar or
different definitions or do not recognize the concept at all. Another one of the factors to be used
in this determination under U.S. law is "the nature and extent of use of the same or similar marks
by third parties." This factor demonstrates that a potentially famous mark owner's antidilution
rights, like rights against infringement, can be lessened or lost by its failure to police its mark.
Thus, trademark holders became quite concerned about the activities of parties that had
registered domain names that either were confusingly similar to their trademarks or diluted their
famous marks. Generally, to cause a likelihood of confusion under trademark law,70 a domain
name must be used in connection with an offering of goods and services or it must be used
commercially for antidilution laws to apply, but cybersquatters raised a slightly different problem.
Cybersquatters are generally defined as domain name speculators who register a domain name
that incorporates a trademark owned by another party, not in order to use the domain name, but
with the intent of reselling the registered domain name to that party for an amount that far
exceeded the cybersquatter's registration cost. The most contentious examples of cybersquatting
were those in which domain names incorporating trademarks (or phonetic or typographical
variants of them) were associated with Web sites that included pornographic content or other

66 This also means that the failure to use a registered mark can result in the loss of rights.
67 See J. Thomas McCarthy, McCarthy on Trademarks and Unfair Competition § 24:88 (2003).
68 See id. An example would be if Mack Trucks, Inc. began using the mark "BIG MAC" in association
with a new model of truck. It is unlikely that the McDonald's Corporation would argue that there was a
likelihood of consumer confusion between Mack's truck and McDonald's large sandwich of the same
name. However, it is possible that the McDonald's Corporation would argue that the truck model was
diluting the value of its famous mark, BIG MAC.
69 These factors include "the degree of inherent or acquired distinctiveness of the mark," "the duration and
extent of use of the mark in connection with the goods or services with which the mark is used," and "the
duration and extent of advertising and publicity of the mark." See 15 U.S.C. § 1125(c)(1).
70 See McCarthy on Trademarks, § 25:76, at 25-237.


Prepublication Copy--Subject to Further Editorial Correction
information that would confuse or offend users who reached that Web site by mistake or assumed
that the trademark holder had registered the domain name and that the Web site associated with it
belonged to the trademark holder. The cybersquatter would then offer to sell the domain name to
the trademark owner for a substantial sum, which some companies agreed to pay, if for no other
reason than to make the offending use stop. Such actions, of course, only fueled more
cybersquatting activities. Cybersquatters also figured out that users were trying alternative
domain names (e.g., instead of and began to register many
different combinations, such as and Partially in
response to these actions, many companies began to register various combinations themselves--
so-called pre-emptive registrations71--and to seek remedies through the courts, various dispute
resolution processes, and legislative action.72
At least one court has held that a cybersquatter's registration of a domain name that
incorporated a plaintiff's famous mark, without the sale of any goods or services via the Web site
associated with that domain name, was a violation of the federal antidilution statute because the
defendant made "commercial use" of the trademark when he attempted to sell the domain name
back to the owner of the famous mark.73 Another court has found the attempted sale or licensing
of domain names containing trademarks to be trademark infringement, similarly concluding that
the sale of these domain names was a commercial use of the marks.74 Additionally, the
Anticybersquatting Consumer Protection Act (ACPA), enacted in 1999,75 provides additional
rights to trademark owners against those who register, traffic in, or use, with the bad-faith intent
to profit, a domain name that is identical or confusingly similar to a registered or unregistered
mark that was distinctive or dilutive of a famous mark when the domain name was registered.
Despite the creation of ACPA and other domain name dispute resolution mechanisms,76 the costs
involved with pre-emptive registrations and the enforcement of trademarks ultimately led many
representatives of trademark holder interests to resist efforts to create new TLDs, fearing that
these costs would continue to increase substantially if new additional TLDs were created.
In contrast, the protective efforts by trademark holders in some instances have also raised
conflicts with other legally equivalent rights held by the individuals using the domain names. For
example, suppose a group critical of a corporation wants to create a public space for discussion
and register a domain name associated with that corporation (e.g., Does
such a registration constitute an infringement of the corporation's trademark because it creates
consumer confusion or is it dilutive because it tarnishes the reputation of the corporation, or
rather is it an exercise of a protected right, such as freedom of speech under the First Amendment
of the U.S. Constitution? Discussions related to domain names, trademark concerns, and public
policy issues will continue into the 21st century.77
Other conflicts involving trademarks arose for reasons that had nothing to do with the
above-described conflicts between trademark holders and their cybersquatting antagonists. For
example, Chris Van Allen, then 12 years old, registered the second-level domain name because that was his nickname, and was subsequently ensnarled in a trademark
dispute with Prema Toy Company, owner of the trademarks on the claymation character Gumby

71 Such pre-emptive registrations were encouraged by the marketing strategy of many registrars.
72 In fact, the congressional mandate (P.L. 105-305) for this study came largely as a result of actions by
representatives of the trademark community. This community was very active and visible in the domain
names arena in the late 1990s.
73 See Intermatic, Inc. v. Toeppen, 947 F. Supp. 1227 (N.D. Ill. 1996).
74 See Toys "R" Us, Inc. v. Abir, 45 U.S.P.Q.2d 1944 (S.D.N.Y. 1997).
75 See 15 U.S.C. § 1125(d).
76 See Section 3.5 for an extended treatment.
77 The state of understanding continues to evolve. See, for example, The Taubman Company v. Webfeats,
et al.
, Nos. 01-2648/2725 (6th Circuit, February 7, 2003).


Prepublication Copy--Subject to Further Editorial Correction
and his horse Pokey, which wanted control over the domain name.78 Ultimately, Prema Toy
Company withdrew its complaint.79 In other cases, disputes arose between different entities with
equally valid rights to the same second-level domain name, such as "," which might
have been coveted by many men named Robert, and "," which might have been
coveted by a variety of different companies around the world that have legitimate rights to the
trademark Avon for different products or services in different countries.
This latter example presents a particularly difficult point that cannot be easily resolved by
simply granting the second-level domain name to the entity with the legal right to use it as a
trademark, since multiple entities can have such legitimate rights. Under most countries'
trademark laws, multiple entities can use the same trademark in the same country, provided each
entity uses the trademark for different categories (called classes) of goods or services, as long as
there is no possibility of confusing consumers as a result, and each trademark has to be registered
within each country in which it is used in order to be fully protected. Hence it is entirely possible
to have one company legitimately use the trademark "avon" for automotive tires in the United
States, and a second company to use the same trademark for cosmetics. Unfortunately, however,
there can be only one, and so the first entity to register that domain name has often
been able to use it to the exclusion of all other legitimate users worldwide.80
Beyond Trademark Conflicts

Trademark issues dominated domain name conflicts in the late 1990s and into the
beginning of the 21st century, but other conflicts also demanded attention. For example, some
governments asserted rights to control the assignment of country code TLDs and country names
and the registration of those names, even beyond the second level.81 Some governments assert
that this is an extension of national sovereignty. Similar claims may be asserted by ethnic groups
and indigenous tribes that have not achieved the political status of sovereigns, but that
nevertheless wish to protect or control the use of their collective name. In some jurisdictions, the
subunits of national governments, such as city administrations or port authorities, have claimed
exclusive rights to the use of their name in the DNS.82 In a similar vein, some international
organizations have asserted a right to prevent others from registering domain names identical to
their acronyms or names.83 These claims have more to do with the imputed legitimacy of the
association than with commercial confusion. Here, too, issues arise regarding the balance struck
between the use of the name as an identifier and its legitimate use as a reference to the identified

78 See "Short Take: Pokey Causes Net Trademark Uproar," Courtney Macavinta,, March 23,
1998, available at <>. The dispute also
involved Pokey's Network Consulting, which registered
79 See "Pokey Wins His Domain Name," Heather McCabe, Wired News, April 22, 1998,
80 See Chapters 3 and 4 for a discussion of how trademark conflicts over domain names can be (and should
be) managed.
81 See Principles for Delegation and Administration of ccTLDs presented by the ICANN Government
Advisory Committee, February 23, 2000, at <
82 See, for example, Excelentisimo Ayuntamiento de Barcelona v. Inc. WIPO Case No.
D2000-0505, available at <>; Salinas,
California, National Arbitration Forum, City of Salinas v. Brian Baughn, WIPO Case No.
FA0104000097076, available at <>.
83 For example, international organizations such as the International Monetary Fund (IMF) or the World
Health Organization (WHO). See The Recognition of Rights and the Use of Names in the Internet Domain
Name System
, Report of the Second WIPO Internet Domain Name Process. September 3, 2001. Available
at <>.


Prepublication Copy--Subject to Further Editorial Correction
entity. These claims also raise questions about who in the affected society has the right to control
the name. Also, some legal regimes, which are analogous to trademark law because they are
related to reputation in commerce, attempt to vest regions or localities, rather than specific firms
or products, with exclusive rights to a name for a certain use. These regimes of "controlled
appellations of origin" might be applied, for example, to wines or other agricultural products.84

In addition to nations, regions, and international organizations, many people feel that they
have some ownership right over their personal name and other aspects of their persona. National
systems of law often recognize "rights of personality" when defined as the ability of a person "to
control the commercial use of his or her identity."85 In the United States, there currently is no
federal right of publicity or privacy; rather, the promulgation of such laws has been left to the
states. About half of the states have recognized the right of publicity, either through common law
or statute.86 Other states provide similar protections as a part of the right of privacy.87 Under the
Restatement (Second) of Torts §§ 652A - 652C (1979), invading an individual's right of publicity
is similar to invading her privacy through unauthorized appropriation of her name or likeness.88
One of the primary motives behind passage of the Anticybersquatting Consumer
Protection Act in the United States, for example, was the widespread registration of the names of
U.S. politicians as domain names and their linkage to Web sites that were satirical or critical.89
Communications technology can create new arenas for disputes over rights to names. In
particular, the process of entering an identifier into a network creates numerous opportunities for
conflicts over the boundary of a name right. Of course, many of the underlying issues--
confusion, fraud, competition, fair use, freedom of expression--are familiar from other contexts.

84 For further discussion see, "Geographic Identifiers," in The Recognition of Rights and the Use of Names
in the Internet Domain Name System
, Report of the Second WIPO Internet Domain Name Process.
September 3, 2001. Available at <>.
85 See J. Thomas McCarthy. 1996. McCarthy on Trademarks and Unfair Competition, 4th ed., Clark
Boardman Callaghan, Deerfield, Illinois.
86 See, for example, Carson v. Nat'l Bank of Commerce, 501 F.2d 1082, 1084 (8th Cir. 1974) (recognizing
a common law right of publicity in Nebraska); FLA. STAT. ANN. §540.08 (West 2002) (providing for a
statutory right of publicity in Florida).
87 See, for example, Allison v. Vintage Sports Plaques, 136 F.3d 1443 (11th Cir. 1998) (describing the
appropriation of plaintiff's personality for a commercial use as an invasion of privacy tort in Alabama).
88 In 1953 in the case of Haelan Labs. v. Topps Chewing Gum, the right of publicity was first explicitly
recognized as a right independent of the right of privacy and as an individual's right to the publicity value
of his photograph. The court distinguished the right of publicity from the right of privacy because "many
prominent persons . . . far from having their feelings bruised through public exposure of their likeness,
would feel sorely deprived if they no longer received money for authorizing advertisements [or]
popularizing their countenances." See Haelan Labs. v. Topps Chewing Gum, 202 F.2d 866, 868 (2d Cir.
1953). Thus, the right of publicity has developed into a body of law distinct from, but related to, copyright
law, privacy rights, and the law of unfair competition. While certain states encode publicity rights within
their right of privacy statutes, prominent case law and jurisprudence acknowledge the development of the
right of publicity as an independent body of law. See, for example, Carson v. Here's Johnny Portable
, 698 F.2d 831, 834 (6th Cir. 1983) (stating, "[T]he right of privacy and the right of publicity protect
fundamentally different interests and must be analyzed separately."). When commercial exploitation of
names is involved, personality rights often overlap with, or are informed by a logic that parallels, trademark
rights. Indeed, a person's name is often registered as a trademark or used to brand products or services
(e.g., Michael Jordan). But rights of personality are often asserted even when commerce is not directly
89 U.S. Patent and Trademark Office, "Report to Congress: the Anticybersquatting Consumer Protection
Act of 1999." January 2000. A law passed by the state of California makes it illegal to register someone
else's name as a domain name "without regard to the goods and services of the parties." [Section 17525 of
the California Business and Professions Code]


Prepublication Copy--Subject to Further Editorial Correction
A good part of the advertising economy of the Internet is based on paying for "hits" (i.e., the
exposure of the content of a Web site to a distinct user).90 Thus, the practice of "typosquatting"
developed, wherein entrepreneurs registered domain names that were only a keystroke or two
apart from popular domains. These "typo" domains would then be linked to advertisements in
order to collect pay-per-hit revenue from people who mistyped the locator into the browser. The
cybersquatter John Zuccarini refined the practice of "typosquatting" to an art, registering
hundreds of close misspellings of popular domain names and trapping users into a parade of
cascading Web pages, some of them pornographic.91
There are even fuzzier boundaries to consider. There are businesses that register large
collections of expired domain names in order to collect advertising hits from people who are
looking for the old Web site. 92 Is this an abusive practice or one as innocent as putting up a
billboard on a choice spot on a busy highway?
Beyond Second-Level Domain Names

Thus far, the discussion has focused on second-level domain names. Although less
common, there are disputes involving third-, fourth- and higher-level domain names, as well as
involving directory and file descriptors. For example, in Bally Total Fitness Holding Corp. v.
, 29 F. Supp. 2d 1161 (C.D. Cal. 1998), an infringement suit was brought against a
defendant who used the URL <> to post critical comments
regarding the plaintiff. The court held that "no reasonable consumer" was likely to confuse the
defendant's domain name with the plaintiff's marks BALLY, BALLY TOTAL FITNESS, and
BALLY'S TOTAL FITNESS, because, among other things, the defendant did not use the
plaintiff's mark in his domain name.93 Based on the facts of the case, the court stated that the
result would have been the same even if the defendant's domain name was The court also contrasted the defendant's domain name and the
hypothetical second-level domain name with other cases where likelihood
of confusion was found when the plaintiff's mark was the only mark (e.g.,
used in the defendant's second-level domain.95
In another example, the Usenet newsgroup name space contains numerous descriptors
that use a variety of names to describe the space, including, for instance, the name Disney (e.g.,
alt.disney.disneyland or rec.arts.disney.parks). These newsgroups (which are visible to most
Internet users) are not run by the Disney Corporation, and the content and administration of the

90 See Section 7.2.2 for an extended discussion.
91 See, for example, "Typo-Loving Squatter Squashed," Joanna Glasner, Wired, October 31, 2000, available
at <,1367,39888,00.html>. In 2004, Zuccarini was sentenced to 30
months in prison for using misleading domain names to trick children into visiting pornographic Web sites
in violation of the federal Truth in Domain Names Act. See "U.S. Man Jailed for Luring Children to Porn
Sites," Reuters, February 26, 2004.
92 Other cases include attempts to protect the "nonproprietary" status of a name by excluding it from a
name space. The World Health Organization and World Intellectual Property Organization (WIPO)
proposed to do this with respect to International Nonproprietary Names (INNs), a list of over 3000 names
of pharmaceutical substances. See "International Nonproprietary Names (INNs) for Pharmaceutical
Substances," in The Recognition of Rights and the Use of Names in the Internet Domain Name System,
Report of the Second WIPO Internet Domain Name Process. September 3, 2001. This proposal is
particularly problematic because the list of INNs is not only long, but expands over time. Available at
<>. Religion is another potential source of rights
claims. Certain religions recognize words as sacred and attempt to protect or restrict their use.
93 See id. at 1163-65.
94 See id. at 1165.
95 See id. at 1165.


Prepublication Copy--Subject to Further Editorial Correction
group may or may not have the corporation's approval. In the even more freewheeling world of
AOL screen names, any user can appropriate the name of his or her favorite Disney character
(even in less than flattering variations) and use it as his or her screen name and e-mail address.
While it is clear that no exemption exists for Usenet groups and AOL screen-name aliases, it does
appear that trademark holders have chosen not to pursue many of these uses in these naming
Yet current law and policy regarding domain names erect major distinctions between the
various parts of the domain name used in a URL. Within the generic and most country-code top-
level domains, all (or at least most) of the political and legal conflict over rights to names takes
place over the second-level domain name. The third-level domain and all identifiers to the right
of the domain name are generally outside the scope of challenge through dispute resolution
processes.97 Current law and policy therefore regard the top-level domain as a fixed set of
generic categories or country codes, the second-level domain as the identifier of an organization,
product, or Web site, and the third-level domain as part of a "private" naming system, wherein
assignments can generally be left to the discretion of whoever holds the second-level name.
To further illustrate this point, Yahoo! Inc. has been an active defender of its brand name
in cyberspace. It has challenged the registration of hundreds of second-level domains, including
some rather remote misspellings, such as "jahu" or "yhuu," whenever they appear in the second
level of a domain name. But under current legal precedent, it would likely take no action against
a name such as In all likelihood, however,
Yahoo's decision not to pursue claims for trademark infringement or dilution for alternative uses
of its brand name and mark is less influenced by current precedent than it is by Yahoo's
likelihood of success on the merits, especially in view of decisions such as that in the above noted
Bally Total Fitness Holding Corp. v. Faber
By contrast, second-level domain names are ripe for generating conflicts over rights to
names. They are meaningful, they are perceived as being economically valuable, and they are
part of a global, public naming system administered via collective action. And perhaps most
importantly, they are susceptible to centralized control because of the existence of a single,
central point of coordination, the relevant registry. See Box 2.2.

96 Many trademark holders have not done anything regarding many newsgroup names, in part, because such
activities are difficult to police as well as prove that trademark infringement and/or dilution has occurred.
Indeed, as soon as a name was removed or changed in this space, another of the millions of users could
create a new one.
97 There are some important exceptions, such as the case of .uk, for which most entities register at the
third level (e.g., rather than at the second level. For these exceptions, it is the fourth
level and beyond that are outside the purview of dispute resolution processes. Dispute resolution processes
are further described in Chapters 3 and 5.


Prepublication Copy--Subject to Further Editorial Correction

BOX 2.2
The Institutionalization of .com

For the most part, the initial dominance of .com among the TLDs was a historical
accident, a product of the chance conjunction of the commercialization of the Internet, the rise of
the Web, the openness of the InterNIC registry relative to the ccTLD registries, and the lack of
any other commercially oriented TLD in the original set of gTLDs. Once .com became
established as the most desired TLD for many registrants, other forces contributed to the
solidification of .com's increasing dominance. As discussed elsewhere in this section, "Beyond
Second-Level Domain Names," the rise in value of .com names (whether for navigation or
marketing functions) led to the registration of some domain names for speculative, abusive, or
preemptive purposes. Based on a desire to avoid further registration of domain names for these
same purposes in new TLDs, some resistance developed to the creation of new TLDs, thereby
reinforcing the focus on extant TLDs (with disproportionate advantage to .com , given its
dominant market position). Whether the historical dominance of .com from the mid-1990s will
continue in the future of the DNS is discussed in Section 5.4.

2.5.3. Whois

In concert with the rise in the interest in and demand for domain names was a
corresponding increase in the value of contact information associated with domain names.
Hence, interest in the Whois database continued to rise in the 1990s.
Some of the targeted uses of the Whois data were for old-fashioned marketing
purposes--for example, to send marketing brochures and to make telephone solicitations to
network operators and domain name registrants. As domain names became economically
valuable after 1995, accessing Whois data also became a popular way to find out which domain
names were taken, who had registered them, and the creation and expiration date of the
registration. The Whois database also became an investigation and monitoring tool for
intellectual property rights holders. When a trademark holder discovered a potentially infringing
domain name, the trademark holder could use the Whois database to identify, investigate, and
contact the registrant of the domain name. At that time, the Whois database could also be used to
determine if the same registrant had registered any similar domain names that the trademark
holder did not know about or to search for further evidence of cybersquatting by the registrant.
Trademark holders also discovered that they could use the database proactively to perform
searches for character strings that matched trademarks, and pull up many of the domain name
registrations in the generic top-level domains that matched or contained a trademark. This
automated searching function proved to be so valuable that trademark interests began to demand
that the Whois functions be institutionalized, expanded, and subsidized, including the right to
purchase the complete list and contact data for all of a registrar's customers. The first World
Intellectual Property Organization (WIPO) domain name process, initiated in 1998 in response to
a U.S. government request, as detailed by a U.S. Commerce Department white paper,98

98 For the text of the white paper see <>.


Prepublication Copy--Subject to Further Editorial Correction
recommended that the contact details in a Whois record be contractually required to be complete,
accurate, and up-to-date, on penalty of forfeiture of the domain name.99


Worldwide interest in the DNS developed during the 1990s along with increasing
concern about U.S. dominance of a critical element of global communication and a commercial
resource on which other nations foresaw their economies and societies becoming ever more
dependent. With increasing recognition of this value came a growing desire to participate in the
management and policy decision making with respect to domain names.
One of the issues of particular interest in many countries is access to the Internet and the
DNS using home-country languages and scripts. The DNS was designed and developed based on
the use of Roman characters, generally in the English language. As the number of users in
countries with first languages that are not based on Roman characters increased dramatically
through the 1990s, interest developed in domain names based on non-Roman scripts (e.g.,
Chinese, Hebrew, Arabic, and so on). Several major efforts were undertaken to accommodate
internationalized domain names (IDNs) within the Internet infrastructure.100
Unfortunately, the design of the DNS presents formidable technical challenges for the
accommodation of languages that use non-Roman characters. As a lookup system, the DNS must
be able to determine unambiguously whether there is a match with a query or not. Comparing
strings is much more difficult than most people realize, because the definition of what is "equal"
is often not deterministic. For example, consider the case of the French language in Canada and
in France, for which there are different rules as to whether an accent stays over a character when
it is converted from lower to upper case.101 And some languages (e.g., Chinese) cannot even be
reduced to a relatively small number of standardized characters (e.g., the character set for
English). See Section 4.4 for further discussion of the IDN issue and the increasing interest and
involvement by parties outside the United States in matters related to the DNS.


In the 1980s, the Network Information Center (NIC) managed the registration of domain
names, operating under the auspices of SRI International and funded by the Department of
Defense (DOD), by DARPA and the Defense Information Systems Agency (DISA).102 Jon Postel
and other researchers at the Information Sciences Institute at the University of Southern
California had been given the authority to establish procedures for assigning and keeping track of

99 See paragraph 73 of The Management of Internet Names and Addresses: Intellectual Property Issues.
Final Report of the WIPO Internet Domain Process
. April 30, 1999, available at
100 See discussion in Section 4.4.
101 Another example is the "a" with diaeresis "¨" ("ä"), which in German should be sorted and looked at
exactly as an "a" with diacritical character, but in Swedish has nothing to do with the character "a" except
the look. In the same way Å as the abbreviation for the unit of length "angstrom" is one character, but the
initial character of the word "Ångström" is another (which in turn is different from an "A"). The other
problem with the German "a" with diaresis ("umlaut") is that it may be considered to match the string "ae",
whereas not all names containing "ae" match "ä".
102 Formally, the NIC was the Defense Data Network-Network Information Center (DDN-NIC).


Prepublication Copy--Subject to Further Editorial Correction
protocol and network numbers and controlled the definition of TLDs.103 Overall, the
administration and policy oversight for domain names was relatively straightforward.
In the mid-1980s, the National Science Foundation (NSF) created NSFNet to provide
data communication services to researchers and educators. It selected the Transmission Control
Protocol/Internet Protocol (TCP/IP) as its transport protocol and worked closely with the
Department of Energy, the National Aeronautics and Space Administration, and DARPA to share
facilities to extend this infrastructure in the United States and worldwide. NSF encouraged
campus network investment by focusing its efforts on high-speed and high-capacity long-haul
"backbone"104 and regional networks to connect the campuses. Thus, the responsibility for the
civilian network gradually shifted from the DOD to NSF. (See Box 2.3 for a timeline of the
shifting administration of domain names.) In the early 1990s, NSF made another important
decision--to withdraw as the primary financial benefactor for the backbone of the Internet and to
encourage a commercial market for support of transport facilities.
Continuing on this path, in 1993 NSF replaced DOD as the funding agency for domain
name management. As the workload increased, NSF contracted with Network Solutions, Inc.
(NSI) to manage the registration for most of the gTLDs (.com, .net, .org, .edu, and .gov),
through InterNIC. At this time, NSF, preserving the practice that the registration of domain
names would be free to registrants, subsidized the costs associated with domain name
registration. See Box 2.3.
Increasing scale was not the only impetus for administrative evolution. The increasing
economic and social value of domain names caused new players to become interested in the
realm of domain names. As discussed earlier, holders of highly visible and valuable trademarks
developed an active interest in domain names. Many other entities, from national governments
and public interest groups to the firms in the emerging domain name industry, also developed a
keen interest in all things related to domain names. Thus, the 1990s saw the domain name
community expand radically, both in scale and, especially important to understand, in the scope
of the interests and backgrounds of participants.105

103 Jon Postel had a central role in the DNS from the beginning as co-author of "The Domain Naming
Convention for Internet User Applications," RFC 819. Postel "held leadership positions in several Internet
infrastructure activities. He was founder and head of the Internet Assigned Numbers Authority, RFC
editor, and chief administrator of the .US domain. He was expected to play a crucial role in the future of
Internet administration, which [was] in the process of being transferred to the private sector [the Internet
Corporation for Assigned Names and Numbers (ICANN)]." Source: The Domain Name Handbook, In
Memoriam, Dr. Jonathan B. Postel, August 3, 1943 ­ October 16, 1998. See
<>, accessed March 31, 2004.
104 A backbone is a network that interconnects other networks. Backbone networks often operate over
relatively longer distances than do typical networks.
105 This diversity in the range of participants creates challenges in achieving consensus in the decisions
needed to make progress on various problems. Among other things, conflicting goals and varying
communication styles and vocabulary contribute to these challenges. Even agreeing on something as basic
as defining "DNS" can lead to disputes.


Prepublication Copy--Subject to Further Editorial Correction

BOX 2.3
Administration of the Domain Name System in the 1990s: The Road to ICANN

Responsibility for much of the Network Information Center (NIC) was transferred from SRI
(operating on the behalf of the Department of Defense; DOD) to Government Systems, Inc.,
which then subcontracted the entire operation to Network Solutions, Inc. (NSI). NSI started
operating the NIC in 1992.

The National Science Foundation (NSF) replaced DOD as the funding source for the NIC. NSF
completed a service contract with InterNIC, the umbrella organization for the participating
contractors AT&T (directory and database services), NSI (registration services), and General
Atomics/CERFnet (information services). Thus, NSF engaged NSI to take over domain name
registration services for most of the generic top-level domains (gTLDs) through a 5-year
cooperative agreement.

"Domain Name System Structure and Delegation" (RFC 1591), written by Jon Postel, was
published and gained general acceptance.106

NSF and NSI amended their cooperative agreement, imposing a $100 fee for 2 years of domain
name registration.

The International Ad Hoc Committee (IAHC), a coalition of individuals representing various
constituencies established in 1996, released a proposal for the administration and management of
gTLDs that included a framework for a governance structure, captured in a document known as
the Generic Top Level Domain Memorandum of Understanding (gTLD-MoU).107

The U.S. government created an interagency group to address the domain name issue and
assigned lead responsibility to the National Telecommunications and Information Administration
(NTIA), Department of Commerce. This interagency group reviewed the IAHC proposal and
solicited public comment.

As a part of the Clinton Administration's "Framework for Global Electronic Commerce,"108 the
Department of Commerce was directed to privatize the Domain Name System in a manner that
would increase competition and facilitate international participation in its management. The
department issued a call for public input relating to the overall framework of the DNS.

The NTIA released "A Proposal to Improve Technical Management of Internet Names and
Addresses," also known as the Green Paper. This proposal called for a private, non-profit

106 Available at <>.
107 For further information, see <>.
108 See <>.


Prepublication Copy--Subject to Further Editorial Correction
corporation, headquartered in the United States, to manage domain names and IP addresses, and
for "the addition of up to five new registries."109

A final statement of policy, the "Management of Internet Names and Addresses," also known as
the White Paper, was issued by NTIA. The White Paper reaffirmed the goals of the Green Paper,
while having the U.S. government take a more hands-off approach, and urged the creation of a
new not-for-profit corporation to oversee the management and assignment of domain names and
IP addresses. Goals for the new corporation included ensuring stability, competition, private and
bottom-up coordination, and fair representation of the Internet community.110

NSF transferred authority to the U.S. Department of Commerce to administer the cooperative
agreement under which domain name registration services are provided.111

Internet constituencies (e.g., those that attended the workshops held under the auspices of the
International Forum on the White Paper) discussed how the new entity (the New Corporation, or
"NewCo") might be constituted and structured. A group led by Jon Postel (and under his name)
proposed a set of bylaws and articles for the incorporation of NewCo. The final version of
NewCo's (then named as the Internet Corporation for Assigned Names and Numbers; ICANN)
bylaws and articles of incorporation were submitted to NTIA in October. On November 25,
NTIA and ICANN signed an official memorandum of understanding (MOU), with an initial
termination date of September 30, 2000.

In October, NTIA and NSI extended their cooperative agreement through September 2000. NSI
committed to a timetable for the development of a shared registration system (SRS) that permitted
multiple registrars to provide registration services within the .com, .net, and .org gTLDs.
Also, NSI agreed to separate its registrar and registry operations into separate divisions, to
recognize NewCo, and to make no changes to the root without written approval from the U.S.

By 1996, the belief by some (e.g., Jon Postel) that additional TLDs were needed led to
the establishment of the International Ad Hoc Committee (IAHC) to develop a framework for the
administration of domain names, which became known as the Generic Top Level Domain
Memorandum of Understanding (gTLD-MoU). The IAHC's proposal for an institutional
framework prompted a strong reaction from a few key constituencies and "sent ripples through
the international system," as characterized by Milton Mueller.112 Although the gTLD-MoU was
not implemented, its creation did motivate the discussions leading to the development of the
Green and White Papers (see Box 2.3) and the eventual creation of the Internet Corporation for
Assigned Names and Numbers (ICANN) in late 1998.

NSI exclusively operated the .com, .net, and .org TLDs through 1998. The registry
operations (associated with the management of the TLD databases themselves) and registrar
operations (associated with the retail functions of dealing with customers) were integrated.
NTIA's agreement with NSI in late 1998 required NSI to separate its registry and registrar
functions so that other registrars could enter the market. To facilitate the entry of other firms,
NSI also agreed to establish a shared registration system to enable all registrars (including NSI's

109 See <>.
110 For the text of the White Paper see <>.
111 See <>.
112 From Milton Mueller, 2000, "Internet Domain Names: Property Rights and Institutional Innovation,"
Entrepreneurship and Economic Growth in the American Economy, Vol. 12, 93-131, p. 111.


Prepublication Copy--Subject to Further Editorial Correction
registrar unit) to interact with the registry database. The vibrant market for domain name
registration services in the .com TLD that developed in the late 1990s also spurred interest in the
creation of new TLDs.
Thus, the DNS has experienced an extraordinary evolution since its birth in the early
1980s. Initially intended to address specific technical and operational problems of concern to a
small, relatively homogeneous group of computer scientists and engineers, the DNS came to
involve many other groups such as lawyers, businesspeople, national governments, public interest
groups, non-profit organizations, and others. The issues surrounding the DNS became
increasingly non-technical in nature and increasingly complex and controversial, and so the
founding of ICANN did not end the conflict among constituents, but rather provided the forum
through which the conflicts became further intensified. Chapters 3 and 5 further explore these
conflicts and the alternatives for their possible resolution.


Prepublication Copy--Subject to Further Editorial Correction
The Domain Name System: Current State

The Domain Name System (DNS) in 2005 serves a global Internet far larger and more diverse, in
users and in uses, than the relatively small homogeneous network for which it was first deployed in the
early 1980s. To meet the needs of this expanded and enhanced Internet, the DNS has developed into a
complex socio-technical-economic system comprising distributed name servers embedded in a
multilayered institutional framework. This chapter describes the DNS as it exists in 2005 to establish a
base for consideration of the future of the DNS and of navigation on the Internet.
The chapter begins with an explanation of how the DNS responds to queries, illustrating the
process with a query about the Internet Protocol (IP) address that corresponds to a particular domain
name. It then describes the basic architecture of the DNS: its domain name space, its hierarchical
structure, its basic programs, and its key standards and protocols. The core of the chapter is a description
of the implementation of this architecture at three levels of the DNS hierarchy: the root, the top-level
domains, and the second- and third-level domains. The distinctive characteristics of each level are
examined first, followed by descriptions of the technical system and its institutional framework. The
committee's conclusions about the current performance of the DNS architecture and the implementation
of each level are presented at the end of each section. Open issues affecting the future of the DNS are
collected and analyzed in Chapters 4 and 5.
Many of the contentious issues that arise in the context of the DNS concern domain names
themselves--in particular, the definition of permissible names and the rights to their use. Some of those
issues are introduced and discussed in Chapter 2.


Many things happen when a query to the DNS is initiated. If the DNS were a centralized
database, such as HOSTS.TXT,2 every query would go directly to a central file where the answer would
be found (or its absence noted). However, because the DNS is a hierarchical, distributed database, a
search in response to a query generally requires several steps. If necessary, it can begin at the root and
traverse a course through the tree of files to the one in which the sought-for answer resides. However,
frequently the search can begin further down the tree because previous answers are stored and reused by
the querying client. The design of the DNS ensures that the path down the tree will be followed without
detours or false starts, leading directly to the desired file because the structure of the domain name spells
out the route. This process may best be understood through an example, shown in Figure 3.1, which
illustrates the use of the DNS to find the IP address corresponding to the hypothetical domain name

1 This section elaborates on the high-level explanation in Chapter 2. It draws extensively on: Paul Albitz and
Cricket Liu. 2001. DNS and BIND. Sebastopol, Calif.: O'Reilly & Associates.
2 HOSTS.TXT is the predecessor of the DNS and is described in Section 2.1.
3 The process shown in Figure 3.1 assumes that the querying client has stored no relevant previous answers.

Prepublication Copy--Subject to Further Editorial Correction

Name Server


Name Server
com net

Name Server
IP Address of
Name Server
IP address
DNS System

FIGURE 3.1 Operation of the Domain Name System without a local name server.

Prepublication Copy--Subject to Further Editorial Correction
This is what would happen if, for example, the user wanted to access a Web site at that name, in which
case the requesting application would be a browser. However, the same process would be followed for,
say, an e-mail application or any other application supported by a host on the Internet.
Two versions of the process are described below: first, the version shown in Figure 3.1, which
would be followed if this were a new query from a computer that was not on the same DNS subtree as the server; and then a version shortened by taking advantage of additional information
from shared servers or previous queries.

3.1.1 A New, Remote Query

When a domain name is used in a Web browser, e-mail program, or otherwise, the applications
program forms a request--a query. The example query, "What is the IP address corresponding to the
domain name," goes first to a piece of software called a resolver. Resolver
software is ordinarily incorporated as part of other software resident on the user's computer4 or in a host
to which it is linked. There are two kinds of resolvers: stub resolvers and iterative resolvers. Both types
of resolver send queries to name servers (see below), but they differ in how the resolver selects the name
servers to which it sends the query and how much of the work of answering the query is performed by the
resolver. A name server is a computer running one of a small number of name-serving programs, the
most common of which is the Berkeley Internet Name Domain (BIND) software.5 A stub resolver simply
forwards the query to a local name server and awaits a reply. It places on the name server the burden of
searching the DNS for the answer. An iterative resolver, in contrast, retains control of the search by using
the answer from each successive name server to guide its search. This example assumes that the query
comes from a stub resolver.
Name servers are located throughout the Internet: at the root and the top-level domain registries,
in organizations' intranet infrastructures, and at Internet service providers (ISPs). Name servers can
perform two important functions:

First, they are designed to reply directly to queries concerning the portion of the domain
name space for which they have complete information, which is called their zone and for which they are
said to be authoritative (see Section 3.1.2).
Second, they can, by incorporating an iterative resolver, reply to queries concerning zones for
which they are not authoritative, obtaining information from other name servers in the DNS (described in
this section). The incorporated iterative resolver will in almost all cases also contain a file or cache of
answers obtained as a result of processing previous queries (see Section 3.1.3). In this case, the
combination of name server and iterative resolver is said to be a caching or recursive name server.

In practice, name servers with a heavy query demand or at the top levels of the DNS hierarchy are
often configured to be authoritative only and not to offer caching/recursive services, except through a
separate server. The name servers at ISPs, however, will offer caching/recursive services to their
customers' stub resolvers, but may not be authoritative for any domain.

4 For example, this is part of the Transmission Control Protocol/Internet Protocol (TCP/IP) stack in Microsoft
Windows software, but some applications incorporate their own resolvers. Consequently, a computer may contain
more than one resolver.
5 This name server program was originally written in 1983-1984 by a group of graduate students at the University of
California, Berkeley, with funding from the Defense Advanced Research Projects Agency (DARPA). Name servers
are discussed further in Section 3.2.3.

Prepublication Copy--Subject to Further Editorial Correction
In Figure 3.1, the query is shown as going first to a name server offering recursive services
located at the user's ISP.6 Since in this example the iterative resolver incorporated into the ISP's name
server has not recently seen this query or any portion of it and the name server is not authoritative for any
portion of the query, it sends the query on to the DNS root. It is able to find the root because the IP
addresses of the name servers for the root, called the root hints data, were manually entered into a file on
the computer, the hints file. Some DNS systems automatically detect changes to the list of root name
servers and make use of them, but the software never changes the file because to do so might eliminate
the fallback in case of an attack that maliciously delivered an incorrect list.
There are 13 root name servers (and many satellite copies of them)7 distributed around the world,
and the querying name server will go to one chosen by an algorithm that, although differing among name
server implementations, usually takes the shortest response time into account. The multiple computer
copies of some of the 13 name servers employ a technology called "anycast" addressing (see Box 3.1).
Although geographically distributed, each satellite is capable of responding to queries to the same IP
If, for some reason, one root name server (or its closest satellite) does not respond, the iterative
resolver will continue to try other servers according to its selection algorithm, and so on, until it receives a
response. Similar behavior is common to all iterative resolvers at whatever level in the DNS hierarchy
they are searching.
The response of a root name server, which is configured to be authoritative only, takes the form:
"The address of is not in my zone's name file, but here are the names and/or addresses
of name servers that are authoritative for .edu." The ISP's iterative resolver then sends the same query to
one of the .edu name servers, which responds: "The address of is not in my zone's
name file, but here are the names and/or addresses of the name servers that are authoritative for"
The ISP's resolver then queries one of the name servers, which refers it to a name
server, which is authoritative for the requested domain name and replies with the corresponding IP

3.1.2 Local Query

A name server can answer many queries quickly when these queries request the address of a
domain name for which the name server is authoritative. This is often the case, for example, for name
servers on organizational intranets, where most of the requests are for IP addresses of other computers on
the intranet. In such a case, the name server can respond to the query without going to the larger DNS,
simply by looking up the answer in its local database. For example, if the local name server in the
previous example was authoritative for, it could provide the response directly.

3.1.3 Repeat Query

A caching name server can answer many other queries quickly when it has responded previously
to queries that were identical or matched at a higher level of the tree. (For example, in 2005 virtually
every caching name server is highly likely to have cached the IP address for It
maintains those answers in a cache of information containing the addresses of name servers (and other
data) it has previously obtained. Before going to the root, it searches its cache to find the known name
servers closest (in the DNS hierarchy) to the domain being sought.

6 Where a query goes first is a consequence of an explicit configuration choice made by the user, an ISP, an
enterprise IT department, or by a dynamic configuration protocol whose values are supplied by one of those sources.
7 See Table 3.1 for a listing of the 13 root name servers and their base and satellite locations. See Box 3.1 for a
description of satellite servers.

Prepublication Copy--Subject to Further Editorial Correction
BOX 3.1
Anycast Addressing

Anycast addresses, a special type of Internet Protocol (IP) address, were invented in the early 1990s
to simplify the process of finding replicated services (i.e., services that are provided by multiple and
identical servers).8 Some of the operators of root name servers have implemented anycast addressing as a
way to facilitate load sharing, to improve service, and to reduce vulnerability to attacks.
The use of anycast addresses allows a root name server operator to install copies of the root zone file
at different servers (in this report, those servers that replicate the root zone file are called satellites).
Properly configured and located, each of the satellites will get a share of the traffic for the root name
server. Although the shares will, in most cases, not be equal, the load of queries will be distributed and
thus relieve the load burden on the root name server. Satellites that are located at the same physical site
are using local anycast addressing, also known as load balancing, which is widely deployed among the
root name server operators.
From the user's perspective, the great advantage derived from the adoption of anycast addressing is
improved service. The satellites are typically placed at topologically diverse locations in the Internet.
Queries can therefore be answered more swiftly. An additional benefit is that the DNS queries use, in the
aggregate, fewer network resources, because servers will tend to be "closer" on the network to the sources
of the queries.
The use of anycast addressing can sharply reduce the impact of an attack on a root name server: In
the short run, physically disabling a root name server does not affect the operation of its satellites, and
physically disabling a satellite disables only that satellite. In the long run, there is the question of how
satellites would obtain updated root zone information. It is also much harder to mount an effective
electronic attack--because queries are routed to the closest satellite (or the root name server itself, if it is
the closest). An attacker would need to place (or acquire) machines close to each of the satellites and the
root name server if the attacker wished to disable all access to the service.9 A single attacking machine
might disable the closest server--whether a satellite or the root name server itself. The other servers
would be affected only in a minimal way, through a slightly increased load if one server were rendered

Therefore, the adoption of anycast addressing by the root name server operators is a positive
development. However, more general use of anycast addressing is problematic because current methods
for deploying these addresses waste a number of IP addresses.10 Given the importance of a robust DNS,
this wastage is acceptable for the operation of the root name servers, but not necessarily for other domain
name servers.
Monitoring the performance of satellites presents root name server operators with a difficult problem.
Such monitoring involves the placement of monitoring devices within the part of the Internet that each
satellite serves and can represent a significant logistical challenge because the satellites may be widely

8 The initial motivation for the creation of anycast addressing was to reduce the need manually to configure
information about basic services such as DNS resolvers. The basic idea is that a "host transmits a datagram [a data
packet carrying its own address information so it can be independently routed from its source to the destination
computer] to an anycast address and the internetwork [Internet] is responsible for providing best effort delivery of
the datagram to at least one, and preferably only one, of the servers that accept datagrams for the anycast address."
See Craig Partridge, Trevor Mendez, and Walter Milliken, "Host Anycasting Service," Request for Comments
(RFC) 1546, November 1993. See Box 3.2 for an explanation of RFCs. All RFCs are available at <http://www.rfc->.
9 And even then other root name servers and their satellites would be accessible.
10 As of April 2004, anycast addresses were not fully supported in the current version of IP (version 4). In
particular, the Border Gateway Protocol (BGP, which is the cross-provider routing protocol) was not designed to
accommodate anycast addresses and, therefore, a workaround is used that wastes about 256 IP addresses for each
root name server that employs anycast addressing.

Prepublication Copy--Subject to Further Editorial Correction

For example, if the ISP's caching name server in the previous example were to receive a query
for the address of, it would check to see if it already knew the address of the
name server authoritative for If it did, its iterative resolver would send the query
directly to that server, shortening the path that must be taken to obtain an authoritative response. If it did
not, it would then check to see if it had the address of the name server authoritative for, and
finally .edu. Only if it had none of those addresses would it go to the root.
This property of caching--that it limits the number of queries that are sent to the root name
server--has been crucial to the manageability of the growth in the query load on the root system. If all
DNS queries were to start at a root name server, the capacity of the root system (as a whole) would have
to be of an entirely different magnitude, posing more formidable technical and economic challenges as a
The Internet and the many services on it are subject to constant, sometimes rapid change. As a
result, cached information can become outdated. To reduce that problem, the administrator of each zone
assigns a time to live (TTL) to each datum that it sends out in reply to a query. After the corresponding
amount of time has passed, the name server is expected to eliminate the datum from its cache.
Often a name server will receive information that a domain name being sought does not exist.
That can happen because the query is ill formed, contains a typographical error, is based on a user's
incorrect guess about a desired domain name, refers to a name that does not exist or no longer exists, or
refers to a domain name on a private network that is not on the public DNS.11 Since such inquiries do not
correspond to a cached address, even the caching name server system will not normally relieve the load
on the root name servers related to such requests. To minimize the load on the network and improve
response time, however, it is desirable that name servers store information about such non-existent
domains. That practice is referred to as negative caching (as introduced in Section 2.3.1). The need to
assign a TTL also applies to negative caching, since a previously non-existent domain may come into
existence and would be missed if the negative cache did not eliminate responses regularly.


The architecture of the DNS--its conceptual design--comprises its name space, its hierarchical
structure, and the software that specifies operations within that name space and structure.12 The software
comprises two components: programs, which implement the resolver and name server functions on
various computers; and technical standards, which define the formats of the communication between the
programs, as well as the logical structure13 of the files in a name server.

3.2.1 Name Space

The name space for domain names is the set of all symbol strings that adhere to the rules for
forming domain names specified in the design of the DNS. Those rules define a standard format that
imposes a tree structure on the name space. Each node on the tree has a label, which consists of 1 to 63
characters drawn from a restricted subset of ASCII14 comprising the 26 letters of the Roman alphabet, the

11 A significant portion of queries to the root name servers are ill formed or in error according to studies by
researchers. See for example: Duane Wessels and Marina Fomenkov. 2003. "Wow, That's a Lot of Packets."
Proceedings of Passive and Active Measurement Workshop. Available at
12 The evolution of the DNS is described in Chapter 2.
13 They will be stored in whatever data structures the local database software requires.
14 American National Standards Institute. "USA Code for Information Interchange." X3.4. 1968.

Prepublication Copy--Subject to Further Editorial Correction
10 numerals from 0 to 9, and the hyphen. Labels may not begin or end with a hyphen. The fully qualified
domain name of a node is the list of the labels on the path from the node to the root of the tree written
with the deepest node on the left and with those to the right getting successively closer to the root. In
external presentation form, the labels are separated by dots (.). By convention, the root label has null
length and is written as a dot (.), but its presence is optional. Thus, ""15 and
"" are equivalent domain names. The total length of a domain name must be less than
256 characters. Each node on the tree (including the leaves) corresponds to a collection of data, which
may be empty at the internal nodes, unless they constitute a delegation point. The data are represented by
resource records, which are described in Box 3.2 in Section 3.2.4.

3.2.2 Hierarchical Structure

The hierarchical, tree structure of the DNS facilitates both an efficient response to queries and the
effective decentralization of responsibility for its maintenance and operation. That is accomplished
through the division of a domain into subdomains, which taken together need not directly include all of
the hosts in the domain. The responsibilities for maintaining the name files for some or all of the
subdomains can then be delegated to different organizations. They, in turn, can further divide and
delegate, a process that can be repeated as often as necessary. The parent domain need only retain
pointers to the subdomains so that it can refer queries appropriately.
The hierarchical delegation of responsibility is one of the great strengths of the DNS architecture.
It is up to the organization responsible for a zone to maintain the corresponding zone file (thus, the
organization has considerable motivation to provide satisfactory maintenance)--the data file in the zone's
name servers that contains pointers to hosts in the zone and to the name servers for delegated zones (see
"DNS Zone Data File" in Section 3.2.4). The work of keeping the DNS current is, thereby, distributed to
organizations throughout the entire DNS tree, down to the lowest leaves. Instead of a central organization
being responsible for keeping the DNS data current and accurate, which would have been an impossible
task, there are millions of organizations and individuals across the globe doing the work.
Figure 3.2 illustrates the delegation of responsibility from the root to the .edu domain, whose
name servers are said to be authoritative for the .edu zone. The .edu domain in turn delegates
responsibility, in this limited illustration, for the subdomains to three universities--the University of
California at Los Angeles (UCLA), Southern Methodist University (SMU), and the Massachusetts
Institute of Technology (MIT), whose name servers are authoritative for their corresponding zones--,, and In Figure 3.2 the MIT subdomain is further delegated to two
departments for illustrative purposes--Sloan School and Electrical Engineering and Computer Science,
whose name servers are authoritative for their and Note
that in the example, the zone includes a pointer directly to a
addition to pointers to the delegated zones.
This distribution of responsibility, combined with the distributed handling of tasks by the
technology, is why the DNS scales, or handles growth so well.

3.2.3 Programs: BIND and Others

The DNS requires only two types of programs to operate within the context of the existing
First, there must be resolver software. The resolver, recall from Section 3.1, is a client that
accepts a query, passes the query to a name server, interprets the response, and returns the response to the

15 Note the trailing dot after "edu."

Prepublication Copy--Subject to Further Editorial Correction
source of the query. Generally, resolvers are just a set of library routines within a name server, a browser,
or an operating system.
Second, there must be name server software that performs the functions described in Section 3.1.
As noted above, the most common name server software is BIND,16 which is used on the majority of
name servers on the Internet.17 There have been many versions of BIND produced since it was originally
written in the early 1980s. The most recent releases (as of February 2005) are BIND 8.4.6, which extends
and enhances prior versions, and BIND 9.3.0 and 9.2.4, which is the latest release of a major rewrite of
the underlying architecture undertaken in response to anticipated demands resulting from the rapid growth
of the Internet. Although originally written for Unix operating systems, BIND has been programmed for
other operating systems, including Windows NT and Windows 2000.18 It has also been used as the basis
for many vendor name servers.19
The rest of this section uses BIND as an example because it is widely deployed. However, other
name server software may behave differently in some respects and still conform to the domain name
server standards.
A name server running BIND, or a comparable program, has complete information and authority
over the zones for which it is authoritative. It contains all domain names from the top of its zone down to
its leaves, except for those that are delegated to zones within it. See Figure 3.2.
Each zone must have a primary master (or primary) server that receives the contents of the zone
data file from some manually prepared local file. The primary server is the ultimate source of information
about the corresponding domain. There generally is at least 1 and possibly as many as 12 secondary
servers, which obtain copies of the zone data file from the primary or another secondary.
BIND has not traditionally made severe computational or storage demands on the computers that
run it. It has often been implemented on old machines that have been taken out of front-line service.
However, BIND 9 incorporates security and other features that impose much more severe computational
loads. In general, the computational requirements on a name server depend on the number of queries per
second and the relative distribution of the types of queries (because some types of queries impose greater
computational demands than others), as well as the extent of change of the zone files. The memory
requirements are determined by the size of its cache and zone files. In the latest versions of BIND, the
cache size can be controlled. In general, the cache size for a new name server is determined by
observation of the name server's operation over a few weeks to determine how much memory is required
to respond to the query demand at its installation.

16 BIND also contains a resolver library.
17 According to a survey by Internet Systems Consortium, Inc. in January 2004 it was used by almost 75,000
servers; the second most popular name server software was provided by Microsoft for almost 15,000 servers. For
more information about the survey and BIND, see "ISC Internet Domain Survey," March 10, 2004, available at
18 See "DNS Server Software" at <> for a survey of DNS servers available on
various operating systems.
19 A current list is posted at <>.

Prepublication Copy--Subject to Further Editorial Correction
= delegation

edu zone
smu zone zone

mit zone
eecs zone zone
.edu domain

FIGURE 3.2 The .edu domain divided into zones. SOURCE: Based on Figures 2-8 and 2-9 of Paul
Albitz and Cricket Liu, DNS and BIND, Fourth Edition, O'Reilly, 2001.


Prepublication Copy--Subject to Further Editorial Correction
3.2.4 Standards

The queries and responses that flow between name servers must be in a protocol that is readily
interpretable by any name server, no matter which software and hardware it uses. To that end, the data in
the queries and responses are, like the zone data, represented as resource records (Box 3.2).

BOX 3.2
Resource Records

Resource records are used in every DNS zone data file and every DNS message.20 Resource
records begin with a domain name (NAME), which is followed by the type (TYPE) and class
(CLASS) fields. Those fields are followed by the time to live (TTL) and a data field (RDATA),
appropriate to the type and class. Domain names and IP addresses make up a large portion of the data
in a typical zone data file.

NAME: The domain to which the record refers.
TYPE: Describes the type of data to be found in the resource record. The list of possible types is
open-ended; each is associated with a type code. There were more than 50 in January 2003, of which
no more than 20 are used to any extent.

Examples: A = host IP address; NS = an authoritative name server; MX = mail exchange.21
CLASS: Only one class, IN for Internet, is widely used. Four classes have been defined to date,
one of which is obsolete.
TTL: Specifies the time interval (in seconds) that the resource record may be cached before the
source of the information should again be consulted.

Example: 86400 (equivalent to one day).
RDATA: Describes the resource in question.

Example: If the class = IN and the type = A, this field is an IP address.22

If the class = IN and the type = NS, this field is a domain name.

DNS Zone Data File

Each DNS zone has a zone data file (also called the master file) that contains both resource
records describing the zone and actual data records for the zone. The former group specifies:

The authority for a zone, including the IP address of the primary name server and the e-mail
address of the responsible person, as well as times associated with updating the secondary name servers;
All the name servers that are authoritative for the zone; and
The IP addresses of the name servers (name-to-address mappings) that are in the zone.

There can also be data files that contain reverse mappings (i.e., tables for conversion from IP
addresses back to computer names). The domain names in these files look like IP addresses turned back to
front, and they all end in

20 For further detail, see RFC 1035.
21 For an explanation of "mail exchange," see Section 2.3.3.
22 Since each Type-A resource record in Class IN can include only one IP address, a domain name that maps to
multiple IP addresses will have multiple resource records--one for each IP address.

Prepublication Copy--Subject to Further Editorial Correction
The zone data file may contain some additional types of resource records. The specification of
resource records allows additional types of data to be added in the future, as required.

DNS Message Format

The DNS message format comprises five sections, some of which may be empty:

Question: the question for the name server (includes domain name),
Answer: resource records answering the question,
Authority: resource records pointing toward an authority, and
Additional: resource records holding additional information.

The lengths of the four sections that follow it are specified in the header section.
The practical limitation on the number of root name servers to 13 is a consequence of the DNS
message format and the design decision to use datagrams employing a minimal protocol--the User
Datagram Protocol (UDP)24--to send DNS queries and responses so as to achieve high performance.
Because in current practice there must be room within the datagram answer section for a list of all root
servers--in order to update the list of root servers25 in every iterative resolver--the number of root
servers is constrained by the maximum length of that section.26 To maximize the number of root servers,
their names were shortened and standardized for more efficient compression, increasing to 13 the
maximum number that could fit in the space available.

3.2.5 Functions and Institutions

There are two critical functions that the DNS institutional framework performs in support of the
standards and the programs that define the DNS. The first is maintenance of the DNS standards, and the
second is ensuring the availability of DNS name server software. DNS client software, primarily stub
resolvers, is widely available in standard software--operating systems, Web browsers--and the specifics
of their provision are not further considered here.

Maintenance of the DNS Standards--The Internet Engineering Task Force

The definition and maintenance of the basic standards for the DNS are the responsibility of one
informal organization--the Internet Engineering Task Force (IETF) (see Box 3.3). The IETF has
attracted experienced and knowledgeable technical talent, who volunteer considerable time to its
activities. The requests for comments (RFCs) process has provided a means for this diffuse technical
community to build a freely available, peer-reviewed store of knowledge and the successful standards and
protocols that enable the Internet and the DNS to function reliably and to adapt smoothly to the additional

23 Addresses are converted to names in the .arpa domain for DNS lookup. The zone contains the
hosts associated with all registered IP addresses.
24 For more information on the User Datagram Protocol, see John Postel, "User Datagram Protocol," RFC 768,
August 28, 1980.
25 As noted above, every iterative resolver needs to know where the name servers for the root zone are in order to
have a starting point for a non-local, non-repeat query.
26 In theory, the response could be a list of name servers that contain the names and locations of root name servers.
Also, the hints file, which is local, could contain information about more than 13 name servers. However, the
limitation to 13 has not yet been judged to be a large enough issue to change current practice in either respect.

Prepublication Copy--Subject to Further Editorial Correction
requirements imposed by increased scale, new applications, and new classes of users. Although the
IETF's standards and protocols underpin the worldwide Internet and the DNS, it does not have the
authority, the political or economic power, or the interest to force their adoption or validate their
implementation. Rather, their universal use has been the result of the practical benefit of having freely
available high-quality standards that all can share, and that give no organization proprietary advantage.
As a volunteer collaborating body, the IETF periodically restructures its processes. In 2003, the IETF
identified a number of problems, both routine and structural, in its operations and initiated a process of
problem resolution.27 As is typical, it did so publicly via the RFC process.

BOX 3.3
The Internet Engineering Task Force and Requests for Comments

The Internet Engineering Task Force (IETF)28 is a voluntary, non-commercial
organization comprising individuals concerned with the evolution of the architecture and operation of
the Internet. It is open to anyone who wishes to participate and draws from a large international
community. However, almost all its participants are technologists from universities, network
infrastructure operators, and firms in related industries. Although the IETF holds three meetings each
year at locations around the world, much of its work is conducted through the circulation of e-mail to
electronic mailing lists. The IETF divides its activities among working groups, organized into areas
that are managed by area directors (ADs). The ADs, in turn, are members of the Internet Engineering
Steering Group (IESG). The Internet Architecture Board (IAB) provides general oversight on Internet
architecture issues and adjudicates appeals that are unresolved by the IESG, but it is not actively
involved in standards development or implementation. The Internet Society (ISOC) charters the IAB
and IESG.29
Requests for comments (RFCs) were established in 1969 to document technical and
organizational aspects of the Internet (originally the ARPANET). RFC memos discuss protocols,
procedures, programs, concepts, and various other aspects of the Internet. The IETF defines the
official specification documents of the Internet Protocol suite that are published as "standards-track"
RFCs. RFCs must first be published as Internet-Drafts--a mechanism to provide authors with the
ability to distribute and solicit comments on documents that may ultimately become RFCs. An
Internet-Draft, which can be published by anyone, has a maximum life of 6 months, unless updated
and assigned a new version number. The Internet-Drafts that are intended for progression onto the
standards track, and some other documents at IESG discretion, are "last called," which involves an
announcement being sent out to the Internet community that the IESG wants input on the document.
The Last Call is usually of a few weeks duration. Using input from it, the IESG makes a decision on
further processing of the Internet-Draft. Such decisions might include rejection of the draft,
publishing it as a standards-track document, or handling it in some other way. Documents that are
considered valuable and permanent, including all standards-track documents, are then submitted to
the RFC editor for publication as RFCs.30

27 The problem statement and the history of the response are described in E. Davies, ed., 2004, "IETF Problem
Statement," RFC 3774, May, available at <>; and E. Davies and J.
Hoffman, eds., 2004, "IETF Problem Resolution Process, " RFC 3844, August, available at
28 For a full description, see Susan Harris, "The Tao of IETF--A Novice's Guide to the Internet Engineering Task
Force," RFC 2160, August 2001. Available at <>.
29 For more information on the evolving relationship between ISOC and IESG, see
30 For details on this process, see Scott O. Bradner, 1996, "The Internet Standards Process, Revision 3,"
RFC 2026, Internet Engineering Task Force, October, available at <>.
For background on RFCs and a searchable repository of RFCs, see <>.

Prepublication Copy--Subject to Further Editorial Correction
Providing Root Name Server Software--Internet Software Consortium, Inc. and Other Software

Internet Systems Consortium, Inc. (ISC), a not-for-profit organization formerly called the Internet
Software Consortium, has assumed responsibility for continuing maintenance and development of
BIND.31 Although ISC's scope is considerably more focused than the IETF's, it, too, has played a key
role in the smooth development and operation of the Internet and the DNS. By continuing to evolve
BIND and making it readily available worldwide, free of charge, ISC has provided a widely adopted
implementation of the DNS standards promulgated by the IETF.
Implementations of BIND and other DNS server software are also available from Microsoft and
other software companies as well as from various providers in the form of freeware and shareware.

3.2.6 Assessment

Conclusion: The architecture of the Domain Name System has demonstrated the ability to scale
to support the Internet's rapid growth, never holding up its development because of an inability to meet
the challenges of vast increases in the number of domain names, users, and queries. Furthermore, it has
demonstrated robustness, never failing so extensively as to prevent the Internet from functioning for even
a few minutes. What failures there have been have occurred in subzones of the system and have been
insulated from the rest of the DNS by its hierarchical, distributed structure.

The hierarchical architecture and caching have been critical to the ability of the DNS to scale
smoothly. First, they ensure that many queries are resolved locally, never reaching the higher levels of the
DNS. Second, even queries that do reach higher levels of the DNS do so only the first time they are made
by a specific name server in a given period (the TTL), with all subsequent queries from the same name
server during that period being answered locally.
These benefits are amplified by the ability of large ISPs, such as MCI and Verio, to maintain very
large caches and, thereby, to handle a substantial portion of the queries originating from their customers
without ever passing them along to the DNS.

Conclusion: Through its RFC process, the IETF has created a store of knowledge and a body of
standards and protocols that have, thus far, enabled the Internet and the DNS to function reliably and to
adapt smoothly to the additional requirements imposed by increased scale, new applications, and new
classes of users. Though ISC's scope is much more focused than the IETF's and its participation and
decision-making processes are far less open and public, by continuing the evolution of BIND and making
it readily available ISC has brought most of the IETF DNS standards32 into practical effect.

However, the continued growth of the Internet is posing new challenges and placing new
demands on the DNS architecture. The issues of future security and robustness, internationalized domain
names, and the intersection of the DNS with the telephone system are addressed in Chapter 4.

31 Information about ISC is available at <>.
32 This issue is complex since not all of IETF's standards are mandatory or even recommended, and some are
experimental or turn out to be bad ideas. The situation also differs between BIND 8, which is on end-of-life
maintenance, and BIND 9, which is an entirely new code base. BIND 9 is believed to adhere to all the mandatory
specifications, while ISC has eliminated non-compliant code from BIND 8 whenever it has been identified.

Prepublication Copy--Subject to Further Editorial Correction

The top of the DNS inverted tree is its root, or more properly, the root zone (see Section 3.1.1).
The root zone name file (or, simply, root zone file) is stored in 13 root name servers, which use it to
respond to queries to the root. As shown in Section 3.1, while queries can enter lower on the tree if their
resolvers have current cached information, or if the query lies within the zone of the local name server,
the root serves as the assured point of entry to the DNS for any other query.

3.3.1 Characteristics of the Root Zone

Defining Characteristics

The root zone file defines the DNS. For all practical purposes, a top-level domain (and, therefore,
all of its lower-level domains) is in the DNS if and only if it is listed in the root zone file. Therefore,
presence in the root determines which DNS domains are available on the Internet.33 As Internet use has
grown, especially with the explosive growth of the Web and its reliance on domain names as key parts of
Web site addresses, entry of a top-level domain into the root zone file has become a subject of substantial
economic and social importance. Consequently, who controls entry into the root, and by what means,
have become controversial subjects. The current process for resolving these issues is described in Section

Critical Characteristics

The root zone and the root name servers are critical to the operation of the DNS. The effective
and reliable operation of the DNS, and of the Internet, is entirely dependent on the accuracy and integrity
of the root zone file (and its copies) and the security, reliability, and efficiency of the root name servers.
Fortunately, the root zone has proven highly resilient and resistant to errors or disruptions.
One possible source of error is an incorrect entry--a mistyped domain name or an erroneous IP
address--in the root zone file. A consequence could be misdirection of queries seeking a particular top-
level domain (TLD) name server, say .net. That could prevent users searching for the address of any
domain name within that name server's TLD, say any address in .net, from reaching it through the
specific root name server containing the incorrect entry. If the error were updated in all copies of the root
zone file, access would effectively be denied to all domain names in that TLD, except from name servers
that had previously cached the correct address.34 However, relatively simple error detection and correction
procedures can quickly prevent such errors. (For example, most large top-level domains are programmed
to regularly query the root to ensure that it is properly routing queries seeking that TLD.)
A possibly disruptive event would involve one or more of the root name servers being non-
operational for hours or more. This might happen because of the unexpected failure of a computer or
communication link (accidental or purposeful) or because of regular maintenance. However, since the
capacity of each name server is many times greater than its average load, and iterative resolvers can use
any of the root name servers, other name servers can take up the load without degradation in the system's
performance that is perceptible to users. Moreover, in recent years such outages have been very rare.35

33 The IP addresses for the servers of well-known TLDs are widely available, and so to them presence in the root
may be less important. For less well known and new TLDs, however, presence in the root is the critical question.
34 This exception would hold true only until the TTLs of the cached addresses expired.
35 For example, in 2000, four of the 13 root servers failed for a brief period because of a technical mistake. In 1997,
a more serious problem involving the transfer of an incorrect directory list to seven root servers caused much of the
traffic on the Internet to come to a stop. As reported in "ISC Sets Up Crisis Centre to Protect Domain Name

Prepublication Copy--Subject to Further Editorial Correction
Although, as noted, there have been instances in the past of errors in the root zone file that have
had significant effects on the reliable operation of the DNS, there have been none in recent times. At
least for the current rates of queries and of changes in the root zone file, the systems of error checking and
correction and of capacity sharing appear to be working successfully.

Unique Characteristics

The design of the DNS is predicated on the existence of a single authoritative root that ensures
that there is one and only one set of top-level domains. However, some of those who object to the pace at
which new generic TLDs have been added to the root, or to the process by which they have been selected,
have sought a solution through the addition of one or more roots. They have been joined by some who
believe in the principle that having competition in the delivery of root services would benefit Internet
On the other side are those who argue that additional roots would compromise the reliable
operation of the Internet by, among other things, opening the possibility of multiple addresses, associated
with different entities, being assigned the same domain name.36 Most experienced technologists view the
idea of introducing multiple roots for the DNS as threatening the stable operation of the DNS.
Moreover, since which domains are accessible to a user would depend on which of the multiple
roots were used, these technologists see a multiple-root DNS as balkanizing the Internet. Although there
may be some benefits from having competing roots, the presence of network externalities37 encourages
ISPs and name servers to converge on a single root that provides global compatibility. Consequently,
many technologists and economists believe it is unlikely that an alternative root would achieve
widespread success.38 In their view, while competition may serve a valuable purpose in the short term,
the task of maintaining the root zone file will equilibrate on a single, dominant root zone file, albeit an
equilibrium in which operational control is shared among a number of (non-competing) entities.39

There have been several attempts to create alternate roots that have data about the TLDs that are
recognized by the current root servers plus some additional TLDs that the operator of the alternate root is
trying to promote. However, these attempts have generally been unsuccessful, in large measure because
of the lack of global compatibility among the domain names recognized only by alternative roots, which
has made users unwilling to use them. However, one to have reached
financial profitability, to have registered over 100,000 domain names, and to be potentially accessible to
175 million Internet users through allied ISPs and browser plug-in software. It offers 12 additional TLDs
in English as well as in each of Spanish, French, Portuguese, Italian, and German.40 Although the
continued existence of can be seen as a challenge to the DNS and ICANN's management of the
root zone file, it has not thus far appeared to have had any significant effect on either. Furthermore, in

System," Sydney Morning Herald, October 21, 2003, available at
36 See, for example: M. Stuart Lynn. 2001. "ICP-3, a Unique, Authoritative Root for the DNS." ICANN, July.
Available at <>. Also, Internet Architecture Board. 2000. "IAB Technical
Comment on the Unique DNS Root." RFC 2826. May.
37 Network externalities are the increased benefits received by one user of a system as the number of users of the
system increases.
38 Milton L. Mueller. 2002. "Competing DNS Roots: Creative Destruction or Just Plain Destruction?" Journal of
Network Industries.
Vol. 3. No. 3.
39 If an alternate root attracts a sufficient number of registrations, it raises the possibility that TLDs in the alternate
root and registrations within these TLDs could be created for speculative purposes. If ICANN creates a TLD that
already exists in an alternate root, the organization that controls the corresponding TLD in the alternate root could be
willing to transfer the registered names to the ICANN TLD--for a price.
40 Information obtained from <> in February 2004.

Prepublication Copy--Subject to Further Editorial Correction
September 2004, was acquired by the Vendare Group, an online media and marketing company,
as the basis for a division offering search services, while continuing to offer domain name registration.41

3.3.2 Technical System of the Root Zone

The Root Zone File

The root zone file contains resource records for all the TLDs as well as for the root. In January
2004, there were 258 sets of entries. Of those, 243 were country code TLDs (ccTLDs) and 15 were
generic TLDs (gTLDs). (One of the gTLDs was the domain .arpa that is used for infrastructural
purposes, which is sometimes considered a separate category.) Because there are multiple records in the
root zone file for each of the subdomains, there were a total of 2143 records in the root zone, which
translates to about 78 kilobytes of storage. Moreover, although, as noted earlier, all the root name servers
together execute, in total, several billion searches a day on this file,42 this is well within their
computational capability because of the substantial overprovisioning of the system.

The Root Name Servers

Like other zone files, the root zone initially had a primary or master server accessible from the
DNS and several--in this case, 12--secondary or slave servers. That primary zone file was the most
current of the files, and all updates and changes were made to it; it served as the reference source for the
root zone. The secondary files were updated from it on a regular basis, at least twice daily. Starting in
1996 and achieving adoption by all secondaries by the end of 2001, the role of the primary was
transferred to a "hidden primary," which is a server that is used to update the secondaries but is not itself
accessible from the DNS. VeriSign operates this hidden primary. All 13 of the public root name servers
are now secondaries, including the former primary. Digital signatures, error checking, and correction
processes are in place to minimize the chance of introducing errors or being successfully attacked during
The December 2004 status of the group of 13 named root name servers is shown in Table 3.1.43
They are designated by the letters A through M. The A-root server,, was until
recently the primary but, as noted above, has been replaced by a hidden primary. See Box 3.4.
Although the home locations of some of the root servers are the result of historical accident, they
were originally determined by analysis of network traffic flows and loads, seeking to have at least one
server "close" in terms of message communication time to every location on the network. It is important
to have root servers distributed so that they provide a uniform level of service across the network.44
Having a root server in a well-connected area is fairly unimportant to that area, since users there are likely
to be able to reach several servers. By contrast, if an area is fairly isolated from most of the network, it is
important that the ISPs that serve it acquire sufficient connectivity to enable the area's users to access one
or more root servers at all times.

41 The Vendare Group. 2004. "The Vendare Group, Online Media and Marketing Company, Acquires Search-
Services Provider" Press release, September 17. Available at <>.
42 The F-root server (which includes its satellites), one of 13 root servers, answered more than 272 million queries
per day according to the Internet Systems Consortium in January 2004. See <>.
43 For the current version of this table, see <>. That Web site also contains links to the
root name server operators and to other relevant information.
44 See Lee et al. 2003. "On the Problem of Optimization of DNS Root Servers Placement." Passive Measurement
and Analysis Workshop. Available at <>.

Prepublication Copy--Subject to Further Editorial Correction
BOX 3.4

VeriSign, Inc., founded in 1995, describes itself as providing "intelligent infrastructure services
that support the digital economy." Its acquisition in 2000 of Network Solutions, Inc. made it the registry
for three gTLDs: .com, .net, and .org and the operator of the A-root and the J-root name servers.
It currently operates the A-root and J-root name servers, the hidden root primary, and the name
servers for the .com, and .net TLDs. At the end of 2004, its Naming and Directory Services unit
managed a database of over 38 million names in the .com and .net gTLDs; owned and maintained 13
gTLD name server sites around the globe that handled the more than 14 billion transactions per day for
those two gTLDs; and provided access to the .com and .net gTLD registries for more than 150 ICANN-
accredited registrars that submit over 100 million domain name transactions daily to its Shared
Registration System.
As a registrar--through a subsidiary renamed Network Solutions, Inc. in 2003--it registered
more than 500,000 new domain names during the second quarter of 2002, and an additional 700,000
names were renewed or extended. More than 8.7 million active domain names in .com and .net were
under management by VeriSign's registrar. However, in October 2003, it sold Network Solutions to a
private equity firm for $100 million.45

Considerations of this type are both complex and important but were not sufficient to determine a
unique set of locations for the home sites of the 13 root servers. However, as the Internet has evolved,
these original locations became less satisfactory, which has been one of the reasons for the proliferation
by some operators--notably C, F, I, J, K, and M in Table 3.1--of satellite sites at different locations.
These satellite sites use anycast addressing (see Box 3.1), which enables servers with the same IP address
to be located at different points on the Internet. Copies of the F-root server, for example, can be placed at
multiple locations around the world. The widespread distribution of anycast satellites of the 13 root
servers has improved the level of service provided to many previously less well served locations.
Some have believed that 13 root name servers are too few to meet requirements for reliability and
robustness, which requires sufficient capacity distributed widely enough to protect against system or
network failures and attacks. Others have believed that more root servers are needed to satisfy
institutional requirements. Their concern arose from the belief that to be a full participant in the Internet, a
nation or region must have its own root name server. While technical reasons46 make it difficult to
increase the number of root name server IP addresses beyond 13, the rapidly increasing use of anycast
addressing has enabled the root name server system to respond to both the technical and institutional
requirements through the addition of multiple geographically distributed anycast root server satellites.
Even so, since it addresses the distribution only of servers and not of the operating institutions, the
location issue is likely to continue to add some political acrimony to any selection process that might
follow from the withdrawal of one of the current operators. (See Section 5.3.)
There is no standard hardware and software implementation of the root name servers. Each of the
responsible organizations uses its preferred equipment and operating and file systems, although most of
them run some version of the BIND name server software.47 Although there might be some operational
advantages to having standard implementations, most Internet technologists believe that variation in the
underlying hardware and software of the root name server system is highly desirable, since it ensures that

45 This information was obtained from <> on February 13, 2005.
46 See "DNS Message Format" in Section 3.2.4 for an explanation of the technical limitations.
47 Since BIND 8 and BIND 9 have different code bases, this is less of a problem than it would be if they were
identical. The K root runs NSD name server software from NLnet labs of Amsterdam.

Prepublication Copy--Subject to Further Editorial Correction
TABLE 3.1 The 13 Root Name Servers and Their Anycast Satellites as of December 2004
Name Operator
VeriSign Naming and
Dulles, VA
Directory Services
B Information
Marina del Rey, CA
University of Southern
Cogent Communications
Herndon, VA; New York City; Chicago; Los Angeles
University of Maryland
College Park, MD
NASA, Ames Research
Mountain View, CA
F Internet
Ottawa; Palo Alto; San Jose, CA; New York City; San
Consortium, Inc.
Francisco; Madrid; Hong Kong; Los Angeles; Rome;

Auckland; Sao Paulo; Beijing; Seoul; Moscow; Taipei;
Dubai; Paris; Singapore; Brisbane; Toronto; Monterrey;
Lisbon; Johannesburg; Tel Aviv; Jakarta; Munich;
Osaka; Prague
U.S. DOD Network
Vienna, VA
Information Center
U.S. Army Research
Aberdeen, MD
Stockholm; Helsinki; Milan; London; Geneva;
Amsterdam; Oslo; Bangkok; Hong Kong; Brussels;
Frankfurt; Ankara; Bucharest; Chicago; Washington,
DC; Tokyo; Kuala Lumpur
VeriSign Naming and
Dulles, VA (2 locations); Mountain View, CA; Sterling, USA
Directory Services
VA (2 locations); Seattle, WA; Amsterdam; Atlanta,
GA; Los Angeles, CA; Miami; Stockholm; London;
Tokyo; Seoul; Singapore
Reseaux IP Europeens--
London; Amsterdam; Frankfurt; Athens; Doha; Milan;
Network Coordination
Reykjavik; Helsinki; Geneva; Poznan; Budapest
WIDE Project
Tokyo; Seoul; Paris
SOURCE: This table derives directly from information provided at <> as accessed on
February 13, 2005.

an error in a particular piece of hardware or software will not lead to the simultaneous failure of all of the
root servers.
As the number and rate of queries to the root name servers have increased, hardware and software
upgrades have enabled the servers to keep up.48 However, the pace of inquiries is likely to continue to
grow and it is conceivable that it could, in principle, exceed the capacity of a system comprising even the
most powerful single computers. Because of anycasting and multiprocessing, through which a root server
can comprise multiple processors, the number of computers at each root-server address is effectively
unrestricted. Moreover, it is plausible to expect continued improvements in computing and
communications performance. Consequently, it is unlikely that the query load will ever be able to outrun
the computing capacity of the 13 named root name servers and their satellites.

48 This is only a portion of the total query load to the root zone file because of caching of queries by many name
servers, and especially those of large ISPs, as noted above.

Prepublication Copy--Subject to Further Editorial Correction

3.3.3 Institutional Framework of the Root Zone

Because the root is central and critical to the operation of the DNS, decisions about the root zone
and the root name servers are of substantial importance, and the institutions that bear responsibility for
them take on an important role as stewards of the DNS.
Those institutions carry out four critical functions:

1. Deciding what new or revised entries will be permitted in the root zone file;
2. Creating the root zone file, keeping it current, and distributing it to all the root name servers;
3. Selecting the locations and the operators of the root name servers; and
4. Establishing and continually and reliably operating the root name servers.

The diverse collection of institutions that performs these functions includes a not-for-profit
corporation--ICANN; a U.S. government agency--the Department of Commerce; a corporation--
VeriSign; and an informal group consisting of the commercial, non-commercial, and governmental root
name server operators.

Approving the Root Zone File--U.S. Department of Commerce and ICANN

The fundamental importance of the root zone file to the operation of the DNS and, therefore, of
the Internet means that special attention is paid to the process by which new or revised entries to the file
are authorized. Some process must be in place to decide whether a change is legitimate; otherwise,
persons or organizations with malicious motives or inadequate capabilities could make or revise entries.
Currently the authority to make changes lies with the U.S. Department of Commerce (DOC).49 However,
the day-to-day operational responsibility is at VeriSign. As part of the DOC's delegation of responsibility
to ICANN (see Box 3.5), the process of authorization for new or modified entries in the root zone was
changed. All such requests go first to the Internet Assigned Numbers Authority (IANA) (now a function
of ICANN) and then to the DOC for final approval. Once the addition or change is approved, the DOC
notifies IANA and VeriSign. VeriSign Naming and Directory Services then makes the change in the
hidden primary, which distributes the changed root zone file to the other root name servers (see
"Operating the Root Name Servers" in Section 3.3.3).
There are three kinds of changes to the root zone file. The first kind of change is a modification of
the data associated with an existing resource record. This might entail a change in the IP address of one
or more of the name servers resulting, for example, from a change in the network service provider. The
data entry process is straightforward, and so such changes should be routine and rapid. However, they do
require a stage of verification to ensure that the request is legitimate and not, for example, an effort by a
third party to capture a top-level domain. Nevertheless, such requests should be processed in a few hours
or days.
The second kind of change is a shift in the responsibility for a TLD, typically a ccTLD. In such
redelegation cases, the questions that must be resolved may be more difficult and time-consuming. (See
"Selecting the Organizations Responsibile for the TLDs" in Section 3.4.3 for a discussion of the issues.)
In particular, they may entail judging which organization has the "right" to operate the registry for a
ccTLD, which can become embroiled in national politics. These changes will generally take longer but
should proceed according to a formal and transparent process.
The third kind of change is the addition of data about a new TLD to the root zone file (see
"Selecting the Organizations Responsible for the TLDs" in Section 3.4.3). This is a single process for new
gTLDs. An ICANN selection process recommends both the domain name that shall be added and who

49 See Section 2.7 for a history of that authority.

Prepublication Copy--Subject to Further Editorial Correction

BOX 3.5
The Internet Corporation for Assigned Names and Numbers

As described in Chapter 2, the Internet Corporation for Assigned Names and Numbers (ICANN)
is a not-for-profit corporation founded in October 1998 in California by a group of individuals
interested in the Internet. It was sponsored by the U.S. Department of Commerce (DOC) to serve as a
technical coordination body for the Internet. Consequently, ICANN has assumed responsibility (under
a memorandum of understanding--and its six amendments--with the DOC) for a set of technical
functions previously performed under U.S. government contract by the Internet Assigned Numbers
Authority (IANA) and other groups. However, in practice, many of its most important and
controversial activities have been as a policy-setting, rather than as a technical coordination, body.
IANA continues as a function of ICANN with overall administrative responsibility for the
assignment of IP addresses, autonomous system numbers, top-level domains, and other unique
parameters of the Internet. In addition, ICANN is charged with coordinating the stable operation of
the Internet's root server system.
ICANN's primary governing body is its board of directors, which now comprises 15 voting
members and the president, ex officio. Dissatisfaction with the composition of the board and with the
nature of the selection process inspired a reform effort that resulted in new bylaws that guided the
selection of a new board in 2003. (See Section 5.2.4, Alternative F, for a description of this reform.)
ICANN has three supporting organizations: the Address Supporting Organization (ASO), which
deals with the system of IP addresses; the Country-Code Names Supporting Organization (ccNSO),
which focuses on issues related to the country-code top-level domains; and the Generic Names
Supporting Organization (GNSO), which handles issues related to the generic top-level domains. In
addition, it has four advisory committees: the At-Large Advisory Committee (ALAC) for the Internet
community at-large; the DNS Root Server System Advisory Committee (RSSAC) for root server
operators; the Governmental Advisory Committee (GAC) for governments; and the Security and
Stability Advisory Committee (SSAC) for security. There is also the Technical Liaison Group (TLG)
for standards-setting organizations and an Internet Engineering Task Force liaison who provides
technical advice to ICANN. The supporting organizations and advisory committees together represent
a broad cross section of the Internet's commercial, technical, academic, non-commercial, and user
communities and advise the board on matters lying within their areas of expertise and interest.
As part of the reform effort, ICANN adopted a new mission statement:

The mission of ICANN is to coordinate the stable operation of the Internet's unique identifier systems. In
particular, ICANN:
1. Coordinates the allocation and assignment of three sets of unique identifiers of the Internet--domain
names, IP addresses and autonomous system (AS) numbers, and protocol ports and parameter numbers.
2. Coordinates the operation and evolution of the DNS's root name server system.
3. Coordinates policy-development as reasonably and appropriately related to the performance of these

ICANN is open to the participation of any interested Internet user, business, or organization. It
holds several meetings a year at locations around the world.50
ICANN has been the subject of many controversies regarding its governance, its processes, and
its decisions since its founding. The primary issues are discussed in Section 5.2. A good sense of the
full range of controversies that have surrounded ICANN can be obtained from
<>, CircleID (<>), and ICANNFocus

50 Information about ICANN was derived from <> on February 13, 2005.

Prepublication Copy--Subject to Further Editorial Correction
(<>), through the resources that can be linked to from these sites, and from
innumerable articles written about ICANN.

shall operate it. It is a two-step process for new ccTLDs. Who will be responsible for operations is a
separate recommendation from the decision to add a ccTLD name to the list of ccTLDs. Such changes
should take place within a few days after the appropriate decisions have been made.

Maintaining the Root Zone File--VeriSign

VeriSign, as the operator of the hidden primary root zone server, is responsible for maintaining
the root zone file and for distributing it to the secondary servers. It performs this function under the terms
of a memorandum of understanding (MoU) with the DOC.
Since any errors in the root zone file can affect large numbers of sites and users, accurate and
error-free preparation and distribution of the file are essential. In the first instance, this is a human
function. Someone must enter additions and updates into the database, create a new zone file, and check
it for errors. Individuals at the secondary sites must check to ensure that the file has not been corrupted
during its distribution. However, computer techniques and aids can be used to support this process and
reduce the demands on humans. Furthermore, regular queries of the root by each of the TLD operators
can be used to test the entries corresponding to their TLDs and provide further assurance that no
undetected errors are present in the file.

Selecting the Root Name Server Operators--Self-Selection

The current root name server operators were not selected through a formal evaluation and
qualification process, although they play a fundamental role in ensuring the availability and reliability of
the root. Rather, the group is the cumulative result of a sequence of separate decisions taken over the
years since the establishment of the DNS. It is a loosely organized collection of autonomous institutions
whose names are given in Table 3.1. Ten of them are based in the United States. Of those, three are
associated with the U.S. government (National Aeronautics and Space Administration (NASA),
Department of Defense (DOD), and the U.S. Army), two are universities (University of Maryland and
University of Southern California), two are corporations (VeriSign and Cogent Communications), and
two are not-for-profits (ISC, Inc. and ICANN). Three are based outside the United States: one in Sweden,
one in the United Kingdom, and one in Japan.
One of the responsibilities that ICANN assumed under its agreement with the DOC is
coordinating the stable operation of the root server system. To do so, it established the DNS Root Server
System Advisory Committee, whose responsibilities were spelled out in ICANN's bylaws (Article VII,
Section 3(b)).

The responsibility of the Root Server System Advisory Committee shall be to advise the
Board (of ICANN) about the operation of the root name servers of the domain name system.
The Root Server System Advisory Committee should consider and provide advice on the
operational requirements of root name servers, including host hardware capacities, operating
systems and name server software versions, network connectivity and physical environment . . .
should examine and advise on the security aspects of the root name server system . . . [and] should
review the number, location, and distribution of root name servers considering the total system
performance, robustness, and reliability.
The parties will collaborate on a study and process for making the management of the
Internet (DNS) root server system more robust and secure.


Prepublication Copy--Subject to Further Editorial Correction
ICANN's intent is to enter into an MoU51 with each server operator that will spell out the root name
server performance requirements, such as service levels, reliability, and security. However, as of
February 2005, no MoUs had yet been signed.
In the absence of an agreed oversight role for ICANN, there is, at present, no formal process for
selecting a new root name server operator if one of the incumbents should withdraw, although it is clear
that the set of remaining operators could, and probably would, work together to make the selection. (See
Section 5.3 for a discussion of this issue.)

Operating the Root Name Servers--The Root Name Server Operators

The role of the operators of the 13 root name servers is to maintain reliable, secure, and accurate
operation of the servers containing the current root zone on a 24-hour-a-day, 365 days-per-year-basis.
Each server is expected to have the capacity to respond to many times the rate of queries it receives and
must increase its capacity at least as fast as the query rate increases. An attempt to define the
responsibilities of the root name server operators was made in RFC 2870, issued in June 2000,52 but the
ideal that it describes has not been achieved.
Historically, the operators of the root servers have not charged fees for resolving Internet address
queries, instead obtaining support in other ways for the substantial costs they incur in providing the
service. These operators choose to do so either because (1) they believe that operating a root server is a
public service (and sufficiently inexpensive that they can afford the cost) or (2) they believe that operating
a root server conveys a business, or other, advantage to them. Nevertheless, it is a valuable service,
whose provision is a little-known and little-appreciated gift in kind to all users of the Internet.

3.3.4 Assessment

Conclusion: The system of DNS root name servers currently responds to several billion queries
per day and does so reliably and accurately as shown by its virtually uninterrupted availability and the
very low occurrence of root-caused misdirections. Because the majority of queries are served from cached
answers lower in the hierarchy, the entire DNS responds to many times that number each day with
correspondingly good results. However, the robust operation of the root name servers is potentially
vulnerable to an excessive query load, either inadvertent or malicious, that might slow down their
responses or cause them to fail to respond.

Data collected about root name server operation has revealed that a substantial fraction--between
75 percent and 97 percent--of the load on those servers may be the result of erroneous queries.53 These
errors fall into three categories: stupid--for example, asking for the IP address of an IP address; invalid--
for example, asking for the IP address of a nonexistent domain; and repetitive--for example, continuing
to send an incorrect query even after receiving a negative response. Analysis has revealed that the
sources of many of these errors lie in faulty resolver or name server software and faulty system
management that misconfigures name servers or resolvers and does not monitor performance closely
enough to catch errors. The latter is not purely a technical issue but poses an institutional issue.54

51 The DNS Root Server System Advisory Committee has drafted a model MoU, available at
52 See Randy Bush, Daniel Karrenberg, Mark Kosters, and Raymond Plzak. "Root Name Server Operational
Requirements." RFC 2870. June 2000
53 The 97 percent figure is from the Wessels and Fomenkov study, op. cit. However, according to an anonymous
reviewer, VeriSign's experience on its two servers is closer to 75 percent.
54 These analyses have been sponsored by the Cooperative Association for Internet Data Analysis (CAIDA) and
have been reported in many publications. In addition to the Wessels and Fomenkov study, see, for example: Nevil
Brownlee, k.c. claffy, and Evi Nemeth. 2001. "DNS Measurements at a Root Server." CAIDA, Proceedings of IEEE

Prepublication Copy--Subject to Further Editorial Correction
Software developers, system administrators, and users generally have few incentives to make an effort to
prevent the occurrence of this extra load.
In addition, the system of root name servers may be vulnerable to malicious attempts to overload
it. In October 2002, the root name server system was subjected to such a, denial-of-service attack that
sought to swamp the system with queries.55 Eight of the 13 servers were inaccessible for an hour or more,
but the remaining 5 served the Internet without observable degradation. Although the system successfully
resisted this attack, which lasted only an hour and a half, it should serve as a warning about the potential
for longer and more sophisticated attacks in the future.

Conclusion: The root name servers receive far too many incorrect or repetitive queries,
increasing the load that they must serve. This unnecessary load arrives at the root name servers because
many sites on the Internet employ faulty software, misconfigure their resolvers or name servers, or do not
manage their systems adequately. The root name server operators, however, lack the means to discipline
the sites at fault.

Conclusion: The root name servers are subject to malicious attack, but through overprovisioning
and the addition of anycast satellites have substantially reduced their vulnerability to denial-of-service
attacks. Furthermore, the widespread caching of the root zone file and its long time to live mean that the
DNS could continue to operate even during a relatively long outage of most or all of the root name
servers and their satellites.

Recommendation: To be able to continue to meet the increasing query load, both benign and
malign, the root name server operators should continue to implement both local and global load balancing
through the deployment of anycast satellites.

Conclusion: Notwithstanding the deployment of anycast satellites, the concentration of root
servers in the Washington, D.C., area and, to a lesser extent, in the Los Angeles area remains a risk, since
anycast satellites depend on the 13 base root name servers to obtain root zone information.

Conclusion: The system of root name servers lacks formal management oversight, although the
operators do communicate and cooperate. Not everyone would agree that formal oversight is desirable.
Should one or more of the current root name server operators withdraw from that responsibility, or fail to
exercise it reliably, effectively, or securely, there would be no responsible organization or formal process
for removing the failed operator or for recruiting and selecting a replacement. In their absence, the
informal, collegial processes that led to the current group of operators would likely continue to be used.

Conclusion: The root name server operators have provided effective and reliable service to the
community of Internet users without any form of direct compensation for that service from the users.
With the growth in the scale and economic importance of the Internet, and with the expense of ensuring
secure operation, the root name server operators face increasing costs and potential liabilities that some, at
least, may find too great to meet without compensation.56 Indeed, the current system for maintaining root

Globecom, San Antonio, Texas. November. pp. 1672-1676. Also see
55 See, for example: David McGuire and Brian Krebs. 2002. "Attack on Internet Called Largest Ever." Washington
, October 22. Three root server operators coauthored an authoritative report about the attack. It can be found at
56 As President's Report: ICANN--The Case for Reform, February 23, 2002, notes ". . . some organizations that
sponsor a root name server operator have little motivation to sign formal agreements [with ICANN], even in the
form of the MOU that is now contemplated. What do they gain in return, except perhaps unwanted visibility and the
attendant possibility of nuisance litigation? They receive no funding for their efforts, so why should they take on any

Prepublication Copy--Subject to Further Editorial Correction
servers may well face additional economic pressures, as well as technical ones, as the volume of Internet
traffic increases, especially if the number of TLDs were to be expanded.

The root zone must be kept secure and robust in the face of possible future threats. Moreover, the
continued stable management and financing of root zone operations must be ensured for the DNS to
function. These challenges and approaches to meeting them are discussed in Section 5.3.


The portion of the DNS hierarchy that has captured the most public attention and attracted the
greatest controversy is the level just below the root, the top-level domains. As noted above, in February
2005 there were 258 such domains in the two major categories defined in the early days of the DNS (see
Section 2.2.1): 243 country code top-level domains (ccTLDs) and 15 generic top-level domains (gTLDs).
These TLDs are, in turn, the top of a hierarchy of second-level domain names. Typically, the commercial
gTLDs have very flat structures with many second-level names, but the others vary widely, some having
very deep structures. The ccTLDs also have a wide range of structures, some having several levels of
hierarchy, which may be structured geographically or generically.

3.4.1 Characteristics of the TLDs


The 243 ccTLDs are each associated with a nation (or external territory) and designated by the
two-letter abbreviation for that entity assigned by the International Organization for Standardization in
ISO 3166-1.57 Examples of ccTLDs are .ke for Kenya, .jp for Japan, and .de for Germany.58 The
expectation was that a ccTLD would signify a location and would be supervised and administered by an
organization in the corresponding country, and be limited to registrants with a presence there; however,
these criteria are not effectively enforced. Moreover, a number of small nations have licensed
administration of their domains--for example, .tv for Tuvalu and .cc for the Cocos Islands--to
commercial bodies that register anyone who wishes to use that ccTLD's two-letter domain. Such ccTLDs
are for all practical purposes equivalent to commercial gTLDs. This equivalence is further explored
in"Recharacterizing TLDs" in Section 3.4.1.
The largest ccTLDs are listed in Table 3.2, as are most of those ccTLDs, shown in boldface, that
have contracted for commercial use of their domains. The .us domain, which had been limited to third-
and fourth-level registrations in a locality-based hierarchy, is shown in italics. In April 2002, registration
at the second level in .us was opened with the intent that any U.S. citizen or resident and any business or
organization with a bona fide presence in the United States could register a domain name in .us.59

contractual commitments, however loose?" The same logic raises questions about their incentives to continue to
operate the root name servers.
57 See <> for a list of the abbreviations.
58 There are a few exceptions to the use of ISO 3166-1 two-letter abbreviations. For example, the United Kingdom
is assigned .uk rather than .gb (for Great Britain). ICANN has also assigned ccTLD two-letter codes to each of the
Channel Islands and to Ascension Island, which are not in ISO 3166-1, because it was anticipated that they would be
added to the list.
59 The locality-based use of .us seems to be declining, which raises questions about the relative merits of
registrations at the second level and their associated revenue motivations versus the benefits of the locality-based
structure of .us.

Prepublication Copy--Subject to Further Editorial Correction
Table 3.2 shows the number of domains registered at the first available level under each top-level
domain in February 2003. The total at that time was just over 19 million. By December 2003, almost 24
million domains had been registered in ccTLDs, and by November 2004 it had reached 25.6 million. As
noted above, some country code TLDs have further generic or geographic substructures. In those cases,
the count of domains is the sum of those registered under each second- or third-level domain, depending
on the highest level at which registration by the general public is permitted.


There are 15 gTLDs, 8 that date from the early years of the, .com, .org,
.edu, .gov, .mil, .int, and .arpa--and 7 that were authorized by ICANN in November
2000. As their designations suggest, the expectation was that registration in the original gTLDs would be
by types of organizations. Commercial organizations were expected to register in .com, accredited
educational organizations in .edu, network infrastructure organizations in .net, U.S. government
organizations in .gov and .mil, and international treaty organizations in .int. The .org TLD was
created for organizations that did not fit into one of the other gTLDs. The designation .arpa was
assigned for network infrastructural use.
In announcing the authorization of the seven new, .info, .name, .aero,
.museum, .coop, and .pro, ICANN distinguished between sponsored and unsponsored TLDs.
Those that are sponsored have a not-for-profit organization representing the community of potential
registrants. The charters of the sponsored TLDs specify that registrants are restricted to those satisfying
criteria appropriate to that community. Thus, .aero is restricted to people, entities, and government
agencies that (1) provide for and support the efficient, safe, and secure transport of people and cargo by
air and (2) facilitate or perform the necessary transactions to transport people and cargo by air.
Registrations in .museum are restricted to entities that are recognized by the International Council of
Museums as museums, professional associations of museums, or individuals who are professional
museum workers. And registrations in .coop are restricted to members of the international cooperative
movement, which is further defined by membership in one or more of eight specific classes. ICANN
allows unsponsored TLDs to be either restricted or unrestricted. Thus, .info is unrestricted--anyone
can register a name in .info, whereas .name is restricted to individuals and .biz is restricted to bona
fide business or commercial uses. Determination of who is eligible to register is up to the domain
operator operating under the terms of its agreement with ICANN.
The 15 generic TLDs are shown in Table 3.3 together with, for each, its type and purpose, the
organization responsible for its operation, and the number of second-level domains that are registered in
it. Note, however, that in many domains there are far more registrants at the third and lower levels.

Prepublication Copy--Subject to Further Editorial Correction
TABLE 3.2 Large Country Code Top-Level Domains
Country Name
Registered Domainsa
United Kingdom
United States
Republic of Korea
Cocos (Keeling) Islands
Western Samoa
Russian Federation
South Africa

TOTAL (including ccTLDs not listed above)
NOTE: Individual domain data is for February 2003; the total is for November 2004.
aSOURCE: ICANN's Budget for Fiscal Year 2003-2004. See
<>. Domain name counts where not
available are estimated from the average of all other ccTLDs for which data is available (from Web sites or other
direct information). These estimated counts represent about 16 percent of all ccTLD domain names. The total (for
November 2004) was provided by Matthew Zook, Department of Geography, University of Kentucky.

Recharacterizing TLDs

Although a distinction between generic TLDs and national or country code TLDs is widely
accepted and used in policy discussions, the reality of practice is that the distinctions have been
significantly eroded.
Among the generic gTLDs,,, and .gov--are currently restricted to new
registrations by U.S. organizations and, in principle, could have been established as second-level domains
under the .us ccTLD.
As noted above, among the ccTLDs, a number--including .cc, .tv, .bz, .ws, .nu,
and .to--actively seek global registrants and function as though they were gTLDs.

60 Some non-U.S. educational institutions, such as the University of Toronto and the United Nations University,
retain their registrations from an earlier, less restrictive time. Also, registrations from foreign but U.S.-accredited
educational institutions are currently being accepted.

Prepublication Copy--Subject to Further Editorial Correction
Moreover, the registrants in some gTLDs have not been limited to those originally expected. In
practice, .com, .net, or .org have all been operating as unrestricted TLDs--any person or
organization can register in them.
Since many policy issues concern the desirable number and type of TLDs, it is useful to introduce
a characterization of TLDs that better captures the reality of their current state.
In Table 3.4, the TLDs are recharacterized into eight types, designated by the numbers 1 to 8,
which are given in the first column. The second column, "TLD," indicates whether the domain is
currently considered a gTLD or a ccTLD.
The "Scope" column shows whether the TLD is open to registrants from anywhere on the
globe--global--or is primarily for those who are located within the national boundaries of the country--
national. (Many ccTLDs that are primarily national will accept some non-national registrants, but usually
they must have an association with the country.) In addition, the .arpa TLD is open only to register
elements of the infrastructure of the Internet, so its scope is designated as infrastructural.

TABLE 3.3 Generic Top-Level Domains in 2004
gTLD Type Current
Unsponsored Unrestricted
VeriSign GRS
Public Interest Registry (PIR) as of
January 1, 2003
U.S. accredited higher
educational institutions
U.S. military
U.S. DOD Network Information
U.S. governments
U.S. General Services
Administration (GSA)
Sponsored Internet
Internet Architecture Board/
Internet Assigned Numbers
Authority (IANA)
Unsponsored International
Aflilias, LLC
Unsponsored Businesses
Global Name Registry
Unsponsored Professionals
.museum Sponsored
Air-transport industry
DotCooperation, LLC

aSOURCE: ICANN's Budget for Fiscal Year 2003-2004. See
<>. Domain count data provided courtesy of
Matthew Zook, Department of Geography, University of Kentucky. Ongoing data series is available at his Web site
bNeulevel is a joint venture of Neustar, Inc. and Australia-based Melbourne IT, Ltd. started accepting registrations in the United States on June 1, 2004, from licensed MDs, lawyers, and CPAs;
added Canadian and U.K. professionals in Septemeber 2004; and added engineers in February 2005.
dData provided by Cary Karp, president and CEO, MuseDoma, private communication, February 20, 2004.
eAs listed on the .aero Web site in February 2005. See <>
fAs of February 2004. Carolyn Cooper, dot Coop Operations Center, private communication, March 17, 2004.
gTotal reported on the Zooknic Web site (<>) February 2005.

Prepublication Copy--Subject to Further Editorial Correction
The "Restriction" column in Table 3.4 indicates whether the generic TLD or, within national
boundaries, the ccTLD has second- or third-level domains that are intended to be restricted to members of
specific communities (however, the enforcement of these restrictions varies widely). Most ccTLDs have
no restrictions on registrations at the second level, but some, such as .ar (Argentina) and .au
(Australia), and the .us (United States) until recently, register only at the third level under a limited
number of restricted second-level domains, such as and for commercial organizations.
However, Great Britain, .uk, has some second-level domains that are restricted, such as and, and others, such as, that are not. As with other aspects of the DNS, restriction within
ccTLDs is not really "yes" or "no," but more accurately "yes," "no," or "some." That situation is
reflected in Table 3.4 by treating second-level domains of such ccTLDs as though they were "separate"
TLDs--see the examples in 7 and 8.

TABLE 3.4 Types of Top-Level Domains
Type TLD
Restriction Sponsor
1 gTLD
No .com, .net,
.info, .org
2 gTLD
Yes .aero,
3 gTLD

No .biz, .name,
.pro, .int
4 gTLD National
Yes .mil, .gov,
5 ccTLD
- .cc, .tv, .bz
6 ccTLD National

- .jp, .fr,
7 ccTLD National
8 gTLD Infrastructural
Yes .arpa

The "Sponsor" column indicates whether an organization representative of a community has
responsibility for managing a gTLD and for enforcing registration restrictions, if any. The concept of
sponsor for a ccTLD is less clear. To some degree, for example, the governments of the United States and
of France, for example, can be viewed as the sponsors of their ccTLDs. The corresponding entries are left
Thus, among the 15 current gTLDs, 4 are of Type, .net, .org, and .info.
Type 2 includes three, .aero, and .coop. There are four in Type 3: .name,
.biz, .pro, and .int. That leaves three in Type, .mil, and .gov, which are
discussed below as ccTLDs--and one in Type 8, .arpa.
It has not been possible to assign a type to each of the 243 ccTLDs, but examples of each of the
four types have been identified in Table 3.4.
Type 5 comprises many ccTLDs that function as generic TLDs. As noted above, they are
generally small countries that have recognized, or been shown, that their two-letter designation can be
marketed globally to companies and individuals who would find it commercially or personally valuable.
Thus, the lease of the domain name of Tuvalu (population: 11,000) will provide that Pacific Island nation
with $50 million in royalties over the next dozen years. The .TV Corporation, a subsidiary of VeriSign,
in April 2004 listed on its site such "premium registrations" as, which is available for $1 million,
and, which would cost the registrant $250,000. The administrator of this TLD is the Ministry of

Prepublication Copy--Subject to Further Editorial Correction
Finance and Tourism of Tuvalu, although in fact, the ministry appears to have delegated all of its
decision-making authority to the .TV Corporation as a part of the lease. Other nations are managing their
globally available domains themselves. Western Samoa (population: 178,000), another Pacific island
nation, markets .ws directly and handles the registration locally. Western Samoa had two ISPs and 3000
Internet users in 2002.61
While some ccTLDs function as gTLDs, Type 4 comprises the three gTLDs that function like, .gov, .mil. All three are limited to registrants in the United States that are, in
turn, restricted to specific communities--primarily accredited higher educational institutions,62 civilian
government63 agencies, and federal military agencies.
The remaining two types distinguish between restricted and non-restricted ccTLDs. Almost all
ccTLDs are unrestricted at the second and third levels and fall into Type 6, but some, such as Australia,
Argentina, and Austria, have introduced categories resembling the gTLD categories as their second levels.
They belong to Type 7. Thus, Australians cannot register directly in the .au domain but must instead
use the category appropriate to their status: for commercial organizations, for
associations, for individuals, and so on.
Thus, the apparently simple distinction between gTLDs and ccTLDs is really more complex. The
differences among TLDs lie in the differences in the policies that they operate under, not in whether they
were originally associated with a country code or a generic category.

3.4.2 Technical System of the TLDs

Every TLD has an associated zone file, which is stored in name servers whose addresses are
recorded in the root zone file. ICANN requires that there be at least two name servers for each ccTLD
and at least five for each gTLD with which it has agreements. Some TLDs have as many as 13 name
servers, depending on the query load, the need for security against attack, and their desire to improve
access by their users. Each name server is implemented on one or more computers, most of which run a
version of BIND.64
The zone files on all TLDs are larger, generally very much larger, than the root zone file. At the
extreme, .com contains more than 33 million second-level domains, but even Greenland has more than
1200 domains registered. Moreover, because of the hierarchical nature of DNS search, the ease of
caching the labels and associated resource record sets in the small root zone file, and the long TTLs
within that file, the TLD name servers receive a greater number of queries than the root name servers. For
example, according to VeriSign .com and .net alone receive over 14 billion queries per day, while the
root receives several billion queries per day.
TLD name servers face specific unique technical issues distinct from those that involve the roots.
One issue is increased traffic resulting from low TTLs on second-level zone records. A popular second-
level zone can increase traffic to its parent TLD name servers by lowering its TTL, effectively defeating
the DNS's caching mechanism. (The root name servers do not suffer from this potential problem to the

61CIA. 2004. The World Factbook. Available at <>.
62 As noted previously, some non-U.S. academic institutions have been "grandfathered in" or will be registered if
they are accredited in the United States. Also, there are some other exceptions such as the Thomas Jefferson High
School for Science and Technology, whose domain name is
63 Originally it was limited to federal agencies. It is now open to registration by state and local governments and
Native Sovereign Nations.
64 However, the name server for .org (and for some other TLDs) is operated by UltraDNS, which uses its own
proprietary name server software; and VeriSign, which runs .com, .net, .bz, and .tv, also uses its own
proprietary name server software, Atlas.

Prepublication Copy--Subject to Further Editorial Correction
same extent, since TLD name servers give out mostly referrals. As a result, name server caches
throughout the Internet retain the copy of the TLD records from the root zone with their 48-hour TTLs.65)
TLDs probably receive some bogus queries, but perhaps not as many as the root, since at least a
portion of the query must be correct--the TLD name. Their name servers face the same vulnerabilities
that are faced by the root name servers. However, although the effect would be significant if the
operations of large TLDs such as .com or .uk were to be interrupted, the consequence of a short
interruption of most TLDs, individually, would not be significant for the overall Internet. An attack that
could take out a substantial number of TLD servers would be significant but difficult to sustain because of
the number of distinct targets that would be involved.

3.4.3 Institutional Framework of the TLDs

The effective operation of the top-level domains requires that an institutional framework perform
four functions:

1. Deciding which new TLDs will enter the root zone file,
2. Determining the organization to be responsible for a TLD,
3. Determining the organization to operate a TLD's name server, and
4. Operating the TLD registry.

Many different organizations, international and national, governmental and private, for-profit and
not-for-profit, including ICANN, VeriSign, and the DOC, are engaged in these activities. Their processes
and interactions are complex and, often, controversial.

Selecting New TLDs

The original DNS design assumed a single, unified, name space (in which the set of names that
one user is able to look up is the same set of names that any other user is able to look up).66 Unless
uncoordinated entry into the root zone file is permitted, which would require that the operators of the root
servers recognize any TLD that chooses to operate, some entity must decide how many and which TLDs
there will be and who will operate them. "Selecting new TLDs" means deciding which new TLDs will
enter the root zone file. As described above, that decision is made by the U.S. Department of Commerce
upon the recommendation of ICANN. Therefore, it is in the first instance a decision for ICANN. The
decision process that is employed differs between ccTLDs and gTLDs.

There is generally no need for a complex decision process for entry of ccTLDs since, as noted
earlier, they are available to countries and external territories represented by country codes in ISO 3166-1.
This list has been used as the authoritative source for country codes because, as was stated in RFC 1591,
"the IANA is not in the business of deciding what is and what is not a country."67 When new entities are
assigned a two-letter identifier by the ISO, that entity is automatically entitled to have a ccTLD.
However, two situations have arisen that cannot be resolved solely by this rule. The first occurs when a
country is removed from the ISO list. The two-letter code for the Soviet Union, SU, was removed from
the list in September 1992 and placed in "transitional reserved status," which means that its use should be

65 Information provided by an anonymous reviewer.
66 It is this assumption that alternative or multiple roots violate. This discussion assumes a single root.
67 Jon Postel. 1994. "Domain Name System Structure and Delegation." RFC 1591. Internet Engineering Task Force.
March. p. 6. Available at <>.

Prepublication Copy--Subject to Further Editorial Correction
stopped as soon as possible. But, although this issue is on ICANN's agenda, it has not yet addressed the
complicated issues that arise when a country is removed from the ISO 3166-1 list. Indeed, it is still
possible to register in the .su domain, and this domain remains in the root.
The second situation occurs when an international entity requests a ccTLD that is not on the ISO
3166-1 list. In July 2000, the European Commission wrote to ICANN requesting the inclusion of the .eu
domain in the DNS root.68 Its request noted that the ISO had agreed to use of the two-letter code EU as a
TLD. (The code, although not in ISO 3166-1, had "exceptional reserved status" enabling its use for
specific ISO-approved purposes.) At its September 25, 2000, meeting, the ICANN board passed a
resolution that approved for delegation as ccTLDs those codes from the ISO's exceptional reserved list
for which the reservation permits any application requiring a coded representation of the entity.69 In
March 2005, ICANN authorized the creation of the .eu TLD, which is expected to begin operation in
early 2006.70 It will be open to any person living in the EU, as well as businesses with their headquarters,
central administration, or main base in the EU. 71 The exceptional reserved list is no longer published,
and a policy has been implemented to prohibit the creation or reservation of an unrestricted name that is
not on the ISO 3166-1 list.
The process of deciding who will be delegated responsibility for operating a ccTLD upon its
first entry into the root, or for redelegating responsibility subsequently, can become very complex. This
process is discussed in the next section and in Section 5.5.

The gTLDs are a different matter. They fall into two groups: the first eight, which were selected at
the time of initiation of the DNS--the legacy gTLDs; and the seven that were selected by ICANN in 2000
for addition to the root--the new gTLDs. (An additional 4 to 10 are expected to be added during 2005 as
the result of an ICANN selection process initiated in 2004. See Alternative A under "What Selection
Process Should Be Used" in Section 5.4.2.)
The legacy gTLDs were selected by the developers of the DNS and a group of network and
information center operators. Jon Postel, writing almost 10 years later, still maintained a policy that he
expressed as: "It is extremely unlikely that any other TLDs will be created."72 However, by the following
year, Postel had changed his mind and recommended the creation of new TLDs to compete with NSI,
which had a monopoly in commercial gTLD registrations, and in 1996 he recommended the creation of
150 or 300 new TLDs. At the same time, the rapid growth in size and scope of the Internet, driven by the
introduction of the World Wide Web, created a heavy demand for second-level domain names in the
gTLDs, especially in .com. (See Section 2.5.1.) That, in turn, led to a public demand for additional
gTLDs. In response to that demand (some would argue it was a belated response), ICANN created a
process, which it used during the year 2000 to select the new gTLDs. ICANN treated the addition of
gTLDs as an experiment in order to seek compromises that would satisfy the contending interest groups,
although that did not prevent the additions from becoming controversial. Since that process also entailed
selecting the organizations responsible for the TLDs and the TLD name server operators, its description is
deferred to "Selecting the TLD Registry Operators" below.

68 Letter from Erkki Liikanen, member of the European Commission, to Mike Roberts, (then) CEO and president of
ICANN, dated July 6, 2000. Available at
69 ICANN. 2000. "Preliminary Report. Special Meeting of the Board." September 25. Available at
70 See Ellen Dumout. 2005. "ICANN approves .eu Net domain." March 24. c/net Available at
71 See the October 2004 EU Fact Sheet, "Open for Business in 2005:", available at
<>. There is a more complex
story behind the .eu TLD, but it is beyond the scope of this chapter.
72 Postel, 1994, RFC 1591, p. 1.

Prepublication Copy--Subject to Further Editorial Correction

To at least some degree, TLDs compete with one another for the patronage of those entities that
wish to register domain names, although the maximum prices that ICANN has negotiated with the gTLDs
limit the extent of this competition. This competition might be more intense--both with respect to the
prices charged and the services offered--the larger the number of competitors, although beyond some
point the effect of additional entry is likely to be small and the costs of switching from one domain to
another are likely to give incumbents a strong advantage. In any event, even when firms wish to operate
TLDs, either because they observe incumbents earning large profits or because they believe they can offer
better or cheaper services, entry is constrained by their need to obtain approval from ICANN for inclusion
in the root file.73
Over the past few years, the demand for registrations in the TLDs has gone through several
changes. While the ccTLD registrations have grown continually throughout the period, those in the
previously rapidly growing gTLDs declined in the aftermath of the .com bust and only began to grow
again in mid-2002. According to ICANN writing in mid-2003: "The domain name counts for ccTLDs
have jumped 22% over the numbers used last year [in the budget]. The count for gTLDs, however, has
declined 5%, largely due to a drop-off in .com whose statistics dominate the overall gTLD count because
of its comparative size."74 In fact, the name counts in all three of the largest, .net, and
.org--declined from 2002 to 2003. Both .net and .org declined by 14 percent. However, by
August 2003, another source reported that the total number of .com, .net, and .org domain names
had returned to the high mark of 30.7 million reached in October 2001.75 Sustained growth had returned
in July 2002, and over the next 13 months the total registrations in those three domains grew an average
of 250,000 per month. As further confirmation of the renewed demand for those gTLDS, VeriSign
reported76 that an average of 1.2 million new registrations for domain names ending in .com and .net
were added each month in the third quarter of 2004--a 33 percent increase over the third quarter of

Selecting the Organizations Responsible for the TLDs

ICANN has been delegated authority by the DOC (and subject to the DOC's approval) over
entries in the root zone and, consequently, it can determine which organization is delegated responsibility
for each ccTLD and gTLD. The delegated responsibility entails arranging for the establishment and
operation of (1) name servers for the TLD satisfying Internet technical requirements and (2) a domain
name registration process that meets the needs of the local or international Internet communities.
In the case of ccTLDs, the organizations with delegated responsibility are designated managers or
sponsors, and they are designated the sponsors for sponsored gTLDs. Sponsors are primarily either
government or not-for-profit organizations that provide their own funding, although profit-making
organizations run the commercial gTLDs and some ccTLDs. In both groups of TLDs, the responsible
organization need not operate the required name server and domain registration functions itself and
generally contracts with a specialist organization, which may be a commercial service, to carry out those

73 Some limitations on entry may have a legitimate practical basis. Nor is it clear whether a small number of large
TLDs are to be preferred to a large number of smaller ones. In any event, incumbents can be expected to wish to
limit the competition that they face, whether or not there is a legitimate basis for such limits.
74 Source: ICANN's Budget for Fiscal Year 2003-2004. See
75 Zooknic Internet Intelligence. 2003. "gTLD Domains Returning to 2001 Levels." Press release. August 25.
Available at <>.
76VeriSign. 2004. "Verisign Issues Quarterly Domain Name Industry Brief; Overall Domain Name Registration
Tops Past Record of 66.3 Million." December 1. Available at <
77 See also Linda Rosencrance. 2004. "Domain Name Registrations Hit All-Time High." Computerworld. June 8.
Available at <,10801,93716,00.html>.

Prepublication Copy--Subject to Further Editorial Correction
functions. In the case of the seven unsponsored gTLDs (other than .int), the responsible organization
and the operating organization are the same and had until recently all been commercial services.
However, ICANN designated Public Interest Registry (PIR), a not-for-profit, to manage .org beginning
in 2003. PIR has contracted with a Dublin-based company, Aflilias, to provide the registration services
and Afilias has contracted, in turn, with a commercial provider of DNS services, UltraDNS, to run the
name servers.
For the ccTLDs and the legacy gTLDs, ICANN must have a process for recognizing the
organizations that will be responsible for the TLD when a new ccTLD is added or when a change of
responsibility is desired for whatever reason. The processes used for the ccTLDs and the legacy gTLDs
are different. They are described below. For the new gTLDs, the processes of selecting a manager and
selecting an operator were combined since the prospective manager's choice of operator was one factor in
the manager selection decision. That combined process is described below in "Selecting the TLD
Registry Operators."

IANA began delegating responsibility for ccTLDs shortly after the deployment of the DNS and
most delegations predate the formation of ICANN. ICANN's policy has been not to challenge these
except in cases where a government seeks redelegation.78 In the early days of the DNS, responsibility for
ccTLD management was usually assigned to the Internet pioneers who volunteered for the task.
Generally, there was only one interested organization since, at the time, the commercial prospects were
thought to be minimal. Even so, there was considerable due diligence on many applications to determine
that other possible applicants had been identified and, if there were any, that they supported the active
application. There were, however, some cases of multiple applications and post-delegation challenges by
those who sought various other benefits such as prestige or the ability to leverage the role into sale of
Internet services. The managers agreed to abide by a set of policies for the administration and delegation
of ccTLDs that covered technical requirements and the circumstances under which IANA would make or
change a delegation of responsibility.79 Often the corresponding governments did not know or did not
care about the Internet or the applicable ccTLD.
With the growth in the size and importance of the Internet and the formation of ICANN, that
situation has changed. Governments increasingly want to participate in the selection and oversight of the
manager of their ccTLDs.80 Furthermore, ICANN felt the need for more precisely described agreements
spelling out the mutual obligations and responsibilities among governments, ICANN, and the delegated
managers of ccTLDs. Consequently, the board of ICANN authorized the development of such
agreements with the ccTLDs.
According to ICANN, it was soon realized that no single agreement or, even, structure of
agreement would fit every ccTLD. However, after consideration of the wide variety of specific
circumstances in the 243 ccTLDs, ICANN found that most would fit into one of two general situations
that differ principally in the involvement of the local government or public authority. In the first, legacy
situation, the government is not involved. In the second, triangular situation, it is, thus yielding three
participants: a ccTLD, ICANN, and a local government. ICANN proposed two types of agreements,
corresponding to these two situations.
According to the proposed agreement for the legacy situation, the ccTLD manager would operate
under the oversight of ICANN only, subject of course to the laws of the country. ICANN would have the

78 This section draws extensively on: ICANN. 2001. "Update on ccTLD Agreements." September 20. Available at
<>. See also Jon Postel, 1994, RFC 1591.
79 The general responsibilities of the ccTLD managers are documented in RFC 1591 and in the ICANN publications
ICP-1, Internet Domain Name System Structure and Delegation (ccTLD Administration and Delegation), May 1999,
and ccTLD Constituency Best Practice Guidelines for ccTLD Managers, 4th Draft, March 10, 2001. However,
neither of the ICANN documents appears to have achieved consensus among all the ccTLD managers.
80 See discussion in Section 5.5.2.

Prepublication Copy--Subject to Further Editorial Correction
sole responsibility to ensure that the ccTLD manager operates as a trustee for both the local and the
international Internet communities.
According to the proposed agreement for the triangular situation, ICANN would retain the
responsibility to see that the ccTLD manager meets its responsibilities to the international Internet
community and to any non-national registrants, while the local government would assume responsibility
for ensuring that the interests of the local Internet community are served.
According to ICANN, the decision as to which arrangement to pursue would be reached by the
government and the ccTLD manager (or candidate manager) in consultation with the local Internet
community, with ICANN adopting "a neutral stance."
In the legacy situation, ICANN would have the full authority to select and, when necessary,
change the ccTLD manager. (Although they are not signatories to the agreements, governments are
notified if one is in preparation in case they want to be involved.) In the triangular situation, the relevant
government or public authority would communicate its designation of a ccTLD manager to ICANN,
which would then decide whether or not to accept the designee and, assuming a positive decision, seek to
negotiate an appropriate agreement with the manager.
The proposed agreements with the ccTLDs are intended to cover the following areas: (1) the
delegation of responsibility to the ccTLD and description of the circumstances that would lead to its
termination, (2) specification of the local and global policy responsibilities of the ccTLD, (3)
characterization of the relationship with ICANN and their respective responsibilities, and (4) funding for
Despite ICANN's efforts to get ccTLDs to enter into agreements with it, by December 2004 it
had completed only 12 of them. A number of ccTLDs object to accepting ICANN's formal authority over
their operations. This issue is discussed in detail in Section 5.5.


ICANN has acted to change the organization responsible for several of the eight legacy gTLDs.
The most significant instance was its negotiation with VeriSign Global Registry Services, the legacy
manager of .com, .net, and .org. Because those three gTLDs contain the registrations of the vast
majority of the Internet's gTLD second-level domains and, in particular, almost all of those on its
unsponsored and unrestricted domains, VeriSign's position as the profit-making sole supplier of those
three was felt by many in the Internet community to be detrimental to the long-term health of the Internet.
In May 2001, VeriSign, Inc., ICANN, and the DOC signed a revised agreement in which VeriSign agreed
to give up its operation of .org at the end of 2002 while extending the term of its operation of .net to
the beginning of 2006 and of .com to November 2007. Both of the latter agreements are renewable,
although under different terms: under existing agreements, .net had to be put out for bid by ICANN by
the end of 2005, while VeriSign has presumptive renewal rights for .com, unless it materially breaches
the agreement. ICANN initiated an open bidding process for .net in 2004 and will select an operator in
To select a new manager for .org, ICANN issued an RFP in May 2002. In response, it received
11 proposals from a variety of organizations, both commercial and not-for-profit. According to ICANN,
those proposals were reviewed by three independent evaluation teams that were charged with looking at
technical issues and at the ability of the proposals to meet the specific needs of .org. Comments were
also received from the public and the applicants, which, according to ICANN, were used by the ICANN
staff to prepare an evaluation report and recommendation. The ICANN board accepted the staff
recommendation and ICANN contracted with the PIR, a wholly-owned subsidiary of the Internet Society,
to become the manager of .org, effective January 1, 2003.

81 See ICANN.2004. ".NET Request for Proposals." December 10. Available at <

Prepublication Copy--Subject to Further Editorial Correction
Selecting the TLD Registry Operators

An organization that is responsible for reliably performing the functions of (1) operating the TLD
name servers and (2) registering second-level domains in the TLD is called a registry.82 It may or may
not be the same as the delegated manager for the TLD.
For example, VeriSign is the manager and operates the registry for .com and .net. PIR is the
manager of .org, but, as noted above, it has subcontracted with Afilias to operate the TLD registry,
which has contracted the name server function to UltraDNS. Afilias also operates the registry for .info,
for which it also serves as the manager, and for .vc, the ccTLD for residents of St.Vincent and the
Grenadines, which is managed by the Ministry of Communications and Works of St.Vincent and the
There is always one, and only one, registry for a given TLD, but, as noted above, an organization
can be the registry operator for more than one TLD. While the majority of TLD managers are non-
commercial organizations, some registry operators are commercial organizations that operate for profit;
they register the vast majority of domain names.


The manager of a ccTLD may also be the registry operator, or it may subcontract the registry
services in whole or in part to other organizations. The process for selecting the registry services operator
is entirely up to the ccTLD manager, subject only to whatever restrictions the national government may
impose. The situations differ widely among the 243 ccTLDs.
Since the growth of the World Wide Web has vastly extended the scope, scale, and importance of
the Internet, two phenomena have worked to shape the operational arrangements of ccTLDs in a country.
First, there has been a movement from the early informal arrangements, which often involved voluntary
efforts by the computer science department of a university in the country, to more formal arrangements
that provide legal protection and engage a wider national community in policy-setting roles. Second,
there has been a tendency to contract the actual operations to commercial organizations more willing and
able to undertake the responsibilities than academic institutions.
For example, in Austria, the current manager and registry is Nic.AT, which is a limited-liability
company that since 2000 has been wholly owned by a charitable foundation, the Internet Private
Foundation Austria. Nic.AT was established in 1998 by the Austrian ISP Association to take over
responsibility for the ccTLD from the University of Vienna, which had managed it from its inception but
was faced with an increasing number of registrations, legal questions, and name conflicts beyond its
competence. While Nic.AT handles the name registration, it contracts with the University of Vienna
computer center to run the .at name servers.
In contrast, the new registry and registry operator for the .us TLD is a commercial company,
Neustar, which also is the registry and registry operator for the new gTLD .biz. In the .us case, the
manager, the U.S. DOC, decided to change the operational model from a deeply hierarchical, mostly
geographic, extensively delegated structure to one that would be exploited commercially at the second
level. Unlike the Austrian case, there was no pressure for the change in strategy from the previous
manager or operator, and no significant pressure from users/registrants in the domain--indeed, many of
them argued against it. And the DOC RFP essentially required a commercial operator with commercial

82 Registries can exist at any level, except the root, in the DNS (although in some respects ICANN could be viewed
as the registry for the root). In this section, only TLD registries are considered. The term "registry" is unfortunately
used ambiguously in this context. The database maintained by a registry is also called a registry, as are the
organizations that some registries subcontract with for database maintenance and operation, which are here called
registry operators.

Prepublication Copy--Subject to Further Editorial Correction
Legacy gTLDs
The registries for the legacy gTLDs are the consequence of history. NSI had been operating the
registries for .com, .org, and .net by agreement with the U.S. government when VeriSign
purchased it and assumed the responsibility. As noted above, PIR, upon becoming manager of .org,
contracted with Afilias to run the registry. The organizations shown in Table 3.3 for each of the other
legacy gTLDs are the associated registries (and also sponsors).

New gTLDs

In July 2000, the ICANN board adopted a policy for the introduction of new gTLDs that called for
the solicitation and submission of proposals to sponsor or operate them. In August 2000, the RFP was
published. It specified the contents of the detailed multipart proposal, which was to be accompanied by
extensive supporting documentation and a non-refundable $50,000 application fee. The deadline for
applications was October 2000.
As explained by ICANN, two types of gTLDs were specified: sponsored and unsponsored. In the
latter case, the application was to be submitted directly by the organization proposing to serve as the
registry. In the former case, the application was to be submitted by the sponsoring organization but
would include the proposal of an organization that had agreed to perform the registry functions for the
sponsoring organization. Thus, the registries for the new TLDs were selected through the ICANN
process whether ICANN made the decision directly or accepted the sponsoring organizations' selections.
ICANN announced that the selection criteria would be the following:

The need to maintain the Internet's stability;
The extent to which selection of the proposal would lead to an effective proof of concept
concerning the introduction of top-level domains in the future;
The enhancement of competition for registration services;
The enhancement of the utility of the DNS;
The extent to which the proposal would meet previously unmet types of needs;
The extent to which the proposal would enhance the diversity of the DNS and of registration
services generally;
The evaluation of delegation of policy-formulation functions for special-purpose TLDs to
appropriate organizations;
Appropriate protections of rights of others in connection with the operation of the TLD; and
The completeness of the proposals submitted and the extent to which they demonstrate
realistic business, financial, technical, and operational plans and sound analysis of market needs.

ICANN received 47 applications, of which 2 were returned for non-payment of the fee and 1 was
withdrawn, leaving 44 to be evaluated. The ICANN staff carried out evaluation with the assistance of
outside technical, financial/business, and legal advisors. ICANN's goal was to select a "relatively small
group of applications" that (1) were functionally diverse and (2) satisfied the selection criteria.

At its November 2000 meeting, the ICANN board acted on the staff evaluation and selected the
seven new, .info, .name, .pro, .museum, .aero, and .coop. Four
were, .info, .pro, and .name--and, therefore, amounted to the direct
selection of a registry as well as the TLD. The other three were sponsored, and each included a
designated registry chosen by the sponsor to operate its TLD.
The registries for .com, .net, .org and for the seven new gTLDs have agreed to pay
certain fees and adhere to certain requirements as spelled out in ICANN's sponsored and unsponsored
TLD agreements. (The registries for .edu, .gov, and .mil operate under separate agreements
with agencies of the U.S. government. ICANN is the registry for .int, and the Internet Architecture
Board (IAB) manages .arpa. Therefore, they do not have agreements with ICANN.)

Prepublication Copy--Subject to Further Editorial Correction
ICANN's TLD agreement obligates the sponsor and the registry to (1) satisfy functional and
performance specifications set by ICANN; (2) enter into agreements with any ICANN-accredited registrar
(see "Selecting the Organizations to Register Domains" in Section 3.5.2) desiring such an agreement and
accord the registrar fair treatment; (3) provide query and bulk access to registrant data--Whois
information (see Box 3.6); (4) periodically deposit its registrant data into escrow with an approved escrow
operator; and (5) comply with consensus policies established by ICANN.
The ICANN process for adding gTLDs that was implemented in 2000 was quite controversial.
Many participants and observers complained about the design and implementation of the process. The
issue of whether and how to add new gTLDs is examined in detail in Section 5.4.

BOX 3.6
Whois Service

Whois services provide contact information about the registries, registrars, or registrants in the
DNS. (See Sections 2.3.4 and 2.5.3 for information on the development of the Whois service and
background on the issues surrounding it.)
ICANN-accredited registrars are contractually obligated to collect and provide access to
information about the name being registered, the names and IP addresses of its name servers, the
name of the registrar, the dates of initiation and expiration of the registration, the name and postal
address of the registrant, and the name and postal, telephone, and e-mail addresses of the technical
and administrative contacts for the registered name. These must be made available either directly by
the registrar or, in some cases, by the registry.
There are many separate Whois services on the Internet run by registrars, registries, and other
organizations (often, as with universities, of their own second- or third-level domains). In addition,
there are numerous Web sites that provide links to many of the Whois sites, such as

Operating the TLD Registries

Every TLD registry operator must perform two basic functions: register domain names requested
by registrants and operate the name servers that will link those domain names with their IP addresses and
other critical information. Even these basic responsibilities may be divided between organizations, some
commercial and some non-commercial, as noted above in the cases of Austria and of .org. In contrast,
a single commercial organization, VeriSign, is the manager, runs the registry, and runs the name servers
for .com and .net.
The registration operation produces the entries to the zone file for that domain, the content of the
Whois file, and records of billing and payment, where appropriate. When the TLD has restrictions on who
may register either in the domain (such as restricting registrations to nationals or residents of a country or
to professionals or museums) or in its generic subdomains (such as Australian businesses in or
British limited liability corporations in, each application for a domain name must be examined
to ensure that those restrictions are satisfied.
The ICANN agreements with 12 gTLDs include functional and operational specifications that the
registry operators are responsible for meeting, while the few agreements with ccTLDs specify only
general requirements for Internet connectivity, operational capability, and adherence to key RFCs, as well
as agreements to make financial contributions to ICANN. Ideally, all operators of TLD name servers
should satisfy certain minimal technical conditions to ensure their compatibility with the Internet and that

Prepublication Copy--Subject to Further Editorial Correction
they are configured so as not to pose a danger to the stability of the Internet, although there is no
mechanism for enforcing this for the TLDs not covered by ICANN agreements.

3.4.4 Assessment

Conclusion: The level of technical capability and competence varies widely across the 258 TLD
registries. The gTLDs and the large ccTLDs generally operate at a high level of availability and
responsiveness. Although there are no readily available measures of the performance of the majority of
ccTLDs, they appear to provide adequate service.

Recommendation: Regular and systematic testing of the availability and operation of secondary
servers should be adopted by top-level domain registry operators. Policies and procedures should be
developed to clarify what to do when problems are identified and what measures can be taken when
problems are not resolved within a reasonable period of time.

Conclusion: No single organization has the authority and the ability to oversee the operation of
all the TLDs. ICANN's formal authority extends only as far as the provisions of its agreements with 10
gTLDs and 12 or so ccTLDs and the authority it has to recommend changes in the root zone to
accommodate new or reassigned TLDs.

Even where there is a contract, ICANN's authority has been tested. VeriSign's introduction in
October 2003 of its Site Finder service raised fundamental issues of both a technical and an institutional
nature and has been challenged in the courts.83


The domain names in the DNS hierarchy that Internet users interact with most directly are those
at the second level (or in those ccTLDs with fixed second-level domains, such as, those at the
third level). They are generally the key identifiers in e-mail addresses (such as or a Web address (such as They are the names that
businesses, individuals, government agencies, non-profit organizations, and various other groups acquire
to identify themselves on the Internet--the names on their signposts in cyberspace. Over 70 million
second-level (and, in some instances, third-level) domain names had been registered by early 2005, with
about two-thirds in the gTLDs (more than half in the two largest and .net) and about one-
third in the ccTLDs.84

3.5.1 Technical System of the Second- and Third-Level Domains

Second- and third-level domains may have their own name servers to respond to queries to their
zone files, as most large organizations do, but often the services are provided by ISPs or other Web site
hosting organizations that store the zone file on their name servers. This is the course taken by most small
organizations and individuals, although there are many exceptions.
The zone file of a second- or third-level domain may be very small if it belongs, for example, to
an individual, or it may be quite large, if it is owned by a commercial or governmental organization. In
the latter case, a great many of the entries may be associated with the e-mail addresses of the thousands of

83 See Chapter 4 for a discussion of the Site Finder case.
84 See Tables 3.2 and 3.3.

Prepublication Copy--Subject to Further Editorial Correction
employees of the institution, while several, tens, or hundreds may be associated with the name servers of
lower-level zones. Often, institutions will own multiple domain names (e.g., and that point to the IP address of the same server, enabling access to it under
different domain names.

3.5.2 Institutional Framework of the Second- and Third-Level Domains

The effective operation of the second- and third-level domains requires that three functions be

1. Selecting the organizations to register second-level domains,
2. Registering second-level domains, and
3. Resolving second-level domain name conflicts.

The principal organizations participating in this institutional framework are ICANN, the TLD registries--
the organizations that actually register most second-level domains--the registrars, and the organizations
that provide dispute resolution services. Although many of the organizations at this level are commercial,
numerous not-for-profit and governmental organizations play an active role as well.

Selecting the Organizations to Register Domains

In many cases, especially in the ccTLDs and some of the gTLDs, only the registry carries out the
registration of second- or third-level domains. However, in a 1998 white paper85 the DOC, responding to
a policy goal of privatizing and increasing competition in the market for domain name registration,
recommended opening the business of registering lower-level names in gTLDs to competition. It
subsequently amended its agreement with NSI, then the operator of the .com, .org, and .net registries,
to require it to develop a system of multiple registrars and put it into operation in 1999. The DOC
designated ICANN as the organization that would oversee the establishment of the Shared Registration
System (SRS) and would be responsible for establishing and implementing a system for registrar
accreditation. The SRS and registrar accreditation began operation in 1999. Before the multiple registrar
system, NSI charged $35 per year for registrations in its domains. At the beginning of 2005, registrations
in those domains in the United States could be obtained for less than $10 per year.86
Under the terms of their agreements with ICANN, gTLD registries are required to permit
registrars to provide Internet domain name registration services within their top-level domains. In
addition, these agreements regulate the price that registries can charge registrars--the "wholesale price,"
with the current regulated price being a maximum of $6 per year per registrant. One can think of the
regulation of the wholesale price as intended to constrain the exercise of market power by registries and
the requirement for competing registrars as intended to constrain the retail "margin."

To the extent that regulation of the wholesale price is intended to limit the exercise of market
power in the wholesale market, however, it is not entirely obvious why the wholesale price that can be
charged by new gTLDs, especially new gTLDs that are intended to serve diverse users, must be
regulated.87 Indeed, increased future competition, especially as the number of gTLDs is expanded, might
reduce or eliminate the need to regulate the wholesale price that VeriSign or other registries can charge.
But if regulation of the wholesale price is intended to prevent registries from exploiting existing
customers that may be locked in, there may be a continuing need to regulate wholesale rates even if the

85 Department of Commerce. 1998. "Management of Internet Names and Addresses." Federal Register. Vol. 63. No.
111. June. p. 31741.
86 For example, in February 2005, was offering .com registrations for $8.95.
87 Some specialized TLDs might continue to have market power even if the total number of TLDs were very large.

Prepublication Copy--Subject to Further Editorial Correction
registry market were to become more competitive. The significance of lock-in will depend on the
importance of switching costs, the flow of new registrants relative to the existing stock, and on whether
registries can discriminate between new and existing registrants.
There were in February 2005 more than 460 registrars from more than 20 countries accredited to
register domain names in 1 or more of the 10 eligible gTLD domains.88 Many of them have decided to
operate, at least in part, as wholesalers and suppliers of registrar services; those operations have enabled
many agents to sell domain names without any relationship with, or accreditation by, ICANN. Although
most ccTLDs use authorized agents, many ccTLDs have adopted the notion of multiple registrars, and
many ICANN-accredited registrars also register ccTLD second-level domain names.
ICANN accredits registrars through an open application process. Any organization wishing to
become an ICANN-accredited registrar must complete a detailed application concerning its technical and
business qualifications and pay a $2500 application fee. If approved by ICANN, the applicant must
execute the standard Registrar Accreditation Agreement89 and pay a yearly accreditation fee that is $4000
for the first TLD and $500 for each additional TLD for which the registrar is accredited. In addition, the
registrar must pay a quarterly accreditation fee to cover a portion of ICANN's operating expenses. The
fee is based, in part, on the registrar's share of registrations in the TLDs for which it is accredited.
Registrars that are accredited by ICANN must also enter into accreditation agreements with the registries
in the TLDs in which they want to register domains. Those agreements specify, among other things, the
fees to be paid to the registries for each registration. As noted above, a ceiling is imposed on that fee
structure by ICANN for the gTLDs that have signed agreements with ICANN.
The ICANN Registrar Accreditation Agreement imposes certain requirements on registrars. The
registrar is obligated to (1) submit specified information90 for each registrant to the registry; (2) enable
public Internet access to a file of information about registrants--the Whois file--both in query and bulk
access form; (3) maintain a file of all registrant information submitted to the registry; (4) regularly submit
a copy of the file to ICANN or to an escrow agent; (5) comply with consensus policies established by
ICANN; and (6) have for the resolution of name disputes a policy and procedures that comply with
ICANN's Uniform Dispute Resolution Policy. (See "Resolving Domain Name Conflicts" below.)
However, some of the newer gTLD agreements anticipate the possibility of "thick" registries, for
example, ones in which Whois and similar data are maintained by the registry, not the registrars. This
changes requirement (2) and has some impact on the others.

Registering Domain Names

Registration of second-level (or third-level) domain names occurs according to different
processes in the different types of TLDs. For the 10 gTLDs that use accredited registrars and for a
number of ccTLDs--for example, Great Britain, Australia, Canada, and Denmark--registrars compete to
sell second-level domain names in the TLDs they represent. They are free to set whatever fees they like,
subject only to competitive market forces and their obligations to pay registration fees to the registries.
For the .edu, .gov, .mil, .int, and .arpa gTLDs, as well as for most ccTLDs, registration
by members of the restricted group occurs directly at the registries.
The registrar stage of the DNS process appears to be quite competitive, with entry being
relatively easy91 and competition taking place along a number of dimensions. Registrars for the same

88 For the current list, see the ICANN site at <>.
89 A copy of the agreement can be found at
90 See Box 3.6.
91 This competition is facilitated by the Shared Registration System protocol, which allows registrars to enter names
directly in registries.

Prepublication Copy--Subject to Further Editorial Correction
registry compete with one another for the patronage of registrants92--what is sometimes called intra-
brand competition--with competition being based on both the prices and the services offered. These
services include the efficiency with which registrations take place as well as value-added services that
may be bundled with registration.93 Switching registrars within a given registry is not particularly
difficult, but there have been complaints that registrars have not always responded promptly to requests
for switching and that some registrars have aggressively and misleadingly solicited other registrars'
customers. In addition, domain name theft has been one of the problems associated with inadequate
procedures and security measures put in place by the registrars of domain names. In such cases, a third
party fraudulently claims to be the registrant in order to have the domain name transferred to its
ownership.94 There have also been instances of fraud charges against domain name registries and
registrars of domain names.
State of the Domain report listed 148 registrars in the CNO domains (.com,
.net, and .org) in the first quarter of 2003.95 However, "market shares" in these domains were
skewed, with the VeriSign registrar (now Network Solutions)96 having more than 25 percent of all
registrations in .com--which was, however, a significant decline from its 40 percent share only 15
months earlier;97 Tucows having about 10 percent; having approximately 9 percent; and
GoDaddy having somewhat over 6 percent. Thus, the share of the "market" controlled by the top four
firms in .com--the four-firm concentration ratio--was about 52 percent.98 Seven other registrars each
had more than 1 percent of all .com registrations. (More recent data were not available at the time of this

The situation was somewhat different in the then newly opened .biz domain according to the
SnapNames State of the Domain, first quarter 2003 report. Although the VeriSign registrar (Network
Solutions) was the market leader, its share was only about 19 percent, 6 percent less than its share in the
.com domain. Other registrars with shares in excess of 6 percent were (10 percent),
Tucows (9 percent), and Melbourne IT (6 percent),99 so that the four-firm concentration ratio here was
only about 44 percent. SnapNames reports that there were 127 registrars of .biz names. As a result of
the wide disparity in the sizes of the various domains, Network Solution's approximately 25 percent share
of .com was about 5.8 million names, while its approximately 19 percent share of .biz was only about
170,000 names.100

92 This is not to say that registrars for different TLDs do not also compete with one another, a form of interbrand
competition. In addition to registering new domain names, registrars also participate in the secondary market for
domain names, acting as brokers, as well as in assisting registrants in applying for expired names.
93 SnapNames reported that GoDaddy, a registrar, gained 70,000 registrants in a month when it launched its free
online tax preparation software, presumably available only to its registrants. See the next section for a discussion of
value-added services, some of which are provided by firms that are not registrars.
94 In April 2004, the original operator of received a $15 million settlement from VeriSign because its
registrar service had incorrectly transferred ownership of the name to a fraudulent claimant.
95, Inc. 2003. State of the Domain, First Quarter 2003. May 13. Available at
<>. The SnapNames report showed over 150 ICANN-
accredited registrars operational at that time (pp. 31-34).
96 VeriSign sold its registrar, Network Solutions, to a private investment company for about $100 million in October
97, Inc. 2002. State of the Domain, January 2002. February 26.
98 The quotation marks around "market" and "market shares" are intended to indicate that no claim is being made
that registrar services in the .com domain constitute a relevant antitrust market. SnapNames reported that the top 10
registrars had about 75 percent of the market in the first quarter of 2003, down from about 91 percent at the end of
99 Recall that Melbourne IT is part of a joint venture with NeuStar, the operator of the .biz registry.
100 For further discussion of market issues and presentation of market data, see "Generic Top Level Domain Names:
Market Development and Allocation Issues," Organisation for Economic Co-operation and Development,
Directorate for Science, Technology and Industry, Committee for Information, Computer and Communications

Prepublication Copy--Subject to Further Editorial Correction
Finally, although registrars charge the same prices to all registrants, some registrars offer a back-
order service under which, for a fee, they will track the expiration of a desired domain name and attempt
to register it immediately if the registration lapses. However, multiple registrars may be trying
electronically to register the same name at the instant it becomes available. The one that tries at just the
right instant wins the prize. Because of this chance element, consumers often enter back orders with
multiple registrars. In March 2002, ICANN asked VeriSign to conduct a 12-month trial of a single wait-
listing service, which would take just one order--at a $24 fee directly from the consumer--for an
expiring name on a first-come, first-served basis. Thus, from the consumer's side, the need to use
multiple registrars would be eliminated, but from the registrar's side, the opportunity to derive additional
revenue would be lost. On July 15, 2003, a coalition of name registrars filed a lawsuit against ICANN
seeking to block the launch of VeriSign's global waiting list for domain names.101 At its March 2004
meeting, ICANN's board voted to seek the DOC's agreement to its approval of VeriSign's 1- year trial of
the wait-listing service.

Resolving Domain Name Conflicts

One of the most difficult institutional roles that the operation of the DNS requires is the resolution
of conflicts among competing claimants for domain names. These conflicts arise for a number of reasons
that are discussed in detail in Section 2.5. The one that has attracted the most attention is the use of
trademarked words in domain names, which is covered in "Trademark Conflicts" in Section 2.5.2.
Although domain names can be used for a number of legitimate purposes other than as an address for a
World Wide Web site, such as identifying a host, an e-mail server, an FTP site, and so on, the vast
majority of the disputes involving domain names are associated with their very visible use in association
with World Wide Web sites. Conflicts can also arise over names in which individuals, organizations, or
governments claim a proprietary interest other than a trademark.
Whatever the source, the practical use of the DNS, which assumes that every domain name will
be registered to one and only one entity, cannot proceed without some means for resolving conflicting
claims for the same name. In the early days of the Internet, before strong financial and political interests
were involved, such conflicts were handled informally, usually on a first-come, first-served basis, and
they still are in many ccTLDs.102 However, as domain names appeared on the signposts on the World
Wide Web and in e-mail addresses, and some gained visibility and potentially great value, the need to use
more formal processes became evident. This was especially true for .com at first, and then for .net and
.org as they were more widely marketed to general users. For the reasons described in Section 2.5.1,
they were the most visible names on signposts on the Web.

Possible Remedies to Conflicts Over Names
There are basically two approaches to resolving conflicts over rights to names: one approach
incorporates policies and regulations into the actual administration of the naming system and its
assignment rules. The other approach relies on dispute resolution mechanisms that are external to naming
system administration.

Internal Remedies.
In a completely internalized naming system administration, a single manager owns
the naming system and decides who is entitled to which name. Hence there are no rights conflicts. Public
or quasi-public naming systems, such as the DNS, can also attempt to link the assignment of names to

Policy, Working Party on Telecommunication and Information Service Policies. July 13, 2004. Available at
101 See Susan Kuchinskas. 2003. Embittered Registrars Sue Embattled ICANN. July 15. Available at
102 See, for example, the terms and conditions for the .uk registry available at <>.

Prepublication Copy--Subject to Further Editorial Correction
strict policies and regulations. The available techniques include imposing policies on the assignment of
names at the point of registration, name reservations103 or exclusion, and so-called sunrise proposals.

Policies Imposed at the Point of Registration. Such policies must rely on rules or procedures to
determine eligibility for a name. It is difficult to mechanize such rules. Thus, prior review of registration
is likely to be expensive and slow if it is administered manually and prone to be crude and unfair if it is

Name Exclusions. Name exclusions withdraw specific names or entire classes of names from the
available database. For a time, Network Solutions did not allow registration of six of the Federal
Communication Commission's "seven dirty words" in the domain name space.104 The World Intellectual
Property Organization (WIPO) recommended eliminating a list of "famous" trademarks from the DNS
database and reserving their use to the trademark holder.105 ICANN has provided all of its newly
authorized registries with a list of reserved names that consisted of names and acronyms of organizations
related to the IETF and ICANN.106
Name exclusion is an effective method of protecting the reserved names from abuse, but it is also
a crude instrument, and in a global, public name space its crudeness raises significant public policy
concerns. Withdrawing words from circulation is in effect a form of automated censorship. Exclusions
do not make any distinction between legitimate and illegitimate users; they simply make it impossible to
use the names. A rigid exclusion deprives these organizations of the right to register domain names
corresponding to their acronyms or trademarks. Exclusions and other regulations are less significant
when they are part of the practice of a private naming system, where the owner can be assumed to have
property rights over the naming system as a whole. Such exclusions also ignore the fact that certain
words have a particular meaning only in a particular language. A "dirty" word in English may have no
such meaning in another language.

Sunrise Proposals. Sunrise proposals, in general, establish a period at the start-up of a new name
space within the DNS during which certain entities may register names in which they have established
rights (e.g., famous trademarks) before the space is opened for public registration.

External Remedies.
External methods of resolving rights conflicts include litigation through the courts or
alternative dispute resolution procedures, such as the ICANN Uniform Domain Name Dispute Resolution
Policy (UDRP), which is discussed below. The advantage of these methods is that they are based on
case-specific analysis. Thus, they are sensitive to the specific facts of the conflict and can employ "soft"
interpretive principles to adjudicate a dispute. Their disadvantage, of course, is that they are more time-
consuming and expensive, and litigation is subject to jurisdictional limitations that may not match the
scope of the affected naming system. Litigation is particularly expensive and cumbersome, although it
offers a reliably neutral tribunal in many countries. Alternative dispute resolution techniques such as the
UDRP greatly reduce the cost of external dispute resolution but sacrifice thoroughness in compiling and
verifying facts, and their rapid procedures can put respondents at a disadvantage.

103 The deployment of internationalized domain names involves new processes and challenges with respect to the
reservation of names to prevent some conflicts over domain names. See Section 5.6.3 for discussion.
104 For example, see "Seven Dirty Words," Wikipedia online encyclopedia. Available at
105 See "The Problem of Notoriety: Famous and Well-Known Marks," in The Mangement of Internet Names and
Addresses: Intellectual Property Issues,
First Report of WIPO Internet Domain Name Process, April 30, 1999.
Available at <>.
106 For the list of reserved names see, ICANN, Appendix K, Schedule of Reserved Names. Available at

Prepublication Copy--Subject to Further Editorial Correction
Remedies to Conflicts Over Names in the DNS
Since the remedies to domain name conflicts differ somewhat between the gTLDs and the
ccTLDs, each is described separately below.

Although there are other uses of trademarks with international spillover effects, second-level
domains in gTLDs are unusual in the extent to which they are visible in almost every jurisdiction in the
world, with the resulting difficulty in making either geographic or sectoral distinctions. Ordinarily, the
same trademarks can be used in a single geographic region so long as they are in different economic
sectors where confusion is unlikely. On the Internet such distinctions cannot be made. As a result, the
question arises as to the means to be used to make decisions about domain name conflicts and disputes
that cross national and business sector boundaries and, since such decisions inevitably involve matters of
commercial or political or social importance, to ensure that those decisions are regarded as legitimate and
enforced. Normally, trademark and related disputes are resolved by the courts of the nation in which they
arise, and those in many nations have shown themselves capable of handling domain name disputes. In
addition, the United States has passed specific legislation at both the federal and the state levels
addressing the rights of trademark owners to domain names. (These are discussed further below.) As
noted above, however, legal proceedings can become expensive and time-consuming. Therefore, many in
the Internet community felt the need for a less expensive and quicker means of resolving domain name

Uniform Domain Name Dispute Resolution Policy (UDRP). An answer, implemented by
ICANN in December 1999, based on a recommendation from WIPO, was the Uniform Domain Name
Dispute Resolution Policy (UDRP). The UDRP has been adopted, as ICANN requires, by all registrars in
the .aero, .biz, .com, .coop, .info, .museum, .name, .net, and .org top-level
domains, as well as voluntarily by managers of several global ccTLDs, such as .tv, .cc, and .ws.
In addition, managers of other ccTLDs, such as .ca, have adopted their own policies based on modified
versions of the UDRP.
The policy takes effect through agreements between registrars (or other registration authorities)
and their registrants. Each registrant agrees to be bound by the provisions of the policy when it registers
its domain name. In agreeing to the registration agreement, the registrant also must represent that to its
knowledge its registration of the domain name does not infringe any third party's rights, nor is it for an
unlawful purpose, nor will it be used in violation of any applicable laws or regulations.
By registering the domain name, the registrant is then bound by the policy to submit to a
mandatory administrative proceeding if a complainant asserts that:

registrant's domain name is identical or confusingly similar to a trademark or service
mark in which the complainant has rights; and
registrant has no rights or legitimate interests in respect of the domain name; and
registrant's domain name has been registered and is being used in bad faith.

The complainant must demonstrate all three of these conditions for the complainant to prevail.

The policy asserts examples of actions that would demonstrate bad faith that include registering
and using the domain name:

for the purpose of transferring the registration to the complainant, or to one of its
competitors, for more than the documented out-of-pocket costs of the domain name; or
to prevent the owner of the trademark or service mark from using the mark in a domain
name (provided that registrant engaged in a pattern of such conduct); or
primarily for the purpose of disrupting the business of a competitor; or

Prepublication Copy--Subject to Further Editorial Correction
intentionally to attract, for commercial gain, Internet users to registrant's Web site or
other on-line location, by creating a likelihood of confusion with the complainant's mark
on registrant's Web site or location.

It also describes circumstances that would enable the registrant to demonstrate its rights and
legitimate interests in the domain name. These are as follows:

before any notice to registrant of the dispute, its use of, or demonstrable preparations to
use, the domain name or a name corresponding to the domain name in connection with a
bona fide offering of goods or services; or
registrant has been commonly known by the domain name, even if it has acquired no
trademark or service mark rights; or
registrant is making a legitimate noncommercial or fair use of the domain name, without
intent for commercial gain, to misleadingly divert consumers, or to tarnish the trademark
or service mark.

The mandatory proceeding--which is electronically based--must be held before an accredited
and administrative-dispute-resolution provider that has been approved by ICANN.107 The complainant
selects the provider and is required to pay the fees, except in the case when the registrant elects to expand
the panel from one to three panelists, in which case the fee is split. The registrar, the registry, and
ICANN are not parties to a UDRP proceeding. However, during a UDRP proceeding, the registrar does
confirm that the domain name has been registered by the respondent named in the proceeding and is
required to execute the outcome of a decision. The only remedy available to the complainant through the
proceeding is cancellation or transfer of the domain name.
As of May 10, 2004, 9377 proceedings involving15,710 domain names had been brought under
the UDRP.108 Two-thirds (6262) of these proceedings had resulted in a transfer of the disputed domain
name to the complainant or in a cancellation of the domain name. Approximately one-fifth (1892) had
resulted in a decision for the respondent, and approximately one-tenth (971) of the proceedings had been
disposed without decision or terminated. There were 931 proceedings pending. The 15,710 domain
names that had been disputed in 4 years represent 0.03 percent of the more than 46 million domain names
registered in the gTLDs subject to the UDRP.
Approximately 60 percent of these proceedings have been filed with WIPO, approximately 33
percent have been filed with the National Arbitration Forum (NAF), approximately 6 percent were filed
with eResolution,109 and approximately 0.7 percent have been filed with the Center for Public Resources
Institute for Dispute Resolution (CPR). The Asian Domain Name Dispute Resolution Centre (ADNDRC)
began operation in February 2002.
Over time, there has been a decline in the number of UDRP proceedings filed. As shown in
Figure 3.3, the number of UDRP proceedings filed each month has been steadily decreasing since August
WIPO explained this decline as suggesting that "an expedited on-line dispute resolution service
has been effective in dissuading Internet pirates from hijacking names."110 It may also be partly the
consequence of working through the backlog of cases that UDRP faced upon start-up and entering a more
normal steady-state condition. To some extent, as well, it may reflect a change in the attitudes of

107 The approved providers are listed at <>. There were
four approved providers in February 2005.
108 ICANN. 2004. "Statistical Summary of Proceedings Under Uniform Domain Name Dispute Resolution Policy."
January 30. Latest version available at <>.
109 eResolution ceased operating as a dispute-resolution provider for ICANN in late November 2001.
110 WIPO. "WIPO Continues Efforts to Curb Cybersquatting." Press Release PR/2002/303. February 26. Available
at <>.

Prepublication Copy--Subject to Further Editorial Correction
trademark owners, some of whom may have come to feel that their interests are not jeopardized by every
similar domain name, which are not worth the cost and effort to maintain. The decline in cases might also
be attributed, in part, to learning by cybersquatters about avoiding a finding of bad faith against them.
As a first step in a policy development process, in 2003, ICANN prepared a report that lists many
of the procedural and substantive issues that have been raised about the UDRP.111

Jun- Se

FIGURE 3.3 UDRP fillings as of April 30, 2004.

NOTE: With respect to gTLD UDRP filings commencing during the period from December 1999 through October
2003, data were obtained from the ICANN Web site at <>. With
respect to such filings commencing during the period from November 2003 through April 2004, data were obtained
directly from the Asian Domain Name Dispute Resolution Centre Web sites at
<> and <>, the Center for
Public Resources Institute for Dispute Resolution Web site at <>, the
National Arbitration Forum Web site at <>, and the World
Intellectual Property Organization Arbitration and Mediation Center Web site at
<>. Specialized domain name dispute resolution proceedings
(e.g., Start-up Trademark Opposition Policy (STOP) and the usTLD Domain Name Dispute Resolution Policy
(USDRP)) have been excluded from these statistics.

Start-up Registrations. The UDRP was designed to handle an ongoing rate of domain name
conflicts arising in already established gTLDs. However, several of the seven new gTLDs have faced
unique issues when they entered the start-up phase. Specifically, under the terms of their contracts with
ICANN, they needed a process by which intellectual property holders could challenge bad-faith claimants

111 See ICANN. 2003. "Staff Manager's Issues Report on UDRP Review. August 1. Available at
<>. ICANN also set up a committee to
consider WIPO's proposed amendments to the UDRP. ICANN has not made the committee's report public.

Prepublication Copy--Subject to Further Editorial Correction
in advance of open registration in order to avoid a rush of cybersquatters followed by a heavy demand on
the UDRP process as intellectual property holders asserted their rights. The four most significant--
.biz, .info, .name, and .pro--each adopted somewhat different policies, but all are adopting
or have a UDRP-like dispute resolution policy.

Most ccTLD registries, and their agents and registrars when they exist, publish policies about
who is eligible to register in the second-level domain, or in its third-level domains when they are open for
direct registration. These policies generally also cover the resolution of conflicts over domain names.
Where the TLD limits itself to individuals and organizations that have an association with the country,
many potential conflicts are readily addressed through national administrative, regulatory, and judicial
institutions. For example, matters of trademark priority are handled through the national regulatory and
legal systems, and corporations may have defensible rights only in the names that are legally registered.

U.S. Legislation. The United States has passed legislation at both the federal and the state levels
addressing the rights of trademark owners to domain names.
Federal--The ACPA: On November 29, 1999, President Clinton signed into law the Anti-
cybersquatting Consumer Protection Act (ACPA), which provided a trademark owner with a further cause
of action that was specifically directed to domain names.112 Under the ACPA, a trademark owner can
bring a civil action against a person if that person has a bad-faith intent to profit from a mark and
registers, traffics in, or uses a domain name that, in the case of a mark that is distinctive, is identical or
confusingly similar to that mark; or in the case of a famous mark, is identical or confusingly similar to or
dilutive of that mark; or is a trademark, word, or name protected by law.113 Mere registration of such a
domain in bad faith may be sufficient to violate the trademark owner's rights under the ACPA; there is no
further requirement for any use of the domain name in association with any goods or services.
Under the ACPA, factors affecting the judgment of bad faith include, but are not limited to,
whether (1) the registrant holds any intellectual property rights in the name; (2) the domain name consists
of the registrant's name; (3) the domain name was used in connection with the bona fide offering of any
goods or services; (4) the domain name was used in a way that could be viewed as a bona fide
noncommercial or fair use; (5) the domain name was intended to divert consumers from the mark owner's
online location; (6) the registrant offered to sell the domain name to the mark owner without having used
it in a bona fide manner; (7) the registrant provided false contact information in the registration form; (8)
the registrant acquired multiple domain names that are identical or similar to the trademarks of others; and
(9) the domain name incorporates a mark that is not distinctive and famous.
The ACPA also created a cause of action for individuals with respect to domain names.
Specifically, a domain name registrant can be held liable if the domain name consists of, or is
substantially and confusingly similar to, the name of another living person and the domain name has been
registered without that person's consent with the specific intent to profit from the name by selling it for
financial gain. This particular cause of action, however, is not available for domain name registrations
that occurred prior to the ACPA's enactment date.
With respect to remedies, the ACPA provided that a court may award injunctive relief, including
forfeiture, cancellation, or transfer of the domain name. In addition, if the registration, trafficking, or use
of the domain name occurred after the ACPA's enactment date, a plaintiff can elect to recover, instead of
actual damages and profits, an award of statutory damages in the amount of $1,000 to $100,000 per
domain name (as the court considers just).

Furthermore, if personal jurisdiction over the domain name registrant cannot be obtained, the
trademark owner could file an in rem civil action114 against the domain name itself in the judicial district

112 15 U.S.C. § 1125(d).
113 18 U.S.C. § 706 ("Red Cross") or 36 U.S.C. § 220506 ("Olympic").
114 An in rem action is taken against property directly, in contrast to an action against people (e.g., the owners of a
given piece of property).

Prepublication Copy--Subject to Further Editorial Correction
of the domain name registrar, domain name registry, or other domain name authority that registered or
assigned the domain name. However, the ACPA limits the remedies in an in rem action to forfeiture,
cancellation, or transfer of the domain name.
States: California, Hawaii, and Louisiana also passed laws that address the registration, sale, and
use of domain names within that state and provide for civil remedies in state courts for violations of these

Foreign Legislation. Relatively few countries have chosen to address the rights of trademark
holders in domain names in the same legislative manner as the United States. The European Union (EU)
has relied largely on new telecommunications laws coordinated at the EU level to provide for the
regulation of domain names at the national level. For example, in Spain, the General
Telecommunications Act (1998) was modified in July 2001 to impose a number of conditions on the
registration of domain names under the ccTLD .es, including a requirement that a domain name to be
registered be somehow related to the trademark or name of the company undertaking the registration. In
other countries, the judicial process, rather than the legislative process, has been relied on to address
conflicts between domain names and trademarks. Since at least 1997, courts within the United Kingdom
have been prohibiting the registration of domain names that conflict with trademarks. In 1998, the Delhi
High Court in India likewise extended its form of common law trademark protection to domain names, as
did the Tribunal de Grande Instance of Draguignan in France. Many ccTLDs, 42 in all, including a
number from the EU, have opted to rely on a form of alternative dispute resolution policy.115 Likewise,
the new .EU ccTLD will apply the following rules regarding registrations:

Governments may reserve geographical and geopolitical names.

In a 4-month sunrise phase, "prior rights" holders and public bodies can register .eu domain
names before the general public can do so.

Two months before the sunrise phase starts, technical and administrative measures will be
published in detail.

In the first 2 months of the sunrise phase, registered national and European Community
trademarks and geographical indications as well as names and acronyms of public bodies can
be registered as .eu domain names by the holder/public body.

Two months later, other "prior rights" holders can also register .eu domain names, but only
as far as they are protected under national law in the member state where they are held. This
provision concerns unregistered trademarks, trade names, business identifiers, company
names, family names, and distinctive titles of protected literary and artistic works.

There will be an alternative dispute resolution procedure in place (similar to ICANN's

After the sunrise phase, the domain names will be registered according to the first-come, first

3.5.3 Assessment

Conclusion: The 72 million registered second- and third-level domains are operated by
individuals with a broad spectrum of capabilities. It is notable that the DNS has been able to function
effectively and reliably despite this range of operator capabilities.

Conclusion: The UDRP is a unique cross-border, electronically based process that has resolved
thousands of disputes over domain names without the expense and potential delay of court proceedings.

115 See <> for an example.

Prepublication Copy--Subject to Further Editorial Correction

The issues of dispute resolution and appropriate Whois balance are examined in Chapter 5, where
the alternative approaches are described and the committee's recommendations presented.


Conclusion: The Domain Name technical system reliably and effectively handles the billions of
queries it receives every day. The institutions that manage it perform the required functions adequately,
in many cases without direct compensation.

The DNS technical system can continue to meet the needs of an expanding Internet.
Early in the committee's assessment it became apparent that it would not be fruitful to consider alternate
naming systems. As noted, the DNS operates quite well for its intended purpose and has demonstrated its
ability to scale with the growth of the Internet and to operate robustly in an open environment. Moreover,
significantly increased functionality can be achieved though applications--such as navigation systems--
built on the DNS, or offered independently, rather than through changing the DNS directly. Hence, the
need did not seem to be to replace the DNS but rather to maintain and incrementally improve it.
Furthermore, given the rapidly increasing installed base and the corresponding heavy investments in the
technical system and the institutional framework, the financial cost and operational disruption of changing
to a replacement for the DNS would be extremely high, if even possible at all.

Yet, despite this better than passing grade, the committee's assessments have identified a number
of significant technical and institutional issues whose effective resolution is critical to the DNS's
successful adaptation to the demands on it. Chapters 4 and 5 address those issues.


Prepublication Copy--Subject to Further Editorial Correction
The Domain Name System: Technology Prospects


The Domain Name System, as described in Chapter 3, has met most of the naming needs of the
Internet and the applications that rely on it, even as their uses and usage have expanded rapidly.
However, the broadening and deepening penetration of the Internet and its applications into global
communications, commerce, and culture pose new challenges to the basic technology of the DNS. In
anticipation of and in response to those challenges, the technology community has been developing
modifications of and extensions to the current technology.
This chapter is a review of the challenges and the prospective or actual technology responses to
them. Each challenge and responsive technology is described and evaluated and the implications for the
Internet and its applications are explained. Where the committee is in agreement, its conclusions and
recommendations are presented. In all cases, the goal is to provide a clear description of the challenges,
the technologies, and their prospects in order to inform forthcoming policy deliberations affecting or
affected them.

The following challenges and responsive technologies are addressed in this chapter:

1. Improving the security of the DNS DNSSEC,
2. Linking the telephone and Internet naming systems ENUM,
3. Internationalizing domain names IDN, and
4. Responding to domain name errors aggregation and Site Finder.

Some of these technologies are in or ready for the first stages of implementation, whereas others may
never enter into wide-scale usage. Nevertheless, a basic understanding of each of them will enable wiser
decisions about them and other innovations in the future.


Because of its central role in the operation of the Internet, the DNS is a natural target for
mischievous and malicious attacks. These can take a wide variety of forms depending upon the ingenuity
of the attacker and on which of the potential vulnerabilities is attacked.1 The most severe recent attack
was the denial-of-service attack launched in October 2002 and described in Section 3.3.4. It swamped 8
of the 13 root name servers for up to an hour and a half. However, the remaining 5 servers handled the
regular requests to the root without difficulty. Since that attack, the root name server operators have taken
a number of steps, including the widespread distribution of anycast satellites and diversification of
network connectivity, to reduce their vulnerability to such attacks and to mitigate their effects.

1 See, D.Atkins and R.Austein. 2004. "Threat Analysis of the Domain Name System." August. IETF. RFC 3833.
Available at: <>.

Prepublication Copy--Subject to Further Editorial Correction
Furthermore, although some steps have been taken,2 more could be done to continuously monitor
the performance and traffic flows of the DNS infrastructure so as to enable rapid detection and response
to attacks or outages.
However, another serious vulnerability remains. As described in Section 2.4, "the original DNS
design did not include a mechanism to ensure that a name lookup was an accurate representation of the
information provided by the entity responsible for the information. DNS information was assumed to be
accurate as the result of general notions of network cooperation and interoperation (i.e., based on the
presumption that nobody would deliberately attempt to tamper with DNS information)." In more technical
terms, the initial design of the DNS did not incorporate "data origin authentication" and "data integrity
protection." However, because of increased fear of additional attacks on the DNS, these kinds of security
features have now become a major concern.
Data origin authentication is needed to help ensure that the results of DNS lookups come from
authoritative sources. A widely publicized case that involved the diversion of Internet users to an
undesired Web site drew attention to the lack of such authentication in the DNS.3
Data integrity protection is needed because DNS data flows could be compromised at any point
between the various name servers, resolvers, or other intermediaries; and the corrupted data can remain in
caches for extended periods of time.
To respond to these potential vulnerabilities, the technical community has over a number of years
developed DNS Security Extensions (DNSSEC).4 DNSSEC adds data origin authentication and data
integrity protection to the DNS. It aims to ensure that the recipient can validate that the data was sent
from an authoritative source and that it arrived at its destination unchanged.

4.2.1 Mechanics of DNSSEC

DNSSEC provides end-to-end protection through the use of cryptographic digital signatures that
are created by responding zone administrators and verified by a recipients' resolver software. In
particular, DNSSEC avoids the need to trust intermediate name servers and resolvers that cache or route
the DNS records originating from the responding zone administrator before they reach the source of the
query. DNSSEC also preserves the capacity for localized variations and independence within the DNS
In DNSSEC, resource record sets (RRSets) 6 within a zone are signed based on the model of
public key cryptography.7 To support each signing operation, two keys are generated: A private key (to
sign data) and the corresponding public key that is used to verify that the data were signed by the private
key. The process of signing takes data to be signed and a private key as inputs to produce digitally signed

2 Notably the establishment of the Operations Analysis and Research Center by the Internet Systems Consortium ­
see <> ­ and the on-line performance monitoring by the k-root ­< http://k.root->.
3 In 1997, Eugene Kashpureff diverted Internet users who were seeking Network Solutions Web site to his own site,
although this was intended as a publicity stunt rather than as a malicious attack. See "Locking Up DNS Troubles,"
Network Magazine, Rik Farrow, August 5, 2000, available online at
4 Defined in: Eastlake, D. 1999. "Domain Name System Security Extensions." March. IETF. RFC 2535. Available
at: <>. New RFCs are in preparation that will obsolete this RFC.
5 For example, the control of the private and public keys remains within each respective zone.
6 Resource records that have the same label, class, and type are categorized as belonging to the same RRSet. See
Box 3.2 for a detailed explanation of resource records.
7 For a review of public key cryptography and digital signatures, see Paul Albitz and Cricket Liu. 2001. DNS and
, 4th edition, Chapter 11. Sebastopol, Calif.: O'Reilly Media; and Chapter 4 in Trust in Cyberspace, Fred B.
Schneider, editor. 1999. Computer Science and Telecommunications Board, National Research Council.
Washington: National Academy Press.

Prepublication Copy--Subject to Further Editorial Correction
data as the output.8 However, DNSSEC involves signing the hash value of an RRSet, rather than signing
the full RRSet itself.9 (See Figure 4.1 for an illustration of the DNSSEC signing and verification process.)
Two copies of the RRSet are sent over the Internet to the recipient. One copy is not signed; the
other is hashed and then signed as described above. To verify the origin and integrity of the unsigned
RRSet, it is hashed using the same algorithm used by the sender. It is then compared with the verified, but
still hashed, copy of the RRSet created by the zone administrator. Matching hash values provides a high-
level of assurance that the non-signed RRSet is authoritative and that it has not been altered in transit.
DNSSEC can, when everything works correctly, give the data consumer (validating resolver)
some confidence that the received data is what the data producer (signing zone administrator) has sent. It
provides a basis for trusting that the data has been received without tampering. It does not, however,
assure that the data that the data producer sent is error-free or appropriate for the data consumer's
The DNSSEC extensions are based on four new resource record (RR) types: the public key
(DNSKEY), the resource record digital signature (RRSIG), the delegation signer (DS) and the
authenticated denial of existence (NSEC).10 The public key used to verify the digital signature of an
RRSet is stored in the DNSKEY RR.11 The digital signature is stored in the RRSIG RR, and several
RRSIG RRs may be associated with an RRSet, if more than one cryptographic algorithm is used for
signing the RRSet.
DNSSEC depends on establishing the authenticity of the DNS hierarchy leading to the domain
name in question, and thus its operation depends on beginning the use of cryptographic digital signatures
in the root zone. The DS RR facilitates key signing and authentication between DNS zones to create an
authentication chain, or trusted sequence of signed data, from the root of the DNS tree down to a specific
domain name. To secure all DNS lookups, including those for non-existent domain names and record
types, DNSSEC uses the NSEC RR to authenticate negative responses to queries. NSEC is used to
identify the range of DNS names or RR types that do not exist among the sequence of domain names in a

8 The crucial property of the digital signature is that it could have been produced only by someone with access to the
private key.
9 A hash algorithm is a mathematical process that converts a message to a probabilistically-unique fixed-length
string of digits that represents the original message. A hash algorithm is essentially uni-directional: Given a hash
value, it is nearly impossible to reverse the process to derive the original message. Since a hash value is typically
much less data than contained in an RRSet, it is generally more efficient to sign hash values rather than RRSets.
10 For detailed information about these RRs, see Roy Arends, Rob Austein, Matt Larson, Dan Massey, and Scott
Rose, "Resource Records for the DNS Security Extensions," IETF, RFC 4035, July 15, 2004, to be published in
2005 and will be available at <>.
11 The private key must be closely protected from public access, of course, and so it is not stored in a RR.
12 The implementation of DNSSEC also necessitates other changes that are too detailed to discuss here. For the
specifics on the two "new message header bits" (CD and AD) in DNSSEC, see Roy Arends, Matt Larson, Rob
Austein, Dan Massey, and Scott Rose, "Protocol Modifications for the DNS Security Extensions," IETF, RFC 4034,
July 15, 2004, to be published in 2005 and will be available at <>.

Prepublication Copy--Subject to Further Editorial Correction

Private Key

hash of
Public Key
signature of
RR Set hash

hash of
If Hashed RRSets are =, the Received RR Set is authentic; if not, it is not.
FIGURE 4.1 Use of DNSSEC to authenticate an RRSet.

Prepublication Copy--Subject to Further Editorial Correction
4.2.2 Deployment of DNSSEC

DNSSEC implementation on a global level faces a number of technical and non-technical
challenges. The process of cryptographically signing hash values derived from resource records, along
with the increase in the DNS packet size to accommodate large key sizes, adds significant operational
costs for organizations that manage DNS servers because of the increase in DNS data and the associated
increases in server computations and communications traffic.13 The implementation of DNSSEC also
increases the volume of Internet traffic and that, in turn, could increase the vulnerability of the Internet to
denial-of-service (DoS) attacks--a threat DNSSEC does not protect against, although DNSSEC may offer
more confidence in the responses of anycast satellites, which do provide a measure of defense against
DoS attacks. DNSSEC could also cause more timeouts that would degrade the quality of service for end-
users.14 DNSSEC also introduces more complexity to the DNS, and would add to the administrative
requirements for managing the security mechanism.15 For instance, the administrator of a large zone
would probably experience great difficulty in re-signing his or her entire zone daily. This would require
dividing the task among many smaller parallel operations that could be managed with software; a solution
that is feasible given the DNSSEC design (that makes signatures within a zone remain largely
independent), but would not be without additional costs.
Because public keys for the root zone will need to be replaced with new ones on a regular basis,
key management for the digital signatures presents another problem for DNSSEC. In particular, the
interaction of key revocation with global caching and the distribution of copies of a new public root key
remain unresolved,16 and this adds even more importance to the management of root zone keys. The
consequence of a corrupted root zone key is that it would break the chain of trust for source authentication
and data integrity that serves as the basis of DNSSEC. A related and more fundamental and thorny
problem that technical solutions could only partially resolve is reaching agreement over which
organization should have control of the root zone key. Obvious candidates for the controlling
organization include VeriSign, ICANN, and the Department of Commerce; other entities could also be
considered. However, until a controlling organization is identified, the deployment of DNSSEC is likely
to be delayed.17

While the introduction of DNSSEC imposes significant costs and does not eliminate all Internet
security concerns nor address all Internet threats,18 its implementation would represent considerable
progress in improving the security of the DNS. For example, it would raise the level of protection against
the falsification of DNS data to help in deterring identity-related theft and SPAM problems.19

13 Estimates for the increased computations and communications traffic associated with the introduction of DNSSEC
vary, but range from a five-fold to ten-fold increase. See Albitz and Liu (2001); Beth Cohen, "DNSSEC: Security
for Essential Network Services," May 12, 2003, available online at
<>; and "Securing the Domain Name System," Diane
Davidowicz and Paul Vixie, Network Magazine, January 1, 2000.
14 See David Berlind, "DNS Inventor Says Cure to Net Identity Problems is Right Under Our Nose," August 7,
2003, available online at: <,14179,2914447,00.html>.
15 See Cohen (2003).
16 Distributing new public root keys is difficult as they must be preconfigured in DNS root name servers, but they
cannot be delivered via DNSSEC extensions, and thus they may require an offline distribution mechanism. One
proposed solution involves the publication of a public root key in national and international newspapers, which
illustrates the magnitude of the problem.
17 Several facilities in the Netherlands and Sweden are examining how DNSSEC could operate when it is generally
deployed by examining procedures, such as key rollover, determining parameters for DNSSEC mechanisms, such as
key length and signature lifetimes, and other issues beyond the scope of this discussion. For more information about
current efforts of DNSSEC testing, see: <>.
18 See Derek Atkins and Rob Austein, "Threat Analysis of the Domain Name System," April 5, 2004 (work in
progress), available online at <>.
19 See Berlind (2003).

Prepublication Copy--Subject to Further Editorial Correction
Furthermore, DNSEC provides a basis to build trust on the Internet to support higher-level protocols
facilitating IP telephony and other Web services.20


The security of the DNS would be significantly improved if DNSSEC were
widely deployed among name servers for the root zone and TLDs in particular, and throughout the DNS
in general.

Conclusion: Urgent attention is needed to identify the organization that would maintain
control of the root zone key. The deployment of DNSSEC is likely to be delayed until this organization is

Recommendation: DNSSEC should be deployed throughout the DNS as practical, with
highest priority given to deployment in the root zone and the TLDs.


The Internet and the traditional telephone network operate differently. When a traditional
telephone call is made, switches create a circuit between the caller and the person who is called. That
circuit remains in place for the duration of the call. The process is called "circuit-switching." However,
when a message is sent from one computer connected to the Internet to another computer no such circuit
is established. Rather, the message is broken into packets and each is routed through the network
independently, possibly even following different paths, and reassembled at their destination in the proper
order. That process is called packet-switching. For the most part, the circuit-switched world of telephony
and the packet-switched world of the Internet have remained distinct. However, in recent years, a
convergence between the two has begun to occur, with increasing use of the Internet to transmit telephone
calls through a process called "Voice over Internet Protocol" or VoIP.21
The recognition of the potential convergence of telephony and the Internet was one of the
motivations for consideration by the technical community of ways to bring telephone numbers into the
Domain Name System. Doing so, it was thought, would facilitate communications between the Internet
and the world's telephone networks. The method they developed is called the Telephone Number
Mapping protocol, more commonly known as the ENUM protocol.22 Under the ENUM scheme,
telephone numbers, called E.164 numbers,23 are mapped (via the ENUM protocol) to domain names.
These are then mapped (in the DNS) to various resources by DNS lookups that lead to Uniform Resource
Identifiers (URIs).24 The main premise underlying development of the ENUM protocol is that standard
telephone numbers--familiar, globally unique identifiers easily usable on numeric keyboards--are likely

20 See John Leyden, "DNS Inventor Calls for Security Overhaul," The Register, April 11, 2003, available online at:
21 See, for example, the explanation of VoIP on the Federal Communications Commission Web site,
22 See Patrik Fältström and Michael Mealling. 2004. "The E.164 to Uniform Resource Identifiers (URI) Dynamic
Delegation Discovery System (DDDS) Application (ENUM)." April. RFC 3761. Available online at: <ftp://ftp.rfc->. The mechanism used by ENUM for mapping telephone numbers into the DNS was
first specified in late 1992 as part of an Internet "Remote Printing" model that could substitute for fax and other
telephone-enabled transmission mechanisms. That application and the mapping mechanism are described in: Carl
Malamud and Marshall Rose. 1993. "Principles of Operation for the TPC.INT Subdomain: Remote Printing­
Technical Procedures." October. RFC 1528. Available online at: <>.
23 E.164 refers to the global telephone numbering plan administered by the International Telecommunication Union.
RFC 3761 stipulates that the domain names generated through the ENUM protocol must adhere to the existing
E.164 country (or region) delegations.
24 See Chapter 6 for a definition of URIs.

Prepublication Copy--Subject to Further Editorial Correction
to persist. Consequently, making it easy to link the Internet and telephone naming systems may support
the development of new and improved services that use a telephony-like model.
Applications that can build on the ENUM protocol include voice communications, fax, e-mail,
and messaging. For example, a telephone call might originate from a standard desktop telephone set and
terminate at a telephone connected to the Internet (after passing through a gateway). The implementation
of the ENUM protocol may facilitate the completion of such VoIP telephone calls, although such calls do
not require the use of ENUM as an addressing mechanism. The ENUM protocol enables the use of a
single E.164 number to access applications based on the telephone network, the Internet network, or both
networks. Thus, it may enable increased functionality and/or lower costs for communications in such
interconnected networks.25

4.3.1 Mechanics and Operations of ENUM

The ENUM protocol specifies how telephone numbers are converted into domain names. The
conversion is best explained through an example. Begin with a telephone number such as +46-8-1234567.
Then remove all characters except the digits, put dots between the digits, and reverse the order, which
yields, in the example above, Then a second level domain name is appended, which
for the implementation of the ENUM protocol is ""26 The resulting ENUM domain name is
The deployment of ENUM is typically envisioned in tiers. The highest level within the ENUM
hierarchy--tier 0--corresponds with the selected second-level domain The name server
resource records in this second level domain would point to "national" tier 1 registries, such as (for Indonesia - telephone country code 62) or (for Belgium -
telephone country code 32).28 The delegation beyond tier 1 registries (and the definition of a "tier" itself)
may differ among countries. Various trials are underway in a number of countries to identify the most
effective models for those countries.29
A tier 1 registry could delegate directly to name servers that contain ENUM information.
However, in some models for the implementation of the ENUM protocol, a tier 1 registry would delegate
to multiple tier 2 operators (e.g., divided in a way that is based on how telephone numbers are partitioned
within a country). Tier 2 operators would then operate name servers that contain ENUM information that

25 The resources used to develop this subsection include presentations and discussions at the public forum of the
meetings of the Internet Corporation for Assigned Names and Numbers, Rio De Janeiro, Brazil, March 25, 2003,
available online at <>; materials from the
International Telecommunication Workshop on ENUM, Geneva, Switzerland, February 8, 2002, available online at
<>; the ENUM Web site of the International Telecommunication Union at
<>; "Frequently Asked Questions," available online at <>;
"The History and Context of Telephone Number Mapping (ENUM) Operational Decisions: Informational
Documents Contributed to ITU-T Study Group 2 (SG2)," RFC 3245, John C. Klensin, editor, March 2002; and
"Online Registries: The DNS and Beyond...," Release 1.0, September 16, 2003, New York: EDventure Holdings,
26 is the second-level domain name recommended by the Internet Architecture Board for ENUM use
in RFC 3761. The .arpa TLD is intended to support Internet infrastructure initiatives such as the implementation
of the ENUM protocol.
27 The Réseaux IP Européens (RIPE) Network Coordination Centre (NCC) is the administrator of the
as determined by the Internet Architecture Board.
28 Twenty-six codes have been delegated (twenty-eight have been approved) as of March 4, 2005, as reported by
RIPE NCC at <>.
29 For example, see <>.

Prepublication Copy--Subject to Further Editorial Correction
takes the form of Naming Authority Pointer (NAPTR) resource records.30 These records include NAPTR
records for service-specific addresses (e.g., an e-mail address, cell phone number, fax number, etc.31),
which would all be returned in the response to any DNS query about a particular ENUM domain name.
An important feature of NAPTR records is that they can convey priority ordering (e.g., try this address
first--if there is no response, then try this one). The situation described above is referred to by some as
the "calling party control model" because the DNS query for the NAPTR records retrieves all possible
contact modes--that is, access to this information is determined by the requestor.

Tier 3 services could also be offered. Services at this level could support operations after the
completion of a lookup of ENUM information (i.e., some of these operations might not depend on the
DNS in any way). For example, a lookup from a tier 2 name server could point to a proxy server that
contains tailored user information, rather than to service-specific addresses directly. This tailored user
information could, in turn, provide office addresses to all queries and, in addition, home addresses only to
those requests with particular characteristics. Alternatively, all queries to the NAPTR records could be
directed to this tailored user information, thereby providing the called party with control over what
contact information is made available (i.e., the "called party control model").32

4.3.2 Technical and Public Policy Issues

The deployment of the ENUM protocol raises anew most of the challenges associated with the
DNS and raises a few new ones. Thus, the technical and policy context for ENUM implementation
includes a wide variety of issues that should be resolved prior to the widespread deployment of the
ENUM protocol and serves as an illustrative case study for other applications that might be developed on
top of the DNS. (It also illustrates another one of the core Internet navigation issues: While ENUM
provides mechanisms for mapping from telephone numbers to the DNS and from a domain name to
relevant resources, it does not address the problem of determining a telephone number given some
(possibly inexact) form of the name of a person or organization (and perhaps some additional qualifying
attributes). On a global basis, that navigation problem is far more difficult than the challenges associated
with ENUM.)

Registrars and consumers. An important implementation issue is who has control over the
information in the name servers. Conflicts over the inclusion or content of NAPTR records need to be
resolved in some way. The design of the mechanisms for managing these conflicts can draw from past
experience with the DNS and telephone networks, which has included dealing with slamming (the
unauthorized change in service providers), number portability, and recourse in the event of fraud.
Privacy. Since the records in the DNS are publicly accessible, there is some concern about
the privacy of the personal information stored there. Of specific concern are URIs in the NAPTR records
that refer to personal information that an individual would not wish to have linked to a telephone number
in a freely accessible way.33 Alternatives such as the called party control model described above accord
individuals the ability to specify what kind of information will be publicly available and an opt-in strategy

30 NAPTR records are described in Michael Mealling. 2002. "Dynamic Delegation Discovery System (DDDS) Part
Three: The Domain Name System (DNS) Database." October. RFC 3403.
31 These addresses may be specified using a variety of protocols that include the Session Initiation Protocol (SIP),
which supports the negotiation of the parameters between endpoints for a real-time session. See Mark Handley,
Henning Schulzrinne, Eve Schooler, and Jonathan Rosenberg. 1999. "SIP: Session Initiation Protocol." March.
RFC 2543.
32 This discussion was derived, in part, from "Enum: Mapping Telephone Numbers on to the Internet: Potential
Benefits with Public Policy Risks," April 2003, Center for Democracy and Technology, Washington, D.C., available
online at <>.
33 However, note that the storage of E.164 numbers themselves in name servers is not a privacy issue. The issue
arises when E.164 numbers are linked to personal information.

Prepublication Copy--Subject to Further Editorial Correction
provides individuals with the ability to decide whether his or her telephone number will be included in the
DNS as an ENUM domain name.
Authentication and security. Under a system in which an individual must make a deliberate
opt-in decision, authentication of his or her identity is critical in substantiating that the person who wants
a number is really that person and that he or she has the rights to use that number (and to make
subsequent modifications to ENUM information). In addition, ensuring that the results from lookups to
ENUM information are authentic suggests that the implementation of DNSSEC is as critical for ENUM
deployment as it is for other DNS applications, as discussed in Section 4.2.
Institutions. Because ENUM is also dependent on telephone numbers and the various
policies that pertain to telephone numbers, the institutional framework includes the International
Telecommunication Union (ITU). Also, as ENUM is based on telephone number country codes, national
policies must be considered. A clash of institutional approaches may result, given the strong regulatory
tradition associated with telephone numbering that contrasts with the traditionally less regulated
management of Internet naming and addressing.
An important design characteristic of the DNS is the existence of a root zone that provides the
operational basis for global uniqueness and coherence. By contrast, telephone service providers
determine the country codes that they use for routing telephone calls. Each provider might not use the
same set of codes. For example, the country code +866 is used from many, but not all, locations in the
world to complete telephone calls to Taiwan. The +866 country code is not officially allocated to Taiwan,
but it is being reserved by the ITU, which manages the country codes in the world. Because the ENUM
model as currently implemented requires ITU and national approval for each ENUM delegation, the use
of standard ENUM for communication in and with Taiwan has been prevented by the People's Republic
of China.
Other. The deployment of the ENUM protocol raises other issues that are too detailed to be
discussed here, such as the disposition of an ENUM domain name when an individual terminates the
service for the corresponding telephone number. The references provided in this section provide pointers
to documents that explore these issues.

4.3.3 Alternate Models

In principle, ENUM-like domain names could be based on a unique identifier other than a
telephone number. For example, consider the use of any random identifier that is globally unique, such as
a product bar code. Or one might tie ENUM more closely to the existing ccTLD model, using ISO 3166
numeric codes rather than E.164 country codes to identify the country-specific part of the number.
Another alternative could call for the use (at least in part) of a different domain than Also,
the hierarchy of names need not be based on countries at all. However, it is unclear whether the adoption
of an alternate model instead of the ENUM protocol would provide the basis for a superior deployment.

The deployment of the ENUM protocol could support important new applications. However, it is
also the case that its deployment would reinforce the utility of telephone numbers. Assuming that it is
increasingly desirable to identify an individual or activity rather than a telephone number, the deployment
of the ENUM protocol might not be optimal in the long run because its use could forestall efforts to
develop systems with greater capabilities.

Conclusion: Overall, the plan to deploy the ENUM protocol could lead to applications that use
the DNS without necessitating any changes to DNS protocols or software. However, a number of
important technical and public policy issues would need to be resolved in each country that has an interest
in deploying ENUM. These issues include establishing the rights and requirements of registrars and
consumers, developing practices for the protection of personal privacy, implementation of procedures for
authentication and security, and development of an effective and efficient institutional framework for
operation of ENUM.

Prepublication Copy--Subject to Further Editorial Correction


One of the issues of particular interest in many countries is access to the Internet and the DNS
using home-country languages and scripts. As the number of users in countries with first languages that
are not based on the Roman characters used in the DNS increased dramatically through the 1990s, interest
developed in domain names based on non-Roman scripts (e.g., Chinese, Hebrew, Arabic, and so on).
Several major efforts were undertaken in the effort to accommodate internationalized domain names
(IDNs) within the Internet infrastructure.
Unfortunately, the design of the DNS presents formidable technical challenges for the
accommodation of languages that use non-Latin characters. As a lookup system, the DNS must be able to
determine unambiguously whether there is a match with a query or not. Comparing strings is much more
difficult than most people realize, because the definition of what is "equal" is often not deterministic. For
example, consider the case of the French language in Canada and France, for which there are different
rules of whether an accent stays over the character when it is converted from lower to upper case.34 And
some languages (e.g., Chinese) cannot even be reduced to a relatively small number of standardized
characters (e.g., the character set for English). As these challenges were articulated and analyzed by the
interested communities, it became clear that the widespread deployment of IDNs would necessitate a
number of compromises. Some of these compromises stemmed from intense arguments over the
preservation of cultural identities, the use of names that are semantically correct, and other linguistic
Other compromises have their roots in technical issues. Some of those who were very concerned
about the integrity of the DNS argued that internationalized domain names should be implemented in
applications (e.g., by reworking URL and similar formats to accommodate IDNs directly). Some other
technical experts argued that the deployment of IDNs should be executed through a major overhaul of the
Internet's infrastructure, rather than as an add-on. However, considerable pressure developed within the
interested communities to implement IDNs in the near-term and, therefore, solutions that would require
extensive changes in architectures or standards did not attract very much support. This pressure provided
the impetus for an effort led by the Internet Engineering Task Force (IETF) that culminated in a standard
solution, the Internationalizing Domain Names in Applications (IDNA) mechanism.35

4.4.1 Internationalizing Domain Names in Applications

The central goal of the IDNA scheme is to enable end-user viewing of IDNs (e.g., .cn)
without altering the DNS protocols themselves. Hence, even though an end-user may see an IDN, the
DNS itself only sees the usual LDH-style domain names.36 IDNA is entirely a client-side set of

There are a number of encoding systems for representing various language scripts. In the
discussions leading to the adoption of IDNA as a standard, it became clear that a constraint would be
needed on the number of encoding systems so that the introduction of IDNs would be tractable. Unicode

34 Another example is the "a" with diaeresis "¨" ("ä") which in German should be sorted and looked at exactly as an
"a" with diacritical character, but in Swedish has nothing to do with the character "a" except the look. The same
way Å as the physical unit Ångström is one character, but the initial character of the word Ångström is another
(which in turn is different from an "A"). The other problem with the German "a" with diaresis ("umlaut") is that it
may be considered to match the string "ae", while not all names containing "ae" match "ä".
35 IDNA is described in: Fältström, Patrik; Paul Hoffman, and Adam M. Costello. 2003. "Internationalizing Domain
Names in Applications (IDNA)."March. RFC 3490,which is available online at: <>.
36 These are domain names comprising letters, digits, and the hyphen--a subset of ASCII--as described in Chapter

Prepublication Copy--Subject to Further Editorial Correction
was agreed to be the client-side encoding system for language scripts.37 Thus, any user application based
on other encoding systems would first have to translate its internationalized domain names to a Unicode
representation prior to processing by IDNA-compliant procedures.38

The algorithms that make up IDNA work on the individual parts of a domain name separated by
dots, which are called labels.39 Translation can occur in two directions: from Unicode to LDH format
(ASCII) or the reverse.

The input to the "ToASCII" algorithm is a single label comprising Unicode code points.
However, before labels are processed, they must be normalized because different Unicode strings can
represent the same domain name.40 Thus, a profile ("nameprep") is applied--a string preparation and
normalization procedure for Unicode that is partially derived from Unicode Technical Consortium
(UTC)-specified normalization procedures.41 An encoding system ("punycode") is then used for mapping
"nameprepped" labels into conventional LDH-style labels.42 These labels are then concatenated (with
dots in-between the labels) to generate the resulting domain name. In the process of assembling the
resulting domain name, "xn--" is added as a prefix to denote that the domain name is an IDN;43 an
example of such a domain name is At this point, the DNS is used as
described in Chapter 3.

The process for going from ASCII to Unicode involves the use of the decoding algorithm in
punycode. The details of this process are described in RFC 3490.

Client-Side Support

There is a gap between "deploying IDNA" (i.e., adding IDNA-format (punycode) domain names
in DNS zones) and the actual ability of users to see, or provide as input, domain names expressed in
characters that are "native" to their preferred languages. The actual appearance of a domain name in
native characters on a screen or printout, or the ability to transcribe such characters from a sign on the
side of a bus into a URL to locate a Web page, requires that the relevant applications be upgraded to
recognize the IDNA format and to translate to and from local scripts.

Supporting Web Access

37 "Unicode is a coded character set containing tens of thousands of characters. A single Unicode code point is
denoted by "U+" followed by four to six hexadecimal digits, while a range of Unicode code points is denoted by
two hexadecimal numbers separated by ".." with no prefixes," as described in RFC 3490. Additional information
about Unicode may be found at <>.
38 The mappings to and from Unicode may not be obvious (and may become controversial), as the local encodings
sometimes make distinctions that Unicode does not, or vice versa.
39 For instance, in, there are three labels: www, example, and com.
40 For example, upper case characters would be converted to lower case characters. However, this case mapping
may be problematic, as in the case for handling diacritical marks where characters are mapped to upper case in
modern French as compared to older forms (still used in Québéc and elsewhere). Also, consider the Unicode string
"www.Exa$(not$) that would be normalized to Adopted from Eric A. Hall, "The
IDNA-to-ASCII Conversion Process," Network Magazine, June 1, 2004, p. 60.
41 See "Nameprep: A Stringprep Profile for Internationalized Domain Names," Paul Hoffman and Marc Blanchet,
RFC 3491, March 2003; and RFC 3454. For Unicode normalization, see "Unicode Normalization Forms," Unicode
Technical Report #15, Mark Davis and Martin Dürst, <>.
42 See "Punycode: A Bootstring Encoding of Unicode for Internationalized Domain names in Applications
(IDNA)," Adam M. Costello, RFC 3492, March 2003.
43 The description given here is a simplified one. The many details concerning the various algorithms may be found
in the RFCs referenced above.

Prepublication Copy--Subject to Further Editorial Correction
Some Web browsers have been upgraded to support IDNA names directly, whereas others,
including the most common browser, Internet Explorer, support IDNA through browser plugins.44 These
extensions to browser operations and syntax are not standardized and not consistent, leading to different
users getting different results depending on which tools they choose to use (or, more commonly, have
chosen for them).45 In the worst case, the consequence is a breakdown of the principle of referential
integrity--a putative domain name or URI, when passed from one user to another, acquires a different
meaning (or target) depending on the environments of the two users. Even when support for Web
browsers is achieved on a consistent and widespread basis, it will take a considerable period of time to
replace the million of copies of Web browsers. Forrester Research predicts that it will take at least two to
three years before IDNs can be really used for Web browsing and up to five years until 90 percent of
applications are IDN compatible.46

Supporting E-mail and Other Access

For applications other than the Web, the situation is yet more problematic. There is, in general,
no "patch" option equivalent to the browser plug-ins. While there are only a few heavily-used browsers,
there are very large numbers of client / user interface programs for e-mail, FTP, and other protocols.
Some of these programs are embedded in firmware on portable devices, which generally cannot be
upgraded in a practical way.
Electronic mail poses additional problems. For most users, the Web is largely passive: people
find and view Web pages, but relatively few users create their own Web content or need to establish
addressing or location information for it. E-mail, by contrast, is not passive. Most e-mail users are
actively engaged in the creation and receipt of e-mail. Addresses read over the phone or copied from
business cards or notes may well be more common for e-mail than for the Web. Moreover, Internet e-
mail operates in a "store and forward" mode: unlike, for example, the Web, there is normally no reliable
mechanism for a sender to determine, or negotiate, the capabilities of the receiver such as whether a
receiver can handle internationalized addresses. And, finally, users typically expect the left side or "local
part" (before the "@") of an e-mail address to reflect their names and related conventions (or other
personal identifiers such as nicknames) and to do so accurately.47 People are often extremely sensitive
about the spelling of their names (or other personal identifiers), and efforts to replace e-mail addresses
based on names with ones that involve semi-random strings have rarely been met with enthusiasm. Even
if the domain name part of the address internationalization problem were solved with appropriate user
agent software, it may well lead to more demand for properly-spelled and formatted local parts in local
scripts. Non-English-speaking users who have been using addresses containing Romanized
transliterations of their names are not likely to consider a transition to encoded ASCII strings that have no
mnemonic value and that generally cannot be pronounced to be a step forward, especially with their
expectations for native-script presentation.
Design and standardization efforts for e-mail local parts are in their infancy in 2004. There are
two major proposals. One tries to minimize the number of infrastructure changes that are required (just as
IDNA avoided requiring, or even permitting, any DNS protocol changes). The other proposal assumes

44 For examples of plug-ins, see <>. Internet Explorer does not provide direct support for
IDNA as of July 2004.
45 Improper resolution and browser plug-ins that were not stable enough, complicated, and slow to download were
among the reasons why Network Solutions, Inc. pulled out of the IDN business in early 2004. See "NSI Pulls Out
of IDN Registration, Citing Technical Problems," Washington Internet Daily, January 15, 2004.
46 See Thomas Mendel. "Internationalized Domain Names: Good Idea; Shame About the Execution," March 10.
Forrester Research, which is available online at:
47 Although the "local part" of an e-mail address is not within the scope of IDNA standards, it is an important issue
concerning the internationalization of the Internet more broadly.

Prepublication Copy--Subject to Further Editorial Correction
true internationalization is going to require rethinking e-mail in an internationalized context (and hence
requiring those who wish to take advantage of internationalized addresses to implement some upgrades).

4.4.2 Registries and Registrars

When it approved the initial versions of the IDNA standards, the Internet Engineering Steering
Group (IESG) issued a statement that discussed the scope of those documents and areas in which other
work was needed by other organizations.48 Specifically, it pointed out that IDNA addresses characters
only and that any relevant language or script issues, including near-equivalencies must be dealt with on a
per-registry basis. It also identified the special problems faced by generic top-level domains (gTLDs), the
importance of conservatism in characters that are permitted, and concerns about display issues with
converted IDNA strings.
After some discussion, and essentially as part of the process of getting several new country code
TLDs to sign up to a formal ICANN relationship, ICANN issued a set of guidelines for the deployment of
IDNs in TLDs.49 These guidelines build on the IESG statement and the work of the Joint Engineering
Team (JET; discussed below) and provide that top-level domain registries must use the IDNA standards
and must adopt conservative, language-specific approaches to IDN registration.
To protect their populations from confusion and fraud, and, in at least some cases, to comply with
ICANN guidelines, registries have begun to establish language-based policies for registrations. The key
to these involves specifying which particular characters can be registered out of the Unicode set or, when
language rules are interpreted strictly, which characters are permitted in combination. Of course, there is
no standard list of characters for any given language, and the Unicode Technical Consortium has declined
to make lists of characters that belong to named (by language or otherwise) scripts, so a "language" for
DNS purposes is ultimately a list of characters chosen by registries, with each registry free to make a
different choice. For the convenience of registries that are disinclined to reinvent the wheel, the Internet
Assigned Numbers Authority (IANA) set up a registry/catalog of these language/script/registry tables.50
The criterion for registration is that a TLD registry thinks the table is worthwhile; ICANN is not going
into the "what is really a language" business.
This leaves the more complex question of what scripts and languages a particular registry should
permit. The smaller the number of scripts permitted, the lower the odds of fraud or other undesirable
behavior. Prohibition of mixing of scripts (or languages) within a given label is almost implicit in a
language-based system and will prevent at least some problems,51 but may restrict some registrations that
might be considered reasonable.
All of these issues are much more complex for gTLDs, or others that are seen as serving the
entire world. Within a country, a decision as to which languages or scripts are more important than others
is at least tractable. While it may be very difficult, especially in countries that have more than one official
language, and where there are constitutional provisions prohibiting treating some of those as more
important than others, these are the types of decisions that governments are typically constituted to make.
For generic top-level domains, on the other hand, there is an, at least, implied requirement to treat all
registration applicants equally, which implies that policies such as "this script is preferred over that one"
or "this language is assumed when the choice of language cannot be determined" are much more difficult,
if not impossible to implement.

48 See <>.
49 See <>.
50 See "IDN Language Table Registry," <>.
51 At the time IDNA was adopted, the UTC representatives who were participating in the working group, were
convinced that a "one label, one script" rule would prevent many, perhaps most, potentially fraudulent cases
resulting from confusing one character with another. More examples have been turned up since then, and few, if
any, people actively engaged in IDN issues believe any more that such a restriction would eliminate even a large
fraction of the potential problems.

Prepublication Copy--Subject to Further Editorial Correction

4.4.3 Chinese, Japanese, and Korean Scripts

The scripts of Chinese, Japanese, and Korean (CJK) present special challenges. These languages
are based on Han ideographs, which are derived from pictographs and are constantly evolving. The
"simplification" of Chinese writing in the People's Republic of China (PRC) in the early 1950s created a
sharp divide in methods of writing Chinese, with the simplified characters being used in the PRC, but not
in Taiwan, Hong Kong, Macao, and other Chinese-speaking communities elsewhere in the world.
However, the mapping between Simplified and Traditional Chinese forms is not always one-to-one, but
sometimes requires knowledge of meaning or context.52 In addition, Chinese-based characters are used in
written Japanese (as Kanji) and in Korean (as Hanji). An algorithm for handing mappings between
Traditional and Simplified Chinese characters that was not sensitive to the particular language in use
would map (e.g., Kanji into Simplified Chinese) resulting in the characters becoming, not only incorrect,
but unreadable to a significant fraction of the Japanese population.
To address the CJK script issues, a joint committee, known as the Joint Engineering Team (JET)
was created among the network information centers and registries for Japan, China, and Korea. That
committee created a collection of guidelines for registration of the CJK languages and characters.53
Those guidelines, in turn, introduced two new concepts into DNS management: "variant characters" and
"reserved labels" that could be registered into the DNS only by the registrant of some other, primary,
In the JET model, the script associated with a particular language is defined by the entries of a
table, with the primary (valid) characters being listed, one per row, in that table. If a label were proposed
to be registered that contained any characters that did not appear among these entries, the registration
attempt would be rejected as not conforming to the script. Each one of these characters may be associated
with zero or more "preferred variants"--characters which, if they appear, are considered to be fully
equivalent to the "valid" character--and "character variants"--characters that might be confused or
substituted for the valid one. That is, many Han ideographs look exactly the same or have similar
appearances but are assigned different code points in Unicode. The variants for the individual characters
are then used to generate alternate (variant) versions of the labels. For example, if a label proposed for
registration were ABC, and "B" had variants "X" and "Y," a label set ("IDN package") would be formed
consisting of "ABC," "AXC," and "AYC." All of these labels would be reserved; that is, it would not be
possible for anyone but the registrant of "ABC" to actually register them in the DNS.54 The labels
generated from the "preferred variants" (as well as, of course, the original "ABC") would be
automatically registered; those from the "character variants" would be reserved and could be registered at
the option of the package registrant. Of course, if more than one of the characters in a label had a non-
zero number of variants, the number of variant labels generated by the combinations, and hence the size
of the IDN package, could become quite long--examples have been shown of some Chinese labels that
could generate hundreds of variants.55

52 Increasing communications and commerce among Chinese-speaking groups make it important that simplified and
traditional characters be treated as equivalent. The committee that did the writing simplification work, however,
followed the historical pattern of language reforms over the centuries: they did more than simply replace one written
character or one spelling with another but, instead, made some changes to disambiguate homographs and
consolidated other words.
53 See Konishi, Kazunori, Kenny Huang, Hualin Qian, and Yanwoo Ko. "Joint Engineering Team (JET) Guidelines
for Internationalized Domain Names (IDN) Registration and Administration for Chinese, Japanese, and Korean,"
RFC 3743, April 2004, available at <>. Also see Seng, James, "JET Guidelines for
Internationalized Domain Names," CircleID, May 8, 2004.
54 The JET guideline document uses the term "activate."
55 The JET guidelines view IDN packages as atomic--there should not be a mechanism for moving names in or out
of a package once it is created.

Prepublication Copy--Subject to Further Editorial Correction
Introduction of the "package" concept raises varying economic questions that do not arise with
LDH-style domain names. How should the pricing of IDNs reflect the reality that each registration may
cause other (sometimes many other) domain names to be reserved? Also, given that there is only one
possible authorized registrant for a reserved domain name, what options exist for pricing? Many new
possibilities arise in the realm of domain name pricing structures for IDNs.

Original form: .tw corresponding punycode:
Traditional corresponding punycode:
Simplified (corresponding punycode:
Relevant domain .tw .tw .tw .tw
.tw .tw .tw

FIGURE 4.2 Example of multiple forms of an Internationalized Domain Name.57

The challenges posed by CJK scripts also exist, though perhaps less severely, in alphabetic
languages. Thus, work to generalize the JET guidelines to alphabetic languages is underway. Two
attempts have been made so far to do that generalization work, or at least to construct recommendations
for considerations as to how to do it for particular alphabetic scripts and languages.58

4.4.4 Conclusions

It was recognized on its adoption--and it has become much more obvious since--that IDNA
solved only part of the internationalization problem. Remaining to be addressed are the questions of
consumer confusion--especially those questions that did not involve intellectual property issues; conflict
avoidance or resolution for similar-appearing names; differences in interpretations for different
languages; restrictions on registrations on a per-domain basis; implications for the UDRP and the Whois

56 Of course, it is the case that a domain name of the form xn--{string}.com, for instance, could have been
registered directly, irrespective of any IDN issues. However, as discussed in Chapter 2, there are strong reasons
against the registration of domain names that do not have some kind of semantic or mnemonic significance to
someone, and these specially-coded labels do not have that property. Indeed, an extensive search was done, and
none of these xn--strings actually appeared, at least in the accessible top few levels of the DNS.
57 This example was taken from Vincent W.S. Chen, "IDN Whois Challenges--TWNIC's Case Study," presentation
at the ICANN meeting, October 29, 2003, Carthage, Tunisia, available online at
58 See John C. Klensin, "Registration of Internationalized Domain Names: Overview and Method," draft-klensin-
reg-guidelines-os.txt, November 2004, work in progress.

Prepublication Copy--Subject to Further Editorial Correction
database; security issues raised by IDN;59 the implications of (and alternatives to) "multilingual" top level

Recommendation: Continuing and increased attention to internationalized domain names is
necessary. Efforts to coordinate work across different countries, regions, and language groups should be
undertaken to prevent the balkanization of the Internet.

Conclusion: The relative merit of an approach for implementing internationalized domain names
based on incremental fixes as compared to one that involves an infrastructure overhaul remains uncertain.

Conclusion: Although the ongoing work on internationalized domain names is important, it
addresses only a small fraction of the issues associated with internationalization of the Internet in general.


A challenge that has faced the DNS since its inception is that users often make errors when entering
domain names as part of Web URIs, e-mail addresses, or other applications on the Internet. The errors
may simply be misspellings or they may be the entry of non-existent or inactive domains; often they are
the result of a user guessing an address or remembering one incorrectly. When an erroneous domain name
arrives at some level in the DNS, the standard response is for a "no such domain" message to be returned
to the user. If the application is e-mail, then the response may be from a MAILER DAEMON reporting
that the mail could not be delivered to the non-existent address. If the application is the Web, then either
the ISP (say, AOL or Yahoo!) or the browser (say, IE) may send the erroneous address to a search
engine61 to initiate a search. In any event, the user receives information that the address entered is

However, in addition to the return of error information to the originator as specified by the DNS
technical standards and conventions, two controversial kinds of response to user errors have appeared
recently. Both derive from the fundamental law of Internet commerce: traffic => income. That is, traffic
to an Internet location (generally a Web site) can produce income for the owner of the site (through
advertising sales); the greater the traffic, the greater the income potential. Consequently, the commercial
imperative is to acquire as much traffic as possible. Controversy arises when that imperative leads to
actions that confound user expectations or, more fundamentally, challenge the underlying technology
standards and conventions on which smooth operation of the Internet has been based.

4.5.1 Traffic Aggregation

The first controversial response is called "traffic aggregation." 62 The aggregator sets up a Web site
on which multiple advertisers place advertising text that includes links to their marketing sites. The site
may or may not have a specific theme, such as travel or electronic products. The advertisers contract to

59The Unicode Consortium has published a draft Technical Report that addresses Unicode security issues ­
including IDN issues. See Mark Davis. 2005. Proposed Draft Unicode Technical Report #36. "Security
Considerations for the Implementation of Unicode and Related Technologies." February 20. Unicode Consortium.
Available at <>.
60 The implications of IDN introduction for dispute resolution are discussed in Section 5.6.3 and for Whois data and
the Whois protocol in Sections 5.7.2 and 5.7.3.
61 For a fuller discussion of search engines and various forms of paid advertising on search engines, see Chapter 7.
62 Two traffic aggregators, for example, are and Their Web sites are,
respectively, <> and <>.

Prepublication Copy--Subject to Further Editorial Correction
pay for "click-throughs" to their sites; that is, for visits that originate from the links on the aggregator's
site. To attract traffic, the aggregator invests in domain names that would result from likely user errors,
for example, misspellings or wrong guesses of the domain names of highly-trafficked Web sites. It may
also buy names that have not been renewed by the original registrant and solicit people who invest in
domain names for future re-sale (often called "cyber-squatters") to link their warehoused domain names
to the aggregator's Web site in exchange for a share of the resultant proceeds. The aggregator then awaits
the traffic resulting from users who misspell or incorrectly guess a domain name or attempt to visit a
domain name that is no longer operative or that has not yet been made operative and collects fees from
the advertisers, if any, to which they "click through."

Although users finding themselves at entirely different sites from those intended may be annoyed,
the operation of a traffic aggregation site is in itself neither illegal nor in contravention of the explicit
DNS technical standards. However, by responding to an erroneous query with a Web page that the user
was not specifically seeking, it does arguably contravene DNS conventions and reasonable user
expectations of the DNS service. Depending upon the domain name used to attract the traffic, and the
content on the page, it might also result in trademark infringement, unfair competition, or cyber-squatting
prosecution under the Anticybersquatting Consumer Protection Act or violate a host of consumer and
child protection laws. Nevertheless, apparently a sufficient number of the users find the advertised sites of
interest to follow their links and, thereby, to provide income to the aggregators. Moreover, there does not
seem to be a practical technical or regulatory way to control this practice outside of the listed legal
realms. Therefore, it will not be further considered here.

4.5.2 Site Finder by VeriSign

Far more controversial and subject to control was the offering of the Site Finder service by
VeriSign, which was launched without notice in mid-September 2003.63 That service, which was aimed
at users of the World Wide Web, re-routed any request concerning an unregistered domain name within
the .com and .net zones to a VeriSign-operated Web site featuring paid advertising links and a search
engine, rather than returning the usual "no such domain" error message. VeriSign described it as a value-
added service for users that could, at the same time, generate significant revenue for VeriSign from the
frequent errors in second level domain names in the .com and .net TLDs.64 However, the elimination
of the "no such domain" error across the .com and .net domains, which numerous applications depend
upon for their current operation, had a direct and an indirect impact on the performance of applications
other than the Web, on the DNS, and on the Internet in general. Many in the Internet technical and
operator communities believed that, even though Site Finder was implemented in strict conformance with
DNS standards, it was in conflict with their spirit. As a result of their strong complaints and ICANN's
written demand that it desist,65 VeriSign suspended the service (under protest) in early October 2003 and
pursued the matter in the courts, as is described below.

The intense reactions from the Internet technical and operator communities66 that prompted ICANN
to demand the suspension of the service67 until further review raised issues of two types: technical and

63See VeriSign's Site Finder FAQ at: <
64 VeriSign estimated that the number of misspelling to be 20 million per day, see CircleID, "Facts and Figures."
Available online at: <>.
65 See "Letter from Paul Twomey to Russell Lewis 3 October 2003." Available at:
66 Comments expressing concern about the Site Finder service are available online at
67 ICANN. 2003. "Advisory Concerning Demand to Remove VeriSign's Wildcard." October 3. Available online at:
<>. For a description of ICANN's ability to force

Prepublication Copy--Subject to Further Editorial Correction
institutional. The technical issues were, themselves, of two types: First, whether Site Finder would have
negative effects on the stability and security of the Domain Name System.68 Second, whether VeriSign
had followed an appropriate process for introducing operational changes that have potential effects on
other Internet processes and applications. The unilateral introduction of the Site Finder service by
VeriSign also raised fundamental institutional issues about the relationship between ICANN and VeriSign
and, by extension, the other gTLD registries.

Technical Issues

The Site Finder service introduced changes to the operation of the .com and .net top level
domains (TLDs), through the use of the wildcard address (A) record. Wildcards, which can be set up by
an authoritative name server to stand in for name and class records (see Box 3.2), are used to synthesize
records if no exact match exists in the zone. In the Site Finder case, the wild card entries in .com and
.net synthesized a response that sent requests for non-existent second level domains to the VeriSign
service Web site. The use of wildcards is specified within Internet Engineering Task Force (IETF)
standards for the DNS protocol,69 but their use generally has been localized or confined to an
organization.70 In contrast, the introduction of wildcards in the A records of the .com and .net TLDs
had an impact across a major portion of the DNS. They produced affirmative responses, instead of the
expected "no such domain" response, to every attempt by the numerous applications that use the DNS to
find a non-existent domain name within the two TLDs, as noted in the next section.

Security and Stability Issues

Technical Community Views. According to an assessment made by the Internet Architecture Board
shortly after the introduction of the Site Finder service, VeriSign's unilateral change had direct effects on
the many applications that use the Internet.71 For example, SiteFinder altered the normal operation of e-
mail servers, which would be to return to the sender any e-mail addressed to non-existent domains. The
result of the consistent affirmative responses for the VeriSign operated top-level domains was to send e-
mail addressed to the non-existent domains to the registry-operated server instead. This change affected
users, as the immediate notification of a non-existent domain could be delayed by several days or more.
It also affected network administrators who incurred costs to reconfigure servers, if they chose to bypass
VeriSign's server.72

According to the IAB, these changes also affected the utility of spam filters that rely on
identification of invalid domain names to block messages, and limited the operation of sequential lookups

VeriSign to suspend the SiteFinder service, see Jonathan Weinberg. 2003. "Site Finder and Internet Governance."
December 28, Available online at: <>.
68 For a discussion of the broader social, political, and privacy issues raises, see
<>. See also,
69 See, Paul Mockapetris. 1987. Network Working Group, Request for Comments: 1034 "Domain Names ­
Concepts and Facilities." November, Available online at: <> and Paul
Mockapetris. 1987. Network Working Group, Request for Comments: 1035 "Domain Names -
Implementation and Specification" November, Available online at: <>.
70 An example of a common use of wildcards is for mail resource records, or MX records, as they allow e-mail
server operators to synthesize all records locally that enable immediate notification that a domain name is valid
before a message is sent.
71 See Internet Architecture Board (IAB). 2003. "Commentary: Architectural Concerns on the use of DNS
Wildcards." September 19. Available online at: <
72 IAB, Commentary, September 19, 2003.

Prepublication Copy--Subject to Further Editorial Correction
that require notice of unsuccessful DNS queries to seek information from other sources.

Other direct effects of these changes, according to the IAB, included the inconvenience to users
who were rerouted to an English-language Web site, rather than receiving an error message in their native
language; the potential loss of privacy as a result of e-mail and other Internet traffic being rerouted to an
unintended destination; and the danger to Internet security and reliability caused by routing all the
erroneous traffic to one location, creating a single point of failure and a target for deliberate attacks.73
An indirect effect of this change was the development of various technical countermeasures to
circumvent the VeriSign SiteFinder service. The Internet Systems Consortium (ISC) issued a patch for
BIND, the software used on many domain name servers, that disabled the re-direct to the VeriSign Web
site and responded with an error message instead.74 While patches released by ISC and others75 provided
an immediate solution to the re-direct, the ad hoc decision by network administrators or ISPs to use a
patch or to create another workaround introduced inconsistent changes throughout the Internet,76 that had
the second-order consequence of limiting options for other services that operated within the boundaries of
either the protocols or reasonable user expectations.

VeriSign's View. VeriSign responded to the IAB criticism with a point-by-point rebuttal,77 which
asserted that (1) Site Finder did not violate the DNS standards. (2) VeriSign was working to provide other
language responses in the near future. (3) Site Finder, by adding a wildcard RR set to the .com and .net
zones and updating its server, can relieve the majority of e-mail problems. (4) Applications that rely upon
error messages can achieve the same effect by querying the DNS for the presence of a wildcard A record
in the zone. (5) The detection of erroneous domains is not a widely implemented mechanism for spam
identification and discovery and, in any event, is easily circumvented. (6) VeriSign has published
mitigation strategies for dealing with other protocols. (7) VeriSign's experience in securely and stably
operating redundant .com and .net servers enables them to protect the Site Finder service. (8) VeriSign
does not collect or retain personal information through Site Finder. (9) VeriSign is willing to work with
ICANN and the technical community to deal with Internationalized Domain Names and domains not in
the .com or .net domains. It also said that it shared the IAB's concerns with workarounds to bypass
Site Finder and has written a guide for application developers to help them write software consistent with
the DNS standards for wildcards.

Process Issues

Technical Community Views. While the technical communities recognize that the use of the wildcard
record does not violate DNS protocol specifications,78 its implementation in the two largest TLDs did not
follow principles that have guided the process for making Internet architecture decisions from the initial
development of the Internet to the present, which have sought to minimize the impact of changes on the
network.79 The traditional process of working out proposed changes with the technical communities aims

73 As noted in the IAB, Commentary, September 19, 2003. For an additional list of technical problems, see
74 For a description of the patch, BIND 9.2.3rc2, see <>.
75 Such as Imperial Violet. VeriSign Countermeasures: <>.
76 IAB, Commentary, September 19, 2003. See also, Benjamin Edelman. 2003. "The Aftermath: How ISPs
Responded to Site Finder Around the World," CircleID. October 6. Available at
77 See VeriSign. 2003. "VeriSign's Response to IAB on Site Finder Service." October 3. Available at
78 IAB, Commentary, September 19, 2003.
79 The two principles include, The Robustness Principle, "be conservative in what you do, be liberal in what you
accept from others." Jon Postel, RFC 793, and The Principle of Least Astonishment, "A program should always

Prepublication Copy--Subject to Further Editorial Correction
to maintain flexibility for all applications that use the DNS and balance the impact of changes on network
operators, users, and the overall stability of the DNS and the Internet in general. Furthermore, as a matter
of principle, the technical communities insist that innovation should take place not within the DNS, the
core infrastructure of the Internet, but rather on top of the DNS, at the edges ­ the applications that use the
Internet and the DNS.

Examples of innovation at the edges consist of services similar to Site Finder, but which reroute
misspelled or non-existent domain names at the Web browser, such as Internet Explorer, or the ISP, such
as America Online. Because the service is limited to the Web browser or ISP, other protocols, such as
email and FTP, are not affected by the redirect and will still receive a "no such domain" response.
Changes at the core tend to make service offerings at the edges more difficult, as the redirect
offered at the DNS level overrides the changes made at the Web browser or ISP level, requiring these
services to work around the high-level implementation. While services offered at the edges cause the least
harm to the network overall, they are also the most beneficial to users, as they tend to offer more options
to elect the service the user wants to receive, to disable it, or to switch to another Web browser or ISP.
Shortly after Site Finder was introduced, ICANN requested that its Security and Stability
Advisory Committee (SSAC) undertake a study of Site Finder's implications for the security and stability
of the Internet. After public hearings and comments, the SSAC issued its report in July 2004. 80 Its
primary focus was not on Site Finder, per se, but rather on the facts that: "core registry operations were
modified, thereby changing existing services, and the change was introduced abruptly without broad
notice, testing, refinement or community agreement." It found that Site Finder (1) "disturbed a set of
existing services that had been functioning satisfactorily;" (2) "violated fundamental Internet architectural
principles by blurring the well-defined boundary between architectural layers" and moving more control
toward the center and away from the periphery; and (3) proposed mechanisms "to ameliorate the
undesirable effects of their diversion" that "put VeriSign in the implementation path of every existing and
future protocol that uses DNS."
In addition, the SSAC found that "the abruptness of the change violated accepted codes of
conduct that called for public review, comment and testing of changes to core systems." That process is
intended "to ensure that changes are introduced with minimal disruption to existing services and hence
with minimal disruption to the security and stability of the Internet." Moreover, it "precluded the
possibility that administrators, IT departments, ISPs and other intermediaries on whom end users rely
might be adequately prepared to deal with the consequences." The SSAC found that: "In response,
workarounds and patches were introduced quickly, cumulatively reducing the overall coherence of the
system and again violating the established practices of public evaluation, testing, discussion and review
before core services are implemented and deployed. These workarounds further blurred the functional
layers intrinsic to the Internet's robust architecture and in some instances created additional and
unintended harmful effects."
The SSAC made recommendations to eliminate "synthesized responses" from TLDs that serve
the public and that satisfy several technical conditions and to eliminate shortcomings from the
specification of DNS wildcards and their use. Most significantly, it recommended that "changes in
registry services should take place only after a substantial period of notice, comment and consensus
involving both the technical community and the larger user community." It asserted that the process must
"(i) consider issues of security and stability, (ii) afford ample time for testing and refinement and (iii)

respond in the way that is least likely to astonish the user." [source unknown], IAB, Commentary, September 19,
80Report submitted to the ICANN Board by the Security and Stability Advisory Committee. 2004. "Redirection in
the Com and Net Domains." July 9. Available online at: <

Prepublication Copy--Subject to Further Editorial Correction
allow for adequate notice and coordination with affected and potentially affected systems managers and
end users."81

VeriSign's View. As its use of a wildcard A record in the TLDs did not deviate from the IETF standards
that describe the DNS protocol specifications,82 VeriSign maintains that this implementation is a
legitimate use of the wildcard and a valid service innovation that adds value for user searches that are not
well served by the "page not found" error message.83 Additionally, VeriSign selected its own technical
advisory group to test the Site Finder service before it was deployed, arguing that a comparable process
within the broader technical community would have taken much longer to complete and was incompatible
with the pace and time horizon of business decisions.84

In August 2004, VeriSign published a response to the SSAC's report.85 In VeriSign's view,
SSAC's "purported `findings' and `recommendations' are inappropriate, unsubstantiated, and themselves
contrary to longstanding written standards and specifications for the operation of the DNS and the
Internet." According to VeriSign, since the SSAC did not find that "Site Finder, or wildcards generally,
pose a threat to the security and stability of the Internet's naming and address allocation system," its
"findings" and "recommendations" exceeded the scope of SSAC's charter. VeriSign argued that the
SSAC started its analysis with a predetermined conclusion and its report was written by persons who are
opponents of Site Finder or competitors of VeriSign. Of greater general significance, was its assertion
that the report's findings and recommendations "would in effect restrain technical innovation and
commercial practices on the Internet on the basis of vague and unwritten `codes of conduct' and self-
styled `established practices' that, contrary to the Report, do not represent consistent Internet practices
and conduct."

VeriSign asserted that the "well-defined boundary between architectural layers" claimed by the
SSAC is violated by "multiple technologies widely used on the Internet, such as network translators and
firewalls." Furthermore, VeriSign stated that Site Finder did not change the positioning of the DNS in the
layering of network services, while the SSAC-endorsed processing of IDNs at the DNS level would, by
their own analysis, "blur" the boundaries between architectural layers.

In sum, VeriSign charged SSAC with using "a façade of technical orthodoxy to mask a rigid
adherence to the status quo of the DNS, which is antithetical to the very nature of the Internet and
inconsistent with the RFCs, which themselves recognize the importance of innovation to the Internet."
VeriSign went on to argue that "contrary to ICANN's clear directive, SSAC has failed to quantify or
independently verify any of the purported problems described in the Report, raising serious doubts that
they were real, serious, or widespread."86

Institutional Issues

While the Site Finder service raised contentious issues of adherence to technical standards and
processes, it also brought to the fore a critical and equally contentious issue of authority over and
responsibility for the service offerings of TLD registries, especially gTLD registries that have signed
agreements with ICANN. (In Section 3.4.3 it is noted that, as of October 2004, ICANN had such

81 All quotes are from the Executive Summary of the SSAC Report "Redirection in the Com and Net Domains." Op.
82 VeriSign's Response to IAB Commentary "Architectural Concerns on the Use of DNS Wildcards" dated
September 23, 2003. See <>.
83 VeriSign's Site Finder Implementation. See
84 See, Charles Cooper. 2003. "The Cultural Divide and the Internet's Future." CNET October 16,
Available online at: <>.
85 VeriSign. 2004. "VeriSign, Inc.'s Response to Report from the ICANN Security and Stability Committee re
"Redirection in the Com and Net Domains." August 5. Available at: <>
86 All quotes are from VeriSign, Inc.'s Response, op cit.

Prepublication Copy--Subject to Further Editorial Correction
agreements with 10 of the 15 gTLDs.) The issue is of particular significance to the relationship between
ICANN and VeriSign, the registry for the two largest TLD domains, which contain over 30 million
second level registrations between them.

Specifically, the issue is: To what extent and by what processes can ICANN control the offering
of new services or the modification of existing services by TLD registries with which it has contracts? (It
has no clear authority over most other TLD registries, although it could in theory use the threat of
exclusion from the root to control the behavior of other registries. In practice, that threat is unlikely to be
used or to be effective under current circumstances.)

In his letter to VeriSign87 demanding the suspension of Site Finder services on .com and .net, the
president of ICANN asserted that "our review of the .com and .net registry agreements between
ICANN and VeriSign leads us to the conclusion that VeriSign's unilateral and unannounced changes to
the .com and .net top level domains are not consistent with material provisions of both agreements."
He went on to list six specific provisions of the agreements that VeriSign was, in ICANN's view,

More generally, there is the question of whether the introduction of Site Finder abused the public
trust that accompanies the monopoly position granted to VeriSign as the sole operator of the .com and
.net TLDs. 88 Did it take advantage of that monopoly position to extract profits from unregistered
domain names in unfair competition with ISPs, browsers, and search engines? For example, the second
level domain names directed to the traffic aggregation sites described in Section 4.5.1 must all be
specifically registered and a fee must be paid to the registrar, who in turn pays VeriSign for each name. In
contrast, Site Finder effectively redirected every unregistered second level domain in .com and .net to
VeriSign's service, generating traffic-based advertising revenue for VeriSign. Because of VeriSign's
position as the sole registry for those two TLDs, it did not have to specifically register those names in
order to control the response to a request for them.

From VeriSign's perspective, ICANN overstepped its authority as a technical-coordination
organization and prevented it from continuing to offer services that benefited Internet users.89 In pursuit
of that argument, in February 2004 it filed a federal lawsuit in the U.S. District Court, Central District of
California, charging that ICANN "overstepped its contractual authority and improperly attempted to
regulate VeriSign's business in violation of its charter and its agreements with VeriSign."90 VeriSign
asserted that ICANN "stifled the introduction of new services that benefit Internet users and promote the
growth of the Internet." It asked the court to assess damages against ICANN and for ICANN to treat
VeriSign in a "fair, reasonable, and equitable" fashion in the future.91

VeriSign's anti-trust claims against ICANN were dismissed in May 2004, but the court initially
allowed VeriSign to file an amended complaint to try to strengthen its legal arguments. However, on
August 27, 2004, the Court dismissed the claims with prejudice; specifically, the court held that
VeriSign's claims about competitors controlling ICANN's Board could not be supported.

VeriSign's remaining causes of action based upon contractual matters resulting from the registry
agreements had to be re-filed in California state courts. In August 2004 it made such a filing in the
California superior court in Los Angeles County. VeriSign claims that: "Were VeriSign to defer offering

87 See Paul Twomey letter, op. cit.
88 For more information, see The Cooperative Agreement between the Department of Commerce and VeriSign,
Available online at <>.
89 VeriSign reported receiving 5 million unique visitors per day while the service was operating, see John Leyden.
2003. "Users `vote with their mouses' for Site Finder." The Register. October 9, Available online at
90 VeriSign Press Release. 2004. "VeriSign Files Lawsuit Against Site Finder." February 26. Available at:
91 Declan McCullagh. 2004. "VeriSign sues ICANN to restore Site Finder." February 24. CNET.
Available at: <

Prepublication Copy--Subject to Further Editorial Correction
such services to the public during the effective period of the 2001 .com Registry Agreement, or to
modify such services due to ICANN's conduct and threats, VeriSign will suffer irreparable losses of
revenue from third parties, profits, market share, competitive position, reputation and good will.
Furthermore, millions of Internet users will be deprived of the improved functionality and quality of
VeriSign's services."92 As of October 2004, the suit remains open.

4.5.3 Conclusions

The Internet and the Domain Name System have operated successfully over two decades, despite
manyfold increases in connectivity and connected devices and a great expansion of users and uses. As
described in Chapter 3, their successful adaptation to rapid change has been based on a shared
commitment among the operators of the Internet and DNS to adhere voluntarily to a set of open standards
strictly vetted by the technical community and to a collaborative process of cautious and controlled
change. This commitment has held even though the operators are vastly different in nationality and in
type academic, commercial, governmental, not-for-profit and are not subject to the authority of any
overall controlling body. The commitment is threatened, however, by two external forces. One is the
desire by some governments and international agencies to introduce stronger international regulation of
Internet operations. This force is discussed in Chapter 5. The other is the commercial imperative
described earlier.

The commercial organizations that operate key elements of the DNS are appropriately driven by
the goal of increasing revenue and profit. As observed earlier, in the Internet that goal is best served by
attracting and capturing the attention of user traffic, which can be translated into advertising dollars.
Consequently, there is a natural pressure on commercial operators to find ways to do so in competition
with other operators. This is what happened in the Site Finder case. VeriSign saw an opportunity to
capture substantial traffic from unregistered domain names and, driven by its commercial imperative,
took it. In doing so, it made an unanticipated use of a DNS standard for wild cards. Moreover, it launched
its new service as a surprise, without vetting it with the technical community or informing other operators.
VeriSign has defended its actions as being within its rights to provide innovative new services that offer
benefits to users. However, the technical and operator communities have complained vigorously about
VeriSign's breaking of the shared commitment to standards and process.

Although Site Finder may adhere strictly to published standards and its introduction might even,
in some views, be strictly consonant with VeriSign's rights to innovate, VeriSign's action poses two high-
order challenges to the successful operation of the DNS and the Internet.

First, VeriSign is not like any other TLD registry. It contains roughly half of all second-level
domain name registrations and almost all the commercial and network infrastructure domains. It was
granted the right to operate the registry as a commercial monopoly by the Department of Commerce and
ICANN. Therefore, it is effectively an international public utility whose actions have profound and
widespread effects across the entire Internet. When it is seen in that light, it becomes clear that it would
be reasonable for it to be required to, at the very least, obtain formal approval from its contractual
regulator before introducing any new service with wide ranging consequences.

Second, and more significant, VeriSign's action could set in motion a commercial rush among
other operators of the DNS. Suppose VeriSign's actions were copied by other commercial registries. The
consequence for the stability and predictability of operations of the DNS could be profound. By ignoring
the shared commitment among operators to accept the authority of the technical community on standards
and new services and to adhere to a collaborative and cautious process of change, VeriSign tore a hole in
the invisible web of implicit understandings that has been critical to the success of the DNS and the
Internet. In remains to be seen whether the outcome of its court cases will determine whether that web

92 Paul Festa. 2004. "VeriSign sues ICANN in state court." August 31. CNET Available at:

Prepublication Copy--Subject to Further Editorial Correction
can be mended.

Recommendation: ICANN should strengthen its contracts with TLD operators (especially the
largest ones) to ensure that it has the authority to review proposed changes in their services that could
have a detrimental effect on the DNS or on other services that depend on the DNS. It should establish an
open, transparent, and speedy process of review for such changes that solicits contributions from the
technical community, other DNS operators, other affected Internet operations, and end users.

Recommendation: TLDs and other DNS operators that do not have agreements with ICANN
should voluntarily agree to adhere to published technical standards and to consult the technical
community and conduct public review processes before introducing new services that could have a
detrimental effect on the DNS or on other services that depend on the DNS.

The issues raised by VeriSign's introduction of Site Finder are both technical and institutional. As
such they serve as an appropriate bridge to the next chapter, which addresses the issues facing the
institutional framework of the Domain Name System.


Prepublication Copy--Subject to Further Editorial Correction
The Domain Name System: Institutional Issues

The institutional framework of the Domain Name System (DNS), as described in Chapter 3,
comprises a diverse group of organizations carrying out their various responsibilities with a high degree
of autonomy. No single institution--not even the Internet Corporation for Assigned Names and Numbers
(ICANN) or the U.S. Department of Commerce--has effective authority over all of the participating
organizations. Nevertheless, this group of organizations has successfully managed the DNS through two
decades of rapid growth of the Internet, and of the most significant applications on the Internet such as the
Web and e-mail, which rely on it. However, as the Internet and its applications have grown in importance,
so have the attention and the controversy that the DNS and its institutional framework have attracted.
That critical scrutiny has raised a number of issues concerning the structure, governance, management,
and operations of the organizations that manage the DNS.

Most of those issues are being actively addressed, but the organizations involved face the reality
of multiple, often conflicting interests--both public and private--becoming more actively engaged in
their activities. That makes achieving consensus decisions--in the tradition of the early Internet
community--increasingly difficult and leaves a residue of dissatisfaction with any decision that is taken.
Since attention to the institutional framework of the DNS can be expected to continue to increase with the
Internet's growing significance as a critical global communications infrastructure, so too can the critical
institutional issues be expected to receive increasing scrutiny and to give rise to increasing controversy.

This chapter is a guide to the principal institutional issues that have already or can soon be
expected to come to the fore. For each institutional issue, the principal alternatives that have been publicly
identified are summarized, and the alternatives are compared. Where the committee is in agreement, its
conclusions and recommendations are presented. In all cases, the intent is to provide a clear description of
the alternatives and the arguments for and against them as background for current and future policy
Some of the issues, such as ICANN's management structure, may appear to be resolved for the
time being. In the committee's view, understanding the conflicting proposals that preceded the present
resolution illuminates the pressures that remain in the background and that may, in the future, force the
issue once again onto the policy agenda.
The following issues are addressed in this chapter:

1. Governance of the DNS. How should the DNS be governed? What should be the role of the
U.S. government, international organizations, and ICANN?
2. Management of the DNS. What changes in ICANN, if any, are appropriate?
3. Oversight and operation of root name servers. Is there a need for greater oversight of the root
name server operators? If so, how might it best be conducted?
4. Regulation of generic top-level domains (gTLDs): number and process. Can and should new
gTLDs be added? If so, how many new gTLDs should be added, and how fast? What types should they
be, and how should they and their operators be selected?
5. Oversight of country code top-level domains (ccTLDs). Should anything be done to increase
ICANN's oversight of and authority over the ccTLDs? If so, what form should its increased authority
take, and how can it be implemented?
6. Resolution of conflicts over domain names. Does the Uniform Domain Name Dispute
Resolution Process (UDRP) need to be improved? If so, how should it be improved?
7. Provision and protection of Whois data. What is the appropriate balance among the various
interests in Whois data?

Prepublication Copy--Subject to Further Editorial Correction

In contrast to most of the other chapters in this report, this one deals with issues for which
opinion and values play a significant role. Consequently, many of the citations and references are to
advocacy documents, not to peer-reviewed scientific or technical papers. These references serve as
pointers to places where the proposals being summarized were presented. When there were multiple
similar proposals, one or two have been selected as references, either because they appeared to be the
most representative or because they were the most readily accessible. The committee's use of these
references should in no way be considered an endorsement of the point of view expressed.


Issues: How should the DNS be governed? What should be the role of the U.S. government,
international organizations, and ICANN?

As explained in Chapter 3, the U.S. government currently possesses the final authority to make
key decisions affecting the DNS. Specifically, it must approve all changes in the root zone file and,
thereby, controls the designation of top-level domains (TLDs) and the assignment of responsibility for
their operation. In this way, it functions as the steward of the DNS, exercising its authority and making
decisions for the larger Internet community.
Through a memorandum of understanding (MoU) signed in November 1998, the U.S.
government delegated day-to-day operational authority to ICANN, which makes recommendations to the
U.S. Department of Commerce (DOC) for its decisions affecting the DNS. In this capacity, ICANN
recommends the addition of new TLDs and redelegations of existing TLDs. (As noted in Chapter 3,
ICANN has additional responsibilities for Internet Protocol (IP) addresses and protocols.) ICANN might
be thought of as the registry for the root, sponsored by the U.S. government. ICANN nominally has the
same responsibilities as other registries: it registers entries into the zone file and sees to the distribution of
that zone file to the name server operators. However, as noted in Chapter 3, unlike the TLD registries,
ICANN does not directly contract for operation of the root name servers, nor does it have contracts with
all the TLDs. In fact, in October 2004 it had agreements with only 10 of the 15 gTLDs and 12 of the 243
ccTLDs. However, because it recommends any new and revised entries into the root zone, because its
agreements with the largest gTLDs set the rules for their operations and those of their accredited
registrars, and because those rules strongly influence the operations of the other TLDs, ICANN is
generally perceived to be the manager of the DNS.
Against this background, the issue of the proper form of DNS governance can be divided into two
separate but closely interrelated issues: (1) Where should the stewardship--the final authority for key
decisions--of the DNS reside? and (2) How should management authority--the registry function for the
root--for the DNS be exercised? Although these roles are distinct at present, one possible answer is that
they be combined.

5.1.1 Relationship to Governance of the Internet

Before addressing the questions of stewardship and management authority for the DNS, it is
important to emphasize the distinction between governance of the DNS and governance of the Internet.
Clarifying this distinction is necessary because the DOC's and ICANN's visible responsibility for a key
part of the Internet's infrastructure and the fact that decisions affecting the DNS have economic, social,
and political consequences have tended to focus discussions of Internet governance on the DOC/ICANN
and the DNS. However, there are many aspects of the Internet that are or might be subject to governance

Prepublication Copy--Subject to Further Editorial Correction
but that lie outside the DNS and the areas of responsibility of its managers.1 These include, for example,
controlling spam; dealing with use of the Internet for illegal purposes; resolving the "digital divide"
between developed and developing countries; protecting intellectual property other than domain names;
protecting privacy and freedom of expression; and facilitating and regulating e-commerce.2 Furthermore,
there are increasingly important aspects of the Internet that are currently subject to little public
governance, such as search engines and directories. Thus, DNS governance is a part of, but does not
include many aspects of, Internet governance. Since at the time of this writing in December 2004 there
have been no practical proposals for broad governance of the Internet, this chapter focuses solely on
governance of the DNS. (The institutional issues associated with Internet navigation are discussed
separately in Chapter 8.)

Conclusion: Governance of the DNS is part, but not all, of Internet governance. ICANN and the
DNS are not the proper vehicles to address most Internet policy issues.

5.1.2 Where Should Stewardship of the DNS Reside?

The U.S. government acquired responsibility for and authority over the DNS root by virtue of its
historical initiative and financial investment in supporting creation of the Internet and the DNS, as
described in Chapter 2. However, as the Internet has become international in extent, support, and
operation, the formal legitimacy of the U.S. government's continued authority over the root has come
under challenge, such as in the context of the World Summit on the Information Society (WSIS).3

Critics, including representatives of the governments of Brazil, South Africa, Russia, and China,4
argued there that the DNS is so central to the reliable and effective operation of the now fully global
Internet that its control by a single nation is no longer justifiable. Some would prefer that the ultimate
stewardship reside in an intergovernmental body, such as a U.N.-affiliated agency--for example, the
International Telecommunication Union--or a new organization specifically negotiated by treaty, such as
a "World Internet Organization"; others prefer an international body that includes governments, the
private sector, and civil society. Many of the critics are national governments that have felt left out of the
ICANN process, in which their representation is through the Governmental Advisory Committee (GAC),5

1 The issue of Internet governance, in general, has become the subject of intense scrutiny in preparation for the
second phase of the World Summit on the Information Society (WSIS) to be held in November 2005. The ITU held
a workshop on Internet governance in February 2004 at which many views on the subject were presented. For an
extensive discussion of Internet governance and the place of DNS governance within it, see, for example, Don
McLean. 2004. "Herding Schrödinger's Cats: Some Conceptual Tools for Thinking About Internet Governance,"
ITU Workshop on Internet Governance. February. Available at
<>. In preparation for the WSIS meeting in 2005, the
Working Group on Internet Governance (WGIG) was appointed in November 2004.
2 The ITU invited selected experts to present papers on Internet governance at a workshop. These papers, which
present a variety of personal views on this broad range of issues, are available at
3 For information about the World Summit on the Information Society, see <>.
4 See "Global Fight Looms for Net Management," Reuters, September 16, 2003. Available at
5 The Governmental Advisory Committee comprises the representatives of 74 nations, as well as several
international and regional organizations--the African Telecommunications Union (ATU), Asia Pacific
Telecommunity (APT), European Union (EU), International Telecommunication Union (ITU), Organization for
Economic Cooperation and Development (OECD), South Pacific Forum (SPF), and the World Intellectual Property
Organization (WIPO).

Prepublication Copy--Subject to Further Editorial Correction
which until 2003 had advisory status only.6,7 Indeed, Brazil has asserted that "it is a myth that
governments have a say in ICANN's activities via the GAC . . . the influence of governments is
comparable to the influence of non-shareholders in a private company."8

However reasonable the move toward international stewardship might appear in theory, in
practice any change can be made only with the acquiescence and active participation of the U.S.
government Not only would the U.S. government have to be an important party to any transfer, but it
also holds an effective veto because all of the root name server operators would have to agree to accept
the root zone file from a new source, yet 3 of the 12 operators are U.S. government agencies and 6 others
are U.S.-based organizations that may well be reluctant to take actions contrary to the wishes of the U.S.

5.1.3 Alternatives

In this light, an examination of three alternatives to U.S. government stewardship is instructive:
an existing intergovernmental organization, a new international organization formed by treaty, or an
international (but non-governmental) organization, such as ICANN or a successor. (Another possibility
would be to divide DNS governance responsibilities among several organizations, each of which would
have its own steward. Although that may be an outcome of DNS governance decisions, this section
addresses the stewardship of the manager or managers of the DNS, whatever form it or they should take.)

Alternative A: Existing Intergovernmental Organization--International Telecommunication


Of all the existing intergovernmental organizations, the one that claims the most relevant
experience and that has expressed the greatest interest in DNS governance is the Geneva-based
International Telecommunication Union (ITU), a treaty organization affiliated with the United Nations.9
Its membership comprises about 190 nations as well as more than 650 firms and international
organizations in the telecommunications and information technology industries. Its Standardization Sector
(ITU-T) has long experience with the adoption and implementation of international telecommunications
standards, primarily for telephony but also for some computer networking technologies that are now
considered historical in much of the world.10 A separate organization within the ITU, the

6 For a history and analysis of the role of the GAC in ICANN, see: Wolfgang Kleinwachter. 2003. "From Self-
governance to Public-private Partnership: The Changing Role of Governments in the Management of the Internet's
Core Resources." Loyola of Los Angeles Law Review. Vol. 36. Spring. pp. 1103-1126.
7 In September 2004, Norway argued, in a submission to a preparatory meeting for the WSIS, that the GAC needs
stronger funding and "cannot continue to have a mere counseling role to the ICANN." Norway called for a stronger
and better-funded secretariat that would enable the GAC to focus on more strategic and political issues. See World
Summit on the Information Society. Working Group on Internet Governance. Written contribution from Norway,

which is available at <>.
8 As reported by ICANNwatch: "WGIG Will Reassess--or Reassert?--Governments' Role in Internet," which is
available at <>.
9 Other U.N. agencies have expressed interest in the broader question of Internet governance, among them the
United Nations Educational, Scientific, and Cultural Organization, which prepared a position statement for the U.N.
I.C.T. Task Force Global Forum on Internet Governance held in March 2004 in New York. See
10 "The ITU-T performs world-wide administration, and acts as the forum for policy management, of a number of
naming and address allocation systems that are essential for the good functioning of critical infrastructures,
including the physical-layer infrastructure of the Internet itself . . . [and] such well-known examples as the E.164

Prepublication Copy--Subject to Further Editorial Correction
Radiocommunication Sector (ITU-R), has responsibility for radio frequencies and some regulatory
authority under the applicable treaties. Its processes and procedures are mature and, according to the ITU,
"there are sufficient checks and balances in place to ensure that vested interests cannot misuse ITU
processes for their particular interests."11 All member governments and any interested private company
(including registries, registrars, Internet service providers (ISPs), and equipment suppliers) can participate
in ITU-T's work, but unless there is general agreement, only governments can vote and only governments
can be represented in ITU's council, which has overall policy and strategy responsibility between ITU
plenipotentiary conferences.

Because the ITU is an intergovernmental organization, it has sovereign immunity, which obviates
the need for liability insurance or worry about liability affecting its decisions. Since concerns about the
legal liability it might be assuming appear to have affected ICANN's willingness to, for example, enter
into agreements with some ccTLDs that wanted to hold it to predefined performance standards, and could
affect its ability to take on other roles, this advantage of an intergovernmental organization has some
weight. Conversely, however, it may make the ITU a less attractive alternative to those ccTLDs and
others that want to be able to hold a governing organization liable if it fails to perform its functions

The ITU members voted at its 2002 Plenipotentiary Conference in Morocco that it should play an
active role in "discussions and initiatives" related to the DNS and IP address system with the goal of
becoming the forum in which public policy issues of Internet naming and numbering are resolved. In its
resolution from the meeting, the ITU identified "stability, security, freedom of use, protection of
individual rights, sovereignty, competition rules, and equal access for all" as important public policy
issues.12 Since that meeting, the ITU has campaigned for a more active role in governance of the Internet
generally and the DNS specifically. Its sponsorship of the Workshop on Internet Governance in February
2004 and its consultation on forming the Working Group on Internet Governance in September 2004 are
indicative of its continuing active interest.13


The principal arguments in favor of moving stewardship of the DNS to the ITU are the broad
international participation in the ITU, both by governments and industry; its established processes for
policy making; its secure funding and sovereign immunity; and its long experience with management of
international telecommunications resources. Its international treaty status would, perhaps, also give it
greater influence over the ccTLDs and the root name server operators. At the same time, ITU-T cannot
make decisions affecting competing parties or take actions without agreement by its membership. If there
is a disagreement between governments, aside from attempting to mediate, ITU-T can do nothing until a
plenipotentiary conference acts to resolve the matter, since only it has the authority to make
recommendations to governments (which, however, can still ignore them). Furthermore, the ITU's charter
does not allow intervention in disputes within countries, such as might occur over the delegation or
redelegation of a ccTLD registry.

There are, moreover, objections to the ITU's assumption of the DNS stewardship role from the
Internet technical and user communities and, significantly, from the U.S. government.

The technical and user communities have a distrust of the ITU for its long record of opposing the
Transmission Control Protocol/Internet Protocol (TCP/IP) and for its unwillingness to make its standards

numbering resource and the E.212 mobile numbering resource." See Houlin Zhao. 2002. ITU-T and ICANN Reform.
Telecommunication Standards Bureau, ITU, April 17, pp. 3-4. Available at <
11 Zhao, 2002, ITU-T and ICANN Reform.
12 ITU Resolution 102. 2002. Management of Internet Domain Names and Addresses. Marrakesh, Morocco.
Available at <>. Also see Kevin Delaney. 2002. "Global
Organization Seeks Voice in Internet Addressing System." Wall Street Journal, October 21, p. B4.
13 See materials at <>.

Prepublication Copy--Subject to Further Editorial Correction
publicly available for use on the Internet. Additionally, they fear that the ITU's bureaucracy and its
structure and processes, which require that governments sign off on decisions, will lead to extremely slow
decision making unable to keep pace with the requirements of managing the DNS. Many also fear that the
large telecommunications companies, which have a long-standing interest in and presence at the ITU, will
dominate the processes. It has generally been difficult for individuals and public interest groups to
participate in the ITU's processes, although the ITU has expressed a commitment to change. Although it
engaged individuals and groups in the meeting on Internet governance that it sponsored in association
with the WSIS,14 it excluded non-governmental organizations from the decision-making sessions and it
has not moved to change the ITU's charter or ITU-T's internal rules to permit their active participation.

The user communities are also concerned that the ITU would become a vehicle for governments
to exercise control over the registration of domain names and would use that control to enforce other
policies such as the local taxation of Internet commerce, intellectual property protection, and the
restriction of free speech and rights to privacy.

From what is known of current U.S. government attitudes, both in the executive and legislative
branches, transfer of DNS stweardship to an intergovernmental organization is not likely to be supported
now or in the near future, although such attitudes can change.

Conclusion: Despite the interest of the ITU and some of its national members in its assuming
stewardship of the DNS root, that does not appear to be a realistic alternative for the near future.

Alternative B: International Treaty Organization

Description and Evaluation

The negotiation of an international treaty to establish a special agency of the United Nations for
Internet governance is also not likely to be supported by the U.S. government. The example of other such
negotiations has shown that they take many years to complete, especially if governance or binding
regulation are contemplated among its authority, rather than just coordination or standardization, and are
unlikely to succeed in the absence of strong and sustained efforts by many nations, which despite the
currently expressed concerns does not seem likely. So this alternative does not appear to be realistic in the
near term either.

Conclusion: Although it is possible that the U.N.-sponsored 2005 World Summit on the
Information Society will lead to proposals for some form of internationally negotiated, quasi-
governmental or multi-stakeholder organization with oversight or other influence over DNS governance,
specific proposals are not yet (in December 2004) on the table and cannot be evaluated for either their
practicality or their feasibility.

This leaves the third alternative, which follows.

Alternative C: Private Organization with International Participation


The only institution to which the U.S. government has expressed its willingness to transfer its
stewardship of the DNS is ICANN, or a similar private, non-profit successor. Furthermore, this transfer of
stewardship could occur only after ICANN (or a successor) had established its ability to operate
effectively, reliably, and with wide international and constituency support for a number of years. Indeed,

14 See a list of participants at <>.

Prepublication Copy--Subject to Further Editorial Correction
in its statement regarding the 1-year extension of its MoU with ICANN in September 2002, the DOC

The Department of Commerce (Department) continues to support the goal of a private sector
management of the Internet domain name system (DNS). Innovation, expanded services, broader
participation, and lower prices will arise most easily in a market-driven arena, not in an
environment that operates under substantial regulation. To this end, the Department has long
maintained that a private sector organization is best able to respond nimbly to DNS issues in the
rapidly evolving Internet space. Further, in order to garner international respect and function
stably and soundly in the long-term, such an organization must be globally and functionally
representative, operate on the basis of open and transparent processes, and possess robust,
professional management.15

In September 2003, the DOC granted ICANN a 3-year extension of its MoU and indicated that if
the agreement's milestones are met, it is prepared to complete the transition of DNS management to the
private sector at the end of that period.16 More recently, in conjunction with ICANN's July 2004 meeting,
the Assistant Secretary of Commerce for Communications and Information issued a statement expressing
pleasure that "ICANN has timely met the MoU milestones to date. Clearly more work remains to be done
for ICANN to achieve functional, sustainable independence. We look forward to continuing to work
collaboratively with ICANN . . . as we complete the transition to independent, private sector management
of the Internet Domain Name System."17


Although these statements indicate the U.S. government's willingness to give up the ultimate
stewardship of the root, they also demonstrate its unwillingness to do so until it is assured that ICANN
can manage the DNS "in a manner that promotes stability and security, competition, coordination, and
The stewardship role of the DOC, while a matter of political concern to some nations, has not
impeded ICANN's governance role, with the important exception of sometimes substantial delays in
approvals for routine changes in the root zone file, a situation that improved during 2004.19 (See
"Approving the Root Zone File" in Section 3.3.3). For example, the DOC has not overridden any ICANN
recommendations for reasons of U.S. national interest. The issues that have arisen about the governance

15 "Department of Commerce Statement Regarding Extension of Memorandum of Understanding with ICANN,"
September 19, 2002. Available at
16 See amendment 6 to ICANN/DOC Memorandum of Understanding, September 16, 2003, available at
<>, and "Department of Commerce Statement
Regarding Extension of Memorandum of Understanding with ICANN," September 16, 2003, available at
17 "Statement by Assistant Secretary Michael D. Gallagher on ICANN's July Meeting in Kuala Lumpur," press
release, U.S. Department of Commerce, National Telecommunications and Information Administration. July 19,
18"Department of Commerce Statement Regarding Extension of Memorandum of Understanding with ICANN", op.
cit. p. 2.
19 There has been dissatisfaction among the ccTLD managers over the length of time required by ICANN and the
DOC to approve routine changes to the root zone file. In response, ICANN prepared a revised process designed to
cut the time substantially and to keep requesters informed of progress in the approval process. See Internet
Assigned Numbers Authority (IANA). 2003. "Procedures for Handling Requests by ccTLD Managers to Change
Nameservers." May 13. Available at <>.
These revised procedures and other changes have significantly reduced the time taken to make changes in the root
zone file as reported by the general manager of ICANN/IANA at the Kuala Lumpur meeting of ICANN. His full
report, including specific data, is available at <>.

Prepublication Copy--Subject to Further Editorial Correction
of the root have, rather, concerned the way in which ICANN operates in preparing its recommendations to
the DOC and not in the way that the DOC operates once it receives those recommendations. In that
respect, the decision to establish an organization to take on the day-to-day administration of the root has
successfully reduced those pressures on the U.S. government while, at the same time, preserving its
ultimate stewardship.
The Internet in general and the DNS in particular have been developed and governed with the
goal of technically enabling equitable access to all locations on the Internet to users anywhere on the
globe. To best serve the worldwide Internet and the DNS, the U.S. government's influence needs to
continue to be exercised carefully and, in particular, this influence should not be used as an instrument of
U.S. domestic or foreign policy in areas far removed from the Internet.

Conclusion: Governance of the DNS is not an appropriate venue for the playing out of national
political interests.

The technical legacy of the DNS's development and initial implementation in the United States,
such as the use of the ASCII character set for domain names and the concentration of root servers in the
United States, has been a source of concern to some countries. Such concerns have been or are in the
process of being addressed and they should, consequently, be substantially reduced or eliminated.

Although there may be continued objection by some nations to the right of the U.S. government
to exercise its role,20 the United States has committed itself to a contractual 3-year transition to ICANN
under the conditions laid out in the amended MoU. The real question is whether ICANN will be able to
gain full legitimacy in the perception of other national governments and constituencies during this period.

Conclusion: The continued evolution of ICANN to attain legitimacy among its critical
constituencies and, consequently, to receive stewardship responsibility from the U.S. government appears
to be the most feasible path to governance of the DNS that is broadly accepted as international.

The prospects of ICANN's assuming stewardship of the DNS are addressed in the discussion of
ICANN's role that follows.


Issue: What changes, if any, are required in ICANN's organization and management for it to
achieve greater legitimacy?

When the DOC delegated responsibility for day-to-day management of the root to ICANN in
1998, it was with the expectation that ICANN would soon be perceived as a legitimate steward of the root
as well. Yet, although ICANN is not a part of the U.S. government and its board has had an international
membership, its legitimacy was not immediately accepted.21 The critics' concerns have been with (1)
ICANN's scope, (2) its organizational structure, (3) or its management processes, or with all three. The
concern about scope has been the extent to which ICANN has exceeded its specific technical-
administrative responsibilities to, for example, regulate TLD registry operations. The concerns about
structure have included a perceived imbalance in the historical composition of ICANN's board, failings in
the processes by which the board was selected, and the inadequate representation of certain constituency

20 In the recent U.N.-sponsored forums on Internet governance, a few nations have expressed a concern that their
ccTLDs could be removed from the root by the U.S. government for political reasons.
21 See, for example, Jonathan Weinberg. 2000. "ICANN and the Problem of Legitimacy." Duke Law Journal. Vol.
50. pp. 187-257.

Prepublication Copy--Subject to Further Editorial Correction
groups. The concerns about management processes have included the lack of transparency, effectiveness,
accountability, and recourse in ICANN's electoral and decision processes.

5.2.1 Scope of ICANN's Authority

The first controversy affecting ICANN's perceived legitimacy is its appropriate scope--the
extent of its responsibilities and authority. On the one hand, there are those who believe that ICANN
should stick as closely as possible to its technical-administrative charter, eschewing responsibility for
seemingly related matters that require it to make value-laden judgments that have political, commercial,
or social effects.22 Such matters include the administrative approval of new gTLDs, regulation of the
business practices of gTLD registries, and the delegation or redelegation of ccTLD registries. On the
other hand, ICANN currently engages in each of those activities as a result of decisions taken by its staff
and board during its early years. ICANN may have assumed those decision responsibilities because there
was no other organization able to take them on, because they believed that there was no non-judgmental
way of resolving issues that ICANN confronted in connection with its technical-administrative
responsibilities, or because of the views and aspirations of the board and staff.

There should be a connection among the breadth of ICANN's activities, the pressure on it to
broaden membership on its board and engage more constituencies in its decisions, and the acceptance of
its legitimacy by various constituency groups. Were ICANN able, for example, to narrow its scope
primarily to the IANA administrative functions, to use economic processes such as auctions or lotteries to
allocate TLDs, and to delegate to third parties the politically sensitive decisions such as redelegation of
ccTLDs, then the pressures might ease, and acceptance of ICANN's legitimacy as an essentially
administrative body might become easier. However, in actual implementation, even economic processes
require some potentially sensitive decisions to be taken (see Section 5.4.2); delegation entails the question
of delegation to whom; and apparently routine administrative processes, such as contracting with a gTLD
registry, entail judgments about what provisions--if any--are necessary to protect the technical integrity
of the DNS, safeguard the interests of registrants, and meet the needs of intellectual property owners
while preserving freedom of expression.

Conclusion: ICANN is more likely to achieve perceived legitimacy with a narrower scope rather
than with a broader one.

However, it may not be possible or desirable to narrow ICANN's scope to a purely technical-
administrative one because of the difficulty of performing such functions without making some
politically, socially, or economically sensitive judgments. Moreover, ICANN might very well attract
controversy simply because it would be the most visible organization with some, however limited,
authority over the DNS.

5.2.2 Composition of the ICANN Board

The second controversy concerns ICANN's organizational structure. If it is to be widely
perceived as legitimate, what should the composition of the ICANN board be and how should its
members be selected? Those questions have been the subject of dispute since ICANN was formed. To a
large extent, the dispute stems from different views of ICANN's proper scope.
Those who believe that ICANN is principally a technical-administrative body may favor selection
of the board by members of the technical-administrative community from among the members of that
community on the basis of expertise and experience--a traditional process in that community. In contrast,

22 See, for example, Center for Democracy and Technology. 2004. ICANN and Internet Governance: Getting Back
to Basics.
July. Available at <>.

Prepublication Copy--Subject to Further Editorial Correction
those who see ICANN as a major element of Internet governance may favor a board comprising
representatives of the broad Internet user community chosen, so as to achieve legitimacy, through an open
electoral process--Internet democracy. Those who see ICANN's scope as lying between these two views
might favor a board representative of various constituencies selected through some combination of
constituency elections and peer selections. Finally, there are the governments mentioned above that might
feel that because the Internet (and the DNS) has become a central element of the global communications
infrastructure, it requires governmental oversight and regulation and, therefore, ICANN's board should
consist only or principally of governmental representatives.23
Much of the debate about ICANN's board stems from the differences among these perspectives.
In its first days, the 18-member ICANN board was selected without participation by many
Internet constituencies, despite a DOC-imposed requirement in its bylaws that half the board be elected
by a membership. The board's composition quickly became a source of controversy as ICANN addressed
issues of broad significance and social-economic-political consequence, such as the addition of new
gTLDs and the development of the UDRP. This put ICANN under pressure to implement the
requirement for democratically chosen representation of the broad Internet community on the board. That
pressure led it to experiment with a process of open Internet voting for five (the DOC-required one-half
would have been nine) regionally representative members of the board. That experiment, in turn, raised
questions among some ICANN participants about the legitimacy of an open Internet vote, which they
charged was vulnerable to capture by intensely interested, but not necessarily representative, groups and
open to possible fraud. Others felt that it brought onto the board new members whose views were more
representative of those of the broader Internet community than were the views of the previously appointed
members.24 Recently, ICANN has moved away from that model of a partially elected board to one that
more closely follows the in-between case described above. The 15-member board in 2004 comprises six
representatives selected by the three major constituency groups (the Generic Names Supporting
Organization--GNSO; the Country Code Names Supporting Organization--ccNSO; and the Address
Supporting Organization--ASO); eight others chosen by a board-selected nominating committee, which
includes representatives of ICANN stakeholders, from nominations made by constituency groups and the
public; and the ICANN president, serving ex officio.
The legitimacy of the ICANN board has also been undercut by the perception by some that the
selection of its initial members was severely flawed. Critics charge that it was done top-down by
ICANN's managers and did not represent a diversity of views. In their judgment, board members lacked
political experience and ties to constituencies and were, therefore, ill-prepared to supervise the managers
who selected them. Most of the members of that initial board remained in place longer than anticipated,
during which time many of the key policies and processes of ICANN were put in place. These critics see
the selection, composition, and actions of the initial board as having weakened ICANN's claims to
Since disagreements about ICANN's board composition and selection appear to arise in some
measure from different views of ICANN's scope, they are likely to be difficult to resolve unless and until
there is broader agreement on that scope. In the more likely situation, where there is not agreement among
all parties on its scope, ICANN will probably continue to be subject to some criticisms of its legitimacy.
As in all political processes, ICANN's goal will be to ensure that the most important constituencies are
not among the critics. In particular, it will have to convince the Department of Commerce that it has

23 Since those countries have principally been calling for the ITU or another U.N. agency to assume DOC/ICANN
functions, they have not at the time of this writing made this suggestion. However, should they determine that those
options are not feasible, they might turn to this view.
24 A recent academic study of ICANN's experiment in "running a representative and open corporate decision-
making process" judged it to have "largely failed." See John G. Palfrey, Jr. 2004. "The End of the Experiment: How
ICANN's Foray into Global Internet Democracy Failed." Research Publication No.2004-2, Beckman Center for
Internet & Society. Available at <>.

Prepublication Copy--Subject to Further Editorial Correction
achieved legitimacy among the most important constituencies in order to fulfill the terms of its most
recent MoU with the DOC.
It should also be noted that even if there were agreement on ICANN's proper scope, ICANN
would still face numerous practical problems of representation. What would it take, for example, for it to
be seen as truly international? Does it suffice, for example, simply to have one board member from each
region? Or should the distribution of board membership better reflect Internet usage in each country or
region? Or should it, perhaps, reflect the geographic distribution of registered domain names? This
appears to be an inherently difficult problem to which any proposed solution may incur objections from
some of those nations that are not directly represented. (Even direct representation may not suffice. Two
of the countries that have been most critical of ICANN in international forums are Brazil and China, both
of which have nationals on the ICANN board.)
In a similar way, the broadest view of ICANN's appropriate scope implies that the board reflects
the many Internet constituencies having an interest in its decisions. But there is no agreed list of those
constituencies, nor would the list be likely to remain unchanged. Even assuming that a list could be
agreed upon, should one board member be selected from each constituency despite the fact that
constituencies are of different size and degrees of importance? And how should each constituency make
its selection? In the event that elections are used: Who is the electorate? How are they to be reached?
How can the validity of their votes be assured?
These difficulties are not matters of ICANN's making. They arise from the unique situation in
which it finds itself and would confront any other non-governmental organization attempting to manage
an economically, socially, and politically significant component of the global infrastructure.

Conclusion: No composition of the ICANN board, no matter how arrived at, is likely by itself to
confer the perception of legitimacy on ICANN among all its possible constituency groups.

5.2.3 Nature of ICANN's Management Processes

The third controversy concerns ICANN's management processes. Perhaps the goal of perceived
legitimacy could be achieved through refining the processes by which ICANN carries out its work and
makes its decisions. From its inception, ICANN has indicated that it would use bottom-up, consensus-
based, and transparent decision processes. The aspiration for such processes largely reflected the culture
of the early Internet community, especially the Internet Engineering Task Force (IETF). It fit well with
that small, relatively homogeneous group in which technical expertise and experimental results drove
most decisions. However, it has not turned out to be as effective a tool for making the value-laden
decisions that have faced ICANN in an environment with a very large number of users, many of whom
are not technical experts, holding highly diverse and often fundamentally different goals and values that
are often not susceptible to resolution through consensus.

Although ICANN put in place a formal structure of constituency groups with apparent input into
the board's decision processes, prior to the recent reform, at least, it had not succeeded in employing them
in such a way that they were perceived to identify consensus views, or even to ensure that all constituency
opinions were heard during the process. In part, this was the result of failure of the constituency groups to
participate, but there were also questions about the weight given to some constituency groups and the
absence of others.

Thus, although ICANN has made efforts to fulfill its promises using the current processes, it
appears doubtful that--as long as ICANN is more than a purely technical-administrative body--any set of
processes that was both efficient and effective could be restricted to bottom-up, consensus decision
making among imperfectly defined constituency groups.

ICANN's existing processes have also been heavily criticized for their lack of transparency, for
the failure to document the logic of decisions, for the absence of a process of appeal, and for the heavy

Prepublication Copy--Subject to Further Editorial Correction
reliance on non-accountable staff and consultants. Many ICANN observers view accountability as an
"essential component of legitimacy for ICANN."25

Conclusion: Improvement of ICANN's processes appears to be a necessary step toward
strengthening its perceived legitimacy.

For the reasons noted above, it would be difficult to strengthen ICANN's perceived legitimacy if
the focus on bottom-up, consensus decision making were retained. It would be more practical to
concentrate on making conventional majority-vote decisions through processes that are accepted by its
constituencies as being open to input from all those having a legitimate interest, transparent and
observable in all their stages, and fair to all participants.

5.2.4 Alternatives

In response to the criticisms of its structure and management practices, ICANN began
implementing a significant reform in early 2003. During 2002, while it was examining what specific steps
to take, a number of external groups put forward specific proposals of their own. Those proposals
demonstrate the diversity of the interests that attempt to influence ICANN and that remain active or
potential critics of its efforts. Furthermore, should the current ICANN reform not prove successful, those
proposals are likely to reappear in their original or a modified form. Thus, both to show the forces that
ICANN faces and to characterize the alternatives to its current reform, this section summarizes and
evaluates some of the major alternative paths that have been proposed for ICANN to achieve legitimacy
in the eyes of its critical constituencies.
Two distinctly different groups of approaches to ICANN's structure were proposed: broadening
and narrowing. The broadening approaches would keep or extend ICANN's scope and keep or broaden
the number of groups that participate in its processes on the board and through other means. In contrast,
the narrowing approaches would narrow both ICANN's scope and the diversity of stakeholders directly
involved. Each approach would also have consequences for the nature of the processes ICANN employed.
The alternatives described and evaluated below include two broadening proposals--one by the
Markle Foundation and the other by the Non-governmental Organization and Academic ICANN Study
group--and two narrowing proposals, one that ICANN serve solely as the registry for the root and
another that it be a private trade association. The discussion concludes with two proposals that combine a
narrowing of scope with a broadening of participation--a proposal by the Center for Democracy and
Technology, and the actual reform that ICANN adopted in 2003.

Alternative A: Markle Foundation Proposal (2002)


25 Center for Democracy and Technology. 2002. "Comments of the Center for Democracy and Technology to the
Committee on ICANN Evolution and Reform." May 3. Available at
<>. See also Tamar Frankel. 2002. "Accountability and Oversight of
the Internet Corporation for Assigned Names and Numbers (ICANN)," Report to the Markle Foundation. July 12.
Available at <>; and Center for Global Studies. 2002. "Enhancing
Legitimacy in the Internet Corporation for Assigned Names and Numbers: Accountable and Transparent
Governance Structures." Final Report to the Markle Foundation. September 18. Available at

Prepublication Copy--Subject to Further Editorial Correction
The president of the Markle Foundation, Zoë Baird, presented recommendations for improving
ICANN's credibility and legitimacy in a 2002 Foreign Affairs article.26 These recommendations entailed
changes to broaden participation in the board and to improve its processes.

With respect to the board, Baird said:

ICANN's credibility as a global manager of critical parts of the Internet's infrastructure depends
on the board's ability to ensure that all the various private and public interests are represented
[emphasis added]. Government involvement is one step toward providing public-interest
representation but is insufficient on its own. Only with truly broad representation on its board--
including non-profit organizations--can ICANN adequately address the crisis of legitimacy that
plagues it [emphasis added].27

Baird also recommended changes in ICANN's processes:

ICANN must take steps to bolster transparency and accountability. These steps should include
some kind of public oversight by politically accountable officials; development of due-process
principles and clear, publicly available procedures for the resolution of complaints; public
disclosure of its funding sources and budgets; staff and board members who are held accountable
to a clear set of professional norms and standards; open meetings; and documentation of the
rationale for ICANN's policy decisions and actions.28


Although pointing in attractive directions, the Markle proposal's recommendations about ICANN
board composition are, unfortunately, too general to confront and resolve the practical difficulties of their
implementation. For example: What are the groups that constitute "all" public and private interests that
should be represented on the board? How many different governments need to be on the board to
represent all public interests? How many non-profits need to be represented to cover all other public
interests? In sum, how many board members would be required to provide representation of all public
and private interests? How can the number be kept from becoming unwieldy?

The Markle proposal's recommendation on process, although still general, appear to lead more
directly to practical implementation. In the specific case of "public oversight by politically accountable
officials" it should be observed that, in fact, that is the role that the DOC currently plays and effects
through its MoU with ICANN. For reasons adduced before, it is not clear that the U.S. government
would be agreeable to sharing that oversight with officials of other governments. Even if it were, there is
the question, once again, of which governments and which public officials should play that role.

Alternative B: Non-governmental Organization and Academic ICANN Study Proposal (2001)

The Non-governmental Organization and Academic ICANN Study (NAIS) group described itself
as "a collaboration of experts from around the world, formed to explore public participation in ICANN
and the selection of At-Large Directors on ICANN's governing board."29 The NAIS group was self-
selected and directed but was funded by a grant from the Markle Foundation, whose president is Zoë

26 See Zoë Baird. 2002. "Governing the Internet: Engaging Government, Business, and Nonprofits." Foreign Affairs.
November/December. pp. 15-20.
27 Baird, 2002, "Governing the Internet."
28 Baird, 2002, "Governing the Internet."
29 Non-governmental Organization and Academic ICANN Study Group. 2001. "ICANN, Legitimacy, and the Public
Voice: Making Global Participation and Representation Work." August. Available at

Prepublication Copy--Subject to Further Editorial Correction
Baird, author of the Foreign Affairs article in which alternative A appeared. In its report, published in
August 2001, the NAIS group summarized its arguments for broad public participation in ICANN's

"The mission, character, and history of ICANN requires [sic] global public participation and
representation for its long-term legitimacy and stability."
"To the extent possible, the entire affected Internet community--from companies in the
business of providing DNS services, to domain name holders impacted by ICANN's rules, to individual
Internet users and consumers whose activities online could be shaped by ICANN's rules--should be
considered stakeholders in ICANN's activities."
"ICANN's existing supporting organization structures [in 2001], or representation by
governments, do not alone provide appropriate public participation."
"`At-Large' Participatory Structures and Representation on the Board are therefore essential
channels for broader stakeholder involvement and ICANN's legitimacy."

On the basis of these arguments, the NAIS group's report asserted two "overarching principles:
The public membership [of ICANN] should be given structure and the public membership should be
given representation." To fulfill those principles, it made six recommendations:

"ICANN should constitute a broad membership open to all who complete a relatively simple
registration process."
". . . the At-Large Membership should have internal structures that promote policy
deliberation, coalition building and information sharing among Members."
"The public voice in ICANN should be represented at the Board level through a number of
At-Large Directors equal to the number of Directors chosen by the Supporting Organizations."
"At-Large Directors should be chosen through direct election by the At-Large Membership.
Direct elections, while imperfect, are more likely to provide ICANN with global legitimacy than other
proposed options."
Specific processes should be followed to authenticate voters, to achieve both geographic and
global representation, and to refine election policies.
To ensure the public's voice in ICANN, it should develop structural constraints on Board
authority, create additional accountability mechanisms, and should pursue Supporting Organization
reform--structure, processes, and Board representation.

"Thus," the NAIS group's report concluded, "it is essential for ICANN to establish an inclusive,
open At-Large Membership, with a clear means to participate in the decision-making process and
substantial direct representation on the Board."

This is a clear and full expression of the view that ICANN's influence on the Internet is so broad
and important that the Internet's end users must have a strong role (equal to that of the supporting
organizations) in its governance. It avoids the question of determining which constituencies should be
represented by merging them all into a common "public" constituency comprising all who sign up and are
authenticated through the best available method. Representation would be by region, with some global
representatives as well. Voting would be done online, which has the advantages of speed, global
availability, and economy but has been criticized because of the potential for fraud, for capture or
disruption by a determined group, and for possibly wide national differences in participation. In addition,
the populations and Internet participation rates differ greatly among regions, suggesting that equal
representation may not be the best approach.

Prepublication Copy--Subject to Further Editorial Correction
While opening participation to such a broad potential membership would undoubtedly increase
the perception of ICANN's legitimacy among a number of its constituencies, open participation might
have the opposite effect on those constituencies that take a narrower view of ICANN's role and
stakeholders. This illustrates the point made in Section 5.2.2 that differing, perhaps irreconcilable, views
of ICANN's role lead to different views about the proper composition and selection of its board. (The
ICANN reform (alternative F) that is being implemented did create the At-Large Advisory Committee
(ALAC) with a non-voting liaison member on the board.)

Alternative C: ICANN as Registry for the Root30 (2004)


The opposite approach (to alternative B) to enhancing ICANN's perceived legitimacy would be
to focus ICANN on its primary role as registry for the root,31 with responsibility for reliability, security,
accuracy, and availability, and to design its governance and operations accordingly.

Under this approach, ICANN's governance would be simplified by narrowing its controlling
constituencies to four groups of primary stakeholders:

The gTLD and ccTLD registries that depend on the root to direct potential users to them,
The root name server operators that serve the root zone file,
The Internet service providers (ISPs) and intranets that rely on the root to enable them to do
look-ups on the TLDs, and
The technical community that defines protocols and standards affecting the root and its

Only these groups would participate directly in ICANN's governance. Since all of them are
directly affected by or directly affect the operation of the root, they would, in effect, constitute a self-
governing body. Following the tradition established by ICANN, these stakeholder groups could organize
themselves into supporting organizations or forums to consider issues of special interest to their
respective groups and to forward proposals to the overall board.

(The group of domain name registrants and the registrars that serve them are not directly included
in this alternative, although they have a more direct interest in the operation of the DNS than, say, the
much larger group of all Internet users. A variant of this alternative could include them, but it would
make voting and funding much more complex and difficult to balance. This alternative assumes that their
views will be heard either through the registries and ISPs or through direct presentation to the board.)

All other interested parties--international agencies, governments,32 private commercial and non-
commercial users and suppliers, registrars, domain name registrants, individual Internet users, and public
interest groups--would be considered secondary stakeholders that could influence ICANN through one of
the primary stakeholders, by testifying at hearings and board meetings, and by lobbying individually or
through private and public interest organizations.

30 This proposal has been constructed for discussion purposes only. It is not a recommendation by the committee. It
was created to illustrate how a narrowly focused ICANN might operate and be governed. It bears some resemblance
to, but differs in important ways from: Elliot Noss, Timothy M. Denton, and Ross Wm. Rader. 2002. "A New
Approach to ICANN Reform: The Heathrow Declaration." March 25. Available at
31 ICANN's other roles concerning the IP address space and protocols could be retained or turned over to a strictly
technical body.
32 National governments' interests in their ccTLDs would not be subject to this body, except when a conflict over
the operator of the ccTLD registry occurred.

Prepublication Copy--Subject to Further Editorial Correction

Although ICANN would remain a not-for-profit organization, the TLDs, ISPs, and intranets
would be required to pay a fee for listing in the root (if a TLD) or for access to the root zone file (if an
ISP, intranet, or other large user33). A portion of the funds collected would compensate the root name
service providers34 for their service and would subsidize the technical community to conduct testing and
validation activities. The funds would also cover the costs of the IANA function of ICANN and its other
registry activities.

The board's members would be elected for fixed terms by the four or five stakeholder groups.
The number of votes cast by each TLD or ISP would be proportional to its annual fee to ICANN, which
would be in some relationship to its demand for root service, for example, as measured by the number of
registrations for TLDs and the number of IP addresses for ISPs. The number of votes cast by each of the
root service providers and by each technical organization would be proportional to the annual payments it
received, which would be based on its requirements. (Note that the sum of the payment-based votes
would equal the sum of the fee-based votes less a number of votes proportional to the operating expenses
of ICANN itself. These could form a third category of votes that would be cast by the ICANN president.)
Tying board voting power to payments would provide an incentive both for payment and for participation,
particularly by the ccTLDs and root name server operators.

The board would operate like the board of a public agency, taking decisions based on its
collective judgment as informed by public hearings. On special issues, stakeholders could petition for a
stakeholder vote, similar to a shareholder vote in a commercial corporation.


The approach described in alternative C for narrowing ICANN's scope is in strong contrast to the
proposals for a broadly representative board that relies on bottom-up consensus-based processes. It
substitutes a board that is intended to reflect the interests and experience of the immediate providers and
users of root name service only. In doing so, it reconceives ICANN as a narrower, more technically
focused body whose decisions would be limited to those that affect the ability of the root to meet the
needs of those who have direct recourse to it. It would have the means to exercise authority over the root
service providers through its payments to them, while they would gain influence on its decisions through
their consequent voting power. Each TLD would pay a fee to be listed in the root, but the amount would
depend on the number of its direct registrants (not including registrants in the subdomains of its
registrants) and would be proportional to its voting power. Very large TLDs, such as .com, would have
a substantial number of votes, but not enough to overcome the sum of the votes of other TLDs, the root
name service providers, and the technical community. The most difficult issue would be imposing a fee
on the ISPs for access to the root zone file since there is no practical way to prevent access to those that
have not paid. The threat of exclusion from ICANN activities and decisions would be the primary
incentive for payment.

Because it runs counter to the design intent of ICANN, to the cultural traditions of the Internet,
and to ICANN's own recent reforms, the approach of focusing ICANN on its role as registry for the root
is unlikely to be adopted in the near future. However, it illustrates one possible model of a narrowly
focused ICANN and stands, therefore, as a clear contrast to the more expansive models described in
alternatives A, B, E, and F.

33 Small, non-commercial users would be exempt unless they wanted to participate in ICANN governance.
34 Full implementation of this alternative would probably require replacement of the U.S. government-operated root
name servers so that they could receive payments.

Prepublication Copy--Subject to Further Editorial Correction
Alternative D: Proposal--ICANN as a Private Trade Association (2002)


Another narrowly focused model for ICANN was proposed in 2002 by 35 Under this
proposal, ICANN would gain legitimacy by reconstituting itself as an international consensus-based trade
association for "parties interested in issues related to domain names, IP addresses and internet
protocols."36 The association would be the vehicle for developing and promulgating industry-wide
practices and policies. Because policies would be developed through industry consensus, its proponents
believe that these policies and practices would be more likely to be adopted voluntarily by the association
members. They point to the consensus process, for example, that led to film producers' widespread
adoption of the Motion Picture Association of America's film rating practices.

ICANN as a trade association would also be characterized by a reliance on market forces as the
"dominant factor in regulating conduct of persons buying, selling and using Internet-related products and
services."37 The forces of the market would be expected to induce entrepreneurs and companies to
introduce a wide range of innovative services that would succeed or fail based on users' experience with
them, crucially including their interoperability with other services. In other words, the proponents of this
alternative expect that many of the concerns that have driven ICANN to a regulatory model and, in
particular, the protection of a single authoritative root and the controlled pace of new gTLD entries would
be better handled by letting the market decide. These proponents believe that the market would reject any
innovation that harms the functioning of the DNS and the Internet.

Where regulation would be required to supplement industry agreements and market forces, this
alternative would rely on existing practices, using national, regional, and local governments or formal
intergovernmental treaties. In the proponents' view: "It is hubris to assume that there is something so
special about Internet naming issues that the domain name industry requires a unique form of government
that is different from all other industries."38

One consequence of adopting this alternative would be the possible proliferation of DNS root
zones, which might be created by countries unhappy with U.S. control of the "legacy root" or desirous of
one in their national language, or by corporations, like, that see business opportunities in offering
roots containing additional TLDs and operating under different policies. To alleviate the concerns of key
players, the proposal suggests that the United States retain policy control of the legacy root; that other
nations form a ccTLD association that would agree on the contents of the ccTLD component of the root;
and that the United States contract with that association to ensure inclusion of those entries in the legacy
root. Other root zone providers would decide whether to include the legacy root or the ccTLD entries in
their root. Finally, the workings of the market would determine whether any root zone, other than the
legacy root, would survive.


As with alternative C, the approach of ICANN as trade association moves away from the notion
of a broadly representative board for ICANN and replaces it with a smaller set of stakeholders that have a
direct interest in operating the root. However, it retains the basic consensus decision-making process and,
while not specifically limiting the domain of ICANN's action, does indicate a strong preference for
leaving decisions to the operation of market forces. Because of that preference, it loosens restrictions on
the technical system of the DNS (such as maintenance of a single unique root, regulation of TLD registry

35 (see Section 3.3.1) is a commercial organization that offers an alternative root and related search service.
It has an obvious self-interest in a proposal that calls for establishment of alternative roots.
36 2002. "A Proposal for More Realistic Domain Name Governance." March. p. 13. Available at
37, 2002, "A Proposal for More Realistic Domain Name Governance," p. 15.
38, 2002, "A Proposal for More Realistic Domain Name Governance," p. 16.

Prepublication Copy--Subject to Further Editorial Correction
services, and a relatively slow, controlled addition of new TLDs) that many in the technical community
have said are required to retain the Internet's stability, openness, and uniform accessibility.

While obviously serving the interests of its principal proponent in allowing alternative roots and,
thereby, opening the market for gTLDs, alternative D does suggest how a market-based, free-entry
management of the root might be designed. Like the preceding alternative, new.NET's proposal
challenges existing views, in this case those of the Internet technical community, about what constraints
are required to protect the DNS technical system. Although it is unlikely to be put into practice in the near
future, it is another model available for consideration should other reform efforts fail.

Alternative E: Center for Democracy and Technology Proposal--
Narrowed Scope with Broad Participation (2004)

In July 2004, the Center for Democracy and Technology published a report39 that called for both
narrowing the scope of ICANN's mission and broadening participation in its activities. At the same time,
it supported the consensus-based approach to decision making.
With regard to mission, it urged ICANN to:

"Reaffirm the extremely limited mission that [it] was created to accomplish . . . the technical
function of coordinating the assigning of names and numbers . . . and a few inextricably related policy
"Refrain from using [its] coordination role as leverage to engage in policymaking in broader
"Reassess and ensure its contracts with registries provide both the reality and appearance of a
limited approach to coordination of registry activities"; and
"Adopt an approach to coordination that seeks to minimize policy impacts."

With respect to participation in decision making, it urged ICANN to:

"Strengthen activities to engage diverse constituencies around the world in [its] decision-

With reference to decision processes, it urged ICANN to:

"Support the consensus-based approach to decision-making that was core to the original
concept under which [it] was created";
Make its decision making "transparent, predictable and open to broad global participation by
stakeholders, including users"; and
Make its "future policies binding only if they are supported by a demonstrable bottom-up
consensus among affected parties."

It also asserted that ICANN's "only real and legitimate power comes from voluntary [emphasis in
the original] contracts and certain other mutually acceptable relationships and agreements."

39 See Center for Democracy and Technology, 2004, ICANN and Internet Governance. All quotes in this section,
"Description," are from the report.

Prepublication Copy--Subject to Further Editorial Correction
Alternative E was published as a proposal in the context of the World Summit on the Information
Society to refute the arguments and correct the misperceptions of those who see ICANN as a "precedent
or justification" for international centralization of Internet governance. It does so by emphasizing the
limited scope of ICANN's original mission and its charter obligation to apply only policies arrived at by
"bottom-up consensus among affected parties."
Because it focuses on narrowing ICANN's scope and at the same time encourages ICANN to
broaden participation by diverse constituencies around the world and rely on consensus decision-making
policies, alternative E appears somewhat inconsistent. For it would seem that restraining ICANN's role to
the technical-administrative function (and a few tightly linked policy matters) would, at the same time,
reduce the number of interested constituencies and the range of policy issues requiring consensus decision
making, though it would make the latter more feasible. It is primarily a call for ICANN to "get back to
basics" and reverse the steps taken to set and apply non-consensus policies beyond ICANN's original
narrow scope.

Alternative F: Reformed ICANN--Narrowed Scope with Broad Participation (2003)

The issues facing ICANN in 2002 of board composition, constituency participation, funding, and
decision processes were brought into the open and given formal recognition when the then-president, M.
Stuart Lynn, published a critique of ICANN's operations.40 This led to the creation of a committee to
study and make recommendations on ICANN's governance. The committee's report was published in
October 200241 and its recommendations, after board consideration in November 2002,42 have been
The new ICANN bylaws, which took effect in 2003, have remade the governing structure of
ICANN, especially the board and the supporting organizations that represent its constituencies. They are
derived from a mission statement revised and sharpened in response to the widely held perception that
ICANN had suffered from a vague and undefined concept of its mission. The new mission statement is as

The Internet Corporation for Assigned Names and Numbers (ICANN) is the private-sector
body responsible for coordinating the global Internet's systems of unique identifiers.
The mission of ICANN is to coordinate the stable operation of the Internet's unique identifier
systems. In particular, ICANN:
1. Coordinates the allocation and assignment of three sets of unique identifiers of the
Internet--domain names, IP addresses and autonomous system (AS) numbers, and protocol
ports and parameter numbers.
2. Coordinates the operation and evolution of the DNS's root name server system.
3. Coordinates policy-development as reasonably and appropriately related to the
performance of these functions.43

40 M. Stuart Lynn. 2002. "President's Report: ICANN--the Case for Reform." February 24. Available at
41 Committee on ICANN Evolution and Reform. 2002. "Final Implementation Report and Recommendations."
October 2. Available at <>.
42 ICANN Board of Directors Meeting in Shanghai. 2002. "Preliminary Report." November 1. Available at
43 ICANN Evolution and Reform Committee (ERC). 2002. "ICANN: A Blueprint for Reform." June 20. Available
at <>, with revisions from the ERC. 2002.
"Final Implementation Report and Recommendations." October 2. Available at

Prepublication Copy--Subject to Further Editorial Correction
In conjunction with the mission statement, ICANN adopted 11 core values that ICANN will
adhere to, including preservation of Internet stability, reliability, security, and interoperability; limiting
activities to those benefiting from global scope; delegation where feasible; broad, informed participation;
dependence on market mechanisms where feasible; promotion of domain name registry competition; open
and transparent policy development; application of documented policies; speedy action; accountability;
and sensitivity to the public interest and related governmental concerns.

Under the new bylaws, the governing structure consists of a board of directors with 15 voting
members: 8 selected by a nominating committee; 6 selected by three supporting organizations44 (2 from
each); and the president of ICANN, ex officio. In addition, there are 6 non-voting liaisons to the board: 1
each from five advisory committees,45 and 1 from the Internet Architecture Board (IAB)/IETF. The
nominating committee is charged to select directors to ensure that in aggregate the board has functional,
geographic, and cultural diversity; the capacity to understand the global effects of ICANN's mission and
decisions; and credibility.

Each of the supporting organizations has its own internal structure, allowing for the
representation of multiple constituencies. Each is responsible for the development of policy
recommendations to the ICANN board and developing internal consensus. The Nominating Committee
comprises 11 members appointed by constituencies, 3 appointed by advisory committees, and 5
unaffiliated public interest persons appointed by the At-Large Advisory Committee. In addition, there are
2 non-voting liaisons from the other advisory committees and the Technical Liaison Group. The board
appoints its chair.

In a move away from the consensus-based, bottom-up process that guided ICANN initially, the
policy responsibility of the ICANN board has been strengthened:

The ICANN Board of Directors is ICANN's ultimate decision-making body. . . . It is ultimately
responsible for the management of the policy development process. Therefore, while it is highly
desirable to seek and wherever possible find consensus, it does not follow that even proposals that
enjoy consensus support should receive uncritical Board approval. The Board has a fiduciary
responsibility to make decisions on the basis of good faith judgment in furthering the public

In response to concerns about transparency and accountability, the new bylaws call for the
creation of an office of ombudsman, a manager of public participation to encourage full public
participation in ICANN, and a strengthening of the reconsideration process applicable to both the staff
and board and requiring prompt consideration. It also establishes an independent review process to review
whether the board has acted consistently with the bylaws.

To strengthen government participation without directly including government representatives on
the board, the bylaws call for the Governmental Advisory Committee to appoint a non-voting liaison to
the board, a delegate to the nominating committee, and non-voting liaisons to each of the supporting
organizations and to the other advisory committees.

44 The three supporting organizations are the Address Supporting Organization (ASO), which deals with the system
of IP addresses; the Country-Code Names Supporting Organization (ccNSO), which focuses on issues related to the
letter country-code top-level domains; and the Generic Names Supporting Organization (GNSO), which handles
issues related to the DNS and the generic top-level domains.
45 The four advisory committees are the At-Large Advisory Committee (ALAC) for the Internet community at-large;
the DNS Root Server System Advisory Committee (RSSAC) for root server operators; the Governmental Advisory
Committee (GAC) for governments; and the Security and Stability Advisory Committee (SSAC) for security. There
is also the Technical Liaison Group (TLG) for standards groups.
46 ICANN ERC, "ICANN: A Blueprint for Reform," 2002, and "Final Implementation Report and
Recommendations," 2002.

Prepublication Copy--Subject to Further Editorial Correction
In place of having five at-large board members elected by Internet users, that community is
expected to be represented through the At-Large Advisory Committee, which will seek to engage
individual Internet users through regional at-large organizations in five geographic regions.

The restated ICANN mission is to focus on its original mission--coordinating the stable
operation of the Internet's unique identifier systems. If the board adheres to that mission statement, the
effect could be to narrow ICANN's scope from that which has evolved since 1998. The reform, at the
same time, takes a broad view of ICANN's constituencies. The new bylaws define selection and
representation mechanisms that could widen the range and raise the quality of members on the board and,
thereby, strengthen ICANN's perceived legitimacy. However, some critics argue that the board member
selection mechanisms are vulnerable to capture by the board itself, leading to a narrowing of
representation. The structure attempts to respond to concerns about ICANN's processes by defining steps
to increase the specificity, transparency, and accountability of ICANN's processes.
Whether this new structure and these new processes will enable ICANN to achieve the perceived
legitimacy that it has so far failed to attain can be determined only through the practical experience of the
coming years.

Summary of the Alternatives

The six alternative approaches to achieving an ICANN that is perceived as the legitimate steward
of the root are summarized in Table 5.1. They are characterized along two dimensions. The first is the
breadth of ICANN's role--broad or focused. The second is the breadth of its community of
stakeholders--broad or narrow.

TABLE 5.1 Alternatives for Organization of ICANN to Achieve Legitimacy
Constituency Broad
A. Markle (2002)
E. CDT (2004)

B. NAIS (2001)
F. ICANN reform (2003)

Pre-reform ICANN


C. Registry for the root (2004)

D. Private trade association (2002)

5.2.5 Conclusions and Recommendations

The discussion of alternatives in Sections 5.1 and 5.2 leads to four observations. First, the
Department of Commerce has expressed the U.S. government's intention to complete privatization of
DNS governance by transferring its stewardship role to ICANN by 2006, conditional upon ICANN's
satisfying certain pre-conditions. Second, the 2005 WSIS meeting might produce a proposal for a private,
non-governmental organization designed to assume certain Internet governance responsibilities. However,
as of March 2005, whether such a proposal would in fact be presented and what form it would take were
not known. Third, evaluation of any organization proposed by WSIS in late 2005 as a DNS steward
would face practical difficulties: It could be done only on the basis of the proposed organization's design,

Prepublication Copy--Subject to Further Editorial Correction
whereas ICANN will be evaluated on the basis of its record. And it might not be possible to complete an
evaluation before the intended transfer of stewardship to ICANN in 2006. Fourth, the outcome of
implementing ICANN's 2003 reform and other ICANN changes may or may not result in the fulfillment
of the DOC's requirements.
These observations suggest three conclusions.

If ICANN satisfies the Department of Commerce's requirements and is generally
perceived to be a legitimate manager of the DNS in the view of a substantial majority of its
constituencies, and if no preferable alternative results from the 2005 WSIS meeting, the U.S. government
is highly likely to transfer its role as steward of the DNS to ICANN during 2006.

Any private, non-governmental organization proposed as a result of the 2005 WSIS
meeting is likely to be considered by the Department of Commerce for DNS stewardship only if ICANN
fails to satisfy the DOC's requirements.

If ICANN's reforms are not successful and if the 2005 WSIS meeting does not
propose an organization to assume DNS stewardship that is acceptable to the U.S. government, the likely
outcomes will be a continuation of the Department of Commerce's stewardship role and a basic
reconsideration of how DNS governance should be organized.

A transfer of stewardship from the DOC will leave ICANN (and another organization if
stewardship is kept separate) without the benefits and controls that the DOC has provided. It
independently reviewed ICANN's recommended decisions, regularly oversaw ICANN's performance
subject to the sanction of non-renewal of its MoU, and implicitly protected it from attempts by other
governments and organizations to gain control of or strongly influence ICANN's decisions. If the DOC
does transfer its stewardship either to ICANN or to another private body, how will these benefits and
controls be provided?

Without additional protection, legitimacy based on the "consent of the governed"
would be the only basis for ICANN's continued authority and its ability to resist inappropriate pressure
from governments and other powerful interests. Without additional oversight, final responsibility for
satisfying the needs of its constituencies in an equitable, open, and efficient manner would lie solely with
its board.

Recommendation: Before completing the transfer of its stewardship to ICANN (or any other
organization), the Department of Commerce should seek ways to protect that organization from undue
commercial or governmental pressures and to provide some form of oversight of performance.


Issues: Is there a need for greater oversight of the root name server operators? If so, how might it
best be conducted? Should there be a formal process for replacing root name server operators? Should
the root name server operators be compensated for their service, and if so, how?

The root has functioned well as the shared responsibility of a group of 12 diverse, autonomous,
informally coordinated, and independently funded operators (see Table 3.1). Many observers believe that
the diversity and autonomy have been strengths, reducing vulnerability to single point failures. Yet, other
observers feel that the DNS and the Internet have become too important to the global society and
economy to permit such a crucial system to continue to operate without the oversight of an organization
responsible for its continued health--technical, operational, and financial.

Prepublication Copy--Subject to Further Editorial Correction
The tension between these two views--diverse autonomy versus central oversight--leads to
several different potential approaches to the issues identified above. In this section, the current situation is
first reviewed and evaluated and then four alternatives are similarly described and reviewed. The
committee's conclusions and recommendations follow a comparison of the four alternatives.

5.3.1 Current Situation: Diverse Autonomy


The effective daily operation of the root name servers, described in "The Root Name Servers" in
Section 3.3.2, lies squarely in the hands of the root name server operators. The operating organizations
have taken on the responsibility voluntarily, are not compensated by users for their base operations
(although all are subsidized by their home institutions or outside contributors, and some receive payments
from operators of their anycast satellites47), and are self-regulating through extensive and continuous
intragroup coordination.
Although ICANN's MoU with the DOC48 assigns it the responsibility to manage the root name
server system, its authority, as noted in "Selecting the Root Name Server Operators" in Section 3.3.3, has
not been sufficient to enable it to manage or even regulate the root name server operators directly. The
recent extension of the MoU places emphasis on ICANN collaborating with the DOC on "operational
procedures for the root name server system, including formalization of relationships under which root
name servers throughout the world are operated and continuing to promote best practices used by the root
system operators." 49 To increase its authority, ICANN has established the DNS Root Server System
Advisory Committee,50 has sought to enter into formal agreements with each of the 11 other root name
server operators,51 and has prepared a draft "Memorandum of Understanding Concerning Root
Nameserver Operation."52 However, it has thus far been unable to complete an agreement with any of the
operators. This is not surprising since it is likely that the operators do not expect to receive benefits that
would compensate for the additional obligations they would be expected to assume to ICANN or, through
it, to the DOC. In addition, many operators are subsidiary organizations within larger U.S. and foreign
academic, commercial, or governmental entities, which themselves may not wish to incur the obligations
and assume the liabilities that would come with such an agreement.53 Indeed, it is difficult to see what
ICANN could currently do to induce the operators to sign the "Memorandum of Understanding
Concerning Root Nameserver Operation" in the absence of a very attractive quid pro quo. Thus, although
ICANN has been assigned the responsibility for "formalization of relationships" among the autonomous

47 See Box 3.1 for a discussion of anycast satellites of root name servers.
48 "Memorandum of Understanding (MOU) Between ICANN and U.S. Department of Commerce." November 12,
1998. Available at <>.
49 "Amendment 6 to MoU Between ICANN and U.S. Department of Commerce." September 16, 2003. Available at
<>. See also "Department of Commerce Statement
Regarding Extension of Memorandum of Understanding with ICANN." Available at
50 For a description of its responsibilities and activities see <>.
51 ICANN operates the L root name server itself.
52 The draft MoU is available at <>.
53 As the then-president of ICANN, Stuart Lynn, noted ". . . some organizations that sponsor a root name server
operator have little motivation to sign formal agreements [with ICANN], even in the form of the MoU that is now
contemplated. What do they gain in return, except perhaps unwanted visibility and the attendant possibility of
nuisance litigation? They receive no funding for their efforts, so why should they take on any contractual
commitments, however loose?" "President's Report: ICANN--the Case for Reform." February 24, 2002. Available
at <>.

Prepublication Copy--Subject to Further Editorial Correction
root name server operators, it currently has little ability to convince the operators to agree to a stronger
ICANN role.


Fortunately, the current operators have operated the system successfully and without major
incident up to now, despite the enormous and somewhat unanticipated rate of growth in demand for root
name service since the mid-1990s when usage of the Web took off, and despite at least one recent
malicious attack54 on the system. Recently, through the voluntary adoption of anycast technology (see
Box 3.1), the operators have effectively multiplied the number of root name servers severalfold, reducing
the vulnerability of the system to denial-of-service attacks, improving the global accessibility of the root,
and moderating the political pressure for relocation of the core 13 root name servers.

The root name server operators assumed their responsibilities and, with a modest number of
changes, have successfully operated within the context of Internet culture that favors informal, voluntary,
and non-bureaucratic institutions run by technical specialists with primarily altruistic motives. In
addition, advances in computer technology and the ready availability of free name server software have
kept the cost of operating root name servers relatively low,55 especially when run as an adjunct to other,
larger computer operations, as most of them are.
Nor would it be easy to change this situation without the agreement of the operators. Their
incumbency and financial independence protect their responsibility and authority, and their informal,
collegial relationships serve to strengthen their power as a group. Furthermore, although they do not
receive direct financial compensation for their service, they all receive intangible benefits that, together
with their fear of potentially destabilizing change in a system that is working well, are evidently sufficient
to support their continued activity. In the absence of strong justification, it appears that they are not
inclined to favor a change in the current state.
But despite that fact that the current arrangement is functioning well, it might run into difficulties
if for some reason the root name servers' performance deteriorated or some of the operators resigned.
Moreover, ICANN is obliged under the MoU to try to take a more formal role in root name server
operations. Consequently, it is useful to consider the range of alternatives there might be for operating the
root name server system and to see if they might offer advantages over the current arrangement, as well as
in meeting ICANN's MoU obligation.

5.3.2 Alternatives

Two alternative models for structuring management of the operations of the root are (1) funding
and regulation and (2) a competitive market. They are described and evaluated below. A third,
hypothetical possibility--distributing the root zone file--is here raised by the committee as a means of
opening a different kind of approach for consideration. A fourth alternative would be for the DOC to
release ICANN from its MoU requirement that it formalize relationships with the root name server

54 On October 21, 2002, a distributed denial-of-service attack was launched against the 13 DNS root name servers.
For more information on the impact of the attack, see <>.
55 According to discussions in June 2004 with Kurt Erik Lindqvist, managing director of Autonomica AB, which
operates, the average annual cost of operating an independent root server, including the costs of
multiple anycast satellite servers, is about $1 million. However, other root servers are operated as adjuncts to already
well-provisioned secure Internet sites, requiring a minimal incremental expenditure on the order of tens of thousands
of dollars annually.

Prepublication Copy--Subject to Further Editorial Correction
Alternative A: Funding and Regulation

Under alternative A, ICANN (or a successor organization) would acquire the means to assume
responsibilities for the root name server system. As noted above, the most likely source of authority for
ICANN--as a non-governmental body--would be financial. If funds were available to cover all or a
significant portion of the operating costs of the root name server operators, ICANN might be in a position
to use its financial authority to do some or all of the following regulatory tasks:

Establish minimal performance standards for the root name server operators.
Support the implementation of a real-time performance monitoring system for each operator
and the system as a whole, perhaps based on the system under development by the Reséaux IP Européens
Network Coordination Centre (RIPE NCC).56
Monitor the performance of the root name server system.
Enforce performance standards by adjusting compensation to the operators according to the
level at which they performed.
Require performance improvement of operators that do not achieve minimum levels of root
name server performance that would be enforced through cut off or reduction of compensation.
Remove operators whose performance does not meet minimum required levels despite
requests and reduced or eliminated payments.
Identify and qualify new operators to replace removed or resigned operators.
Use the contingent provision of additional funding to provide incentives for the operators to
achieve higher performance levels and introduce new services.

If it performed most or all of those tasks, ICANN would be the central overseer of the root name
server system with full responsibility and authority for its reliable and effective operation. It could
establish a performance-monitoring system and use financial rewards and punishments to maintain and
improve performance. Should change in an operator or in operations be required, ICANN would be in a
position to bring it about. ICANN would, in fact, be carrying out natural responsibilities of the registry for
the root (see alternative C in Section 5.2.4).
ICANN's capabilities as a monitor and decision maker would strongly influence the performance
of the root name server system. If it had capable staff, excellent system understanding, and a
knowledgeable board, it could make timely and effective decisions. However, should any of those
conditions not apply, its influence could in fact be detrimental to the performance of the root name server
Moreover, this alternative depends on the operators responding to financial incentives. Many
might, but it is clear that not all of them would. There is some possibility of a significant number of them
dropping out if the alternative were additional controls from ICANN (or anyone else), and a change in a
number of root server operators at the same time might be destabilizing. In addition, the dropouts could
form an alternative root, which could be further destabilizing.
The three U.S. government agencies that run name servers, for example, would probably not
legally be able to give up their autonomy in exchange for a financial inducement, especially from a
private organization. They would have to be replaced by non-governmental organizations, and that would
probably require the agreement of the U.S. government. The relevant elements of the U.S. government
might, in turn, be reluctant to agree for fear of thereby reducing overall stability.
Nor are the funds necessary for this approach available within ICANN's current budget. They
would have to be incorporated into a future budget and would require a corresponding increase in

56 See <>.

Prepublication Copy--Subject to Further Editorial Correction
ICANN's funding from registries and registrars. (This would be available under "Alternative C, ICANN
as Registry for the Root," in Section 5.2.4.) Another option would be to allocate to this purpose a portion
of any funds received from new TLDs through auction or other processes, as described in "What
Selection Process Should Be Used?" in Section 5.4.2. below.
An alternative proposal would be to charge ISPs and other users an annual fee for access to the
root, which is similar to what is done today by the root server operators when the cost of an anycast
satellite is shared by the ISPs and users that will benefit from it. (However, there is no practical way that
such a fee could be enforced, except through the voluntary agreement of the ISPs and other users.)

Alternative B: Competitive Market

The second alternative would be to create a competitive market for root service. To the
committee's knowledge, competitive service of the same root has not been formally proposed elsewhere.
However, since increasing competition in the provision of DNS services is one of the stated goals of the
DOC, the committee has created an example of what a straightforward approach to such competition
might mean for root name services.57 A less straightforward example, which is actually being
implemented, is also described.

B1: Competing Root Name Server Systems

In the most straightforward form of market for root name service, competition would entail
having two or more distinct groups offering access to the authoritative root zone file. To avoid the
confusion that could arise with multiple root zones, each of the operator groups would have to agree to
offer access to the same root zone file, which would be provided to each of them at the same time by the
authoritative source, say, VeriSign for ICANN.58 Every ISP, intranet, or other organization running a
full-service resolver on the Internet would have to contract with one (or more) of the competing root
server operators, which would charge them a monthly or yearly subscription fee that could be based on
volume of queries or, more practically, on how many IP addresses the organization had assigned to it. A
technical means of ensuring that only subscribers could gain access to an operator's root name servers
would be required.

Evaluation. The committee was unable to conceive a version of this full competitive market for root
name service that would be both technically and operationally feasible and beneficial enough to displace
the current system.

There are two major technical difficulties. First, the communication and computation overhead of
ensuring that only subscribers could gain access to the operators' servers could be a substantial and
undesirable load on the root name server system, with the servers spending significantly more effort
checking permissions rather than answering queries. Second, there is the difficulty that arises because the
content of the root zone file includes the IP addresses of the root name servers. If there is one
authoritative root zone file, then there is only one set of IP addresses that it can contain and, therefore,
only one set of root name servers. Even if these technical problems could be overcome, two major
operational problems would remain.
The first operational problem is that every system that contains a full-service resolver would need
access to the root zone. Thus, every laptop and home system so equipped (and there are many) would
require a contract with one of the competing root name server systems or would have to change to a stub

57 The recent sixth amendment to the MoU includes the statement, "The Department reaffirms its policy goal of
privatizing the technical management of the DNS in a manner that promotes stability and security, competition
[emphasis added], coordination, and representation." p. 2.
58 The authoritative root zone file is currently distributed by VeriSign, which operates the hidden primary. However,
at some future time, ICANN may take over this function itself.

Prepublication Copy--Subject to Further Editorial Correction
resolver and contract with an ISP for that service. This, in turn, would cause problems for traveling users
who typically encounter and would have to contract for root service from multiple different ISPs in the
course of their travels.
The second operational problem is that the switchover to operation of a competitive system would
have to take place in a short time, unless the current set of operators were the default choice, to which
competitors would likely object. That means that every ISP, organizational intranet, and owner of a full-
service resolver would have to sign up for service with one of the competitors. Technically, the root
name server Hints File in every resolver/name server on the Internet would have to be changed over in a
short time to list the IP addresses of the root name servers of one of the providers.
Even if these technical and operational difficulties could be overcome, there are two pragmatic
difficulties that the changeover to a full competitive market for root name service would have to

First, a fee would be introduced for what is currently a free good. That is only likely to be
accepted if it is accompanied by significantly improved performance on dimensions that its customers
care about and if the root zone file can be kept secure from all but subscribers. But there has been no
suggestion that dissatisfaction with current service is such that any level of "improved performance"
would be willingly paid for.

Second, the incumbent operators would have a great advantage, since they have the facilities, the
skilled staff, and the operational experience that newcomers would have to develop. However, that might
not be a problem for companies that already run large DNS name servers, such as Neulevel or UltraDNS,
or for national governments that wanted to establish their own system of local name servers.

These potential difficulties, which mean that the probable costs would significantly outweigh the
prospective benefits, effectively eliminate the approach of a full competitive market for root name
service. However, there is another approach that is already developing.

B2: Competing Providers of Anycast Servers

The current market for anycast satellite servers is a less direct form of competition for root name
service. The competitive providers are a number of the root name server operators who compete on price,
service level, and other features to attract customers, which are ISPs or others that want to have an
anycast satellite server nearby. The result of the development of this market has been a rapid
multiplication and broadened distribution of the number of servers of the root zone file.

Evaluation. Model B2 has shown itself to be a demonstrably feasible path to reducing the geographically
uneven distribution of response times and to increasing the reliability of the root name server system
without incurring the technical, operational, or practical problems of the model first described. Thus, it
provides most of the benefits of model B1, "Competing Root Name Server Systems," without its

Alternative C: Distributed Root Zone File


A third alternative, which might avoid the difficulties of regulation or competing root name
servers, would be for ISPs and organizational intranets to obtain copies of the root zone file from ICANN
(through VeriSign's distribution server) and make local name servers authoritative for the root zone.
Something like this is, in fact, done today through the caching of queries by ISPs and through the
possibility of an ISP obtaining an anycast satellite of one of the root name servers.59 However, in this
alternative, the intention would be to formalize this practice and have ICANN require every ISP or

59 In fact, the root zone file is available for download at <>, but there is no
encouragement to use it.

Prepublication Copy--Subject to Further Editorial Correction
intranet to subscribe to the authoritative root zone file. The current public root name servers could be
retained for use by small and less technically skilled ISPs and intranets, but would be scaled down to
serve the smaller load.


The wide distribution of the root zone file and root name servers would significantly enhance the
security of the root and the reliability of its operation, in much the same way that anycast satellite sites
already have done, but to an even greater extent. Furthermore, the effects of erroneous or malicious
queries would be limited to the ISP or intranet from which they originated, and the ISP would have a
strong incentive to find and eliminate their sources.

Some fear that widespread distribution of the root zone file and the probable large numbers of
poor local configurations and irregular updating of the local name servers would cause havoc at worst,
and poor local service at best. The concern is that widespread availability of the root zone file would
encourage the offering of alternative roots, from which some TLDs had been removed and/or to which
additional TLDs had been added. The latter problem can, however, be partially addressed through DNS
security extensions (DNSSEC; see Section 4.2), which would detect, although not prevent, unauthorized
changes to the authentic root zone file. What actions would be taken if a change is detected would have to
be agreed in advance and enforced through an agreement with the ISPs. The irregular updating problem
might be resolved through the use of a distributor-driven process like that currently used to update the
secondary root name servers.

If it could not be ensured that organizations (especially ISPs, which in turn have customers)
would keep their copies of the root zone file up-to-date and uniform, the approach of having a distributed
root zone file could cause a decrease in the integrity of the DNS, which relies on the root zone file being
consistent across the Internet.

Alternative D: DOC Relaxes MoU Requirement


The DOC could relax its requirements for ICANN if it accepted the success of the current
diverse, autonomous, self-regulating root name server system instead of viewing it as unsatisfactory
because it is subject to neither formal regulation nor the discipline of the market. In place of its
cooperation with the DOC on "operational procedures for the root name server system, including
formalization of relationships under which root name servers throughout the world are operated," ICANN
could be tasked simply to serve as a facilitator, if asked, of the voluntary cooperation among the root
name server operators.


The DOC's relaxation of its requirements for ICANN would, in effect, legitimate the current de
facto relationship between ICANN and the operators and relieve the pressure on ICANN to take on
authority and responsibility that the operators have not yet shown themselves willing to cede.

Summary of the Alternatives

The current state and the four alternatives for oversight of the root name servers are summarized
in Table 5.2. It compares them on two dimensions: first, whether the root name server operators are under
formal or informal oversight by ICANN, and second, whether there is one set of root name server
operators or multiple sets.


Prepublication Copy--Subject to Further Editorial Correction
TABLE 5.2 Alternatives for Root Name Server Oversight
Root Name Servers
One set of operators
A. Funding and regulation
(Current situation ­ diverse autonomy)
B2. Competing providers of anycast
D. DOC relaxes MoU requirement
Multiple operators
C. Distributed root zone file
B1. Competing root name servers

5.3.3 Conclusions and Recommendations

Conclusion: The effective daily operation of the root, and therefore of the DNS and the Internet,
lies squarely in the hands of the root name server operators. Although ICANN has been assigned
responsibility for the stability and security of the root name server system by the Department of
Commerce, its authority has not been sufficient for it to manage or even regulate the root name server
operators directly.

Conclusion: The committee commends the operators of the 13 root name servers for their
reliable and efficient provision of critical root name service as the Internet has undergone rapid growth in
the numbers of its users and providers.

Conclusion: The committee believes that greater oversight of the operators will not be necessary
so long as they operate effectively and reliably and continue to improve the root name system's reliability
and capability.

Conclusion: The committee believes that in the longer term it is desirable for there to be more
formal coordination of the operators and that ICANN would be the most appropriate organization to
assume the coordination role.

Recommendation: ICANN should work with the root name server operators to establish a
formal process for replacing operators that directly engages the remaining root name server operators.

For example, should one of the current operators withdraw, ICANN could convene the remaining
root name server operators as a selection committee to recommend a replacement operator to ICANN's
board. The board could after appropriate consideration (and after approval by the DOC or a future
stewardship organization, if any) then direct IANA to enter the address of the new operator in the root
zone file.

Conclusion: Any central source of funds to compensate all the root name server operators for
their services is likely to carry an unacceptable regulatory or control role for the funding organization and
reduce the diversity of support that is one of the strengths of the current arrangement.

Recommendation: The present independent funding arrangements for the root name servers are
advantageous and should continue, because the multiplicity of sources contributes to the resilience,
autonomy, and diversity of the root name server system.


Prepublication Copy--Subject to Further Editorial Correction


Issues: Can and should new gTLDs be added? If so, how many new gTLDs should be added, and
how fast? What types should they be, and how should they and their operators be selected?

ICANN has faced a demand for the addition of gTLDs since its establishment (see Section 2.7).
At that time, the demand for .com addresses was growing very rapidly. Some applicants that were
unable to obtain their preferred domain names within .com or .net called for the creation of new
TLDs in which they might register. Potential registry operators, seeing an opportunity for profit in the
rapidly growing market for domain names, added their strong voices to the cry for more TLDs.
Although the demand for domain names is not growing as rapidly as it did in those very early
days, there remains a strong interest in adding gTLDs to the root as well as a strong counterbalancing
interest in moderating such additions.
In response to the pressure to add new gTLDs and to a requirement in its 2003 MoU with the
DOC, ICANN agreed60 to deliver by September 2004 a comprehensive evaluation of:

"The potential impact of new gTLDs on the Internet root server system and Internet stability"
"Potential consumer benefits/costs associated with establishing a competitive environment for
TLD registries."

It also committed to:

"Creation and implementation of selection criteria for new and existing TLD registries,
including public explanation of the process, selection criteria, and the rationale for selection decisions"
"Recommendations from expert advisory panels, bodies, agencies or organizations regarding
economic, competition, trademark, and intellectual property issues."

To fulfill this commitment, ICANN at its October 2003 meeting launched a strategic initiative to
allow new generic top-level domains.61 The initiative comprises two stages intended "to move to the full
globalization of the market for top-level domains." One stage is a comprehensive evaluation that includes:

An assessment of technical standards to support multilingual TLDs,
An assessment of the introduction of competition into the TLD market and of possible
business models for the TLD manager­ICANN relationship,
A study of intellectual property issues involved in the introduction of new TLDs,
Reports on technical stability issues related to the introduction of new TLDs, and
A review of consumer protection issues.

These studies were carried out by independent outside organizations62 as well as by ICANN's Security
and Stability Advisory Committee.

60 Reported in: ICANN Board Resolutions in Carthage, Tunisia, October 31, 2003. Available at
61 ICANN announcement: "ICANN Launches Broad Strategic Initiative for New Generic Top-Level Domains.
October 31, 2003. Available at <>.
62 On August 31, 2004, ICANN published a report, prepared for it by Summit Strategies International, entitled
"Evaluation of the New gTLDs: Policy and Legal Issues." The report touches on the second and third topics in the

Prepublication Copy--Subject to Further Editorial Correction

The second stage was an expedited process that was intended to add a new group of gTLDs
before the end of 2004. In this process, each of these gTLDs would be sponsored by a non-profit
organization representing a specific community, whose members would be the only ones eligible to
register domain names in it.

Thus, the question of adding gTLDs is timely and open to careful examination.
Although ICANN has accepted the view that new gTLDs should be added and is employing one specific
means for doing so, it is useful for a full understanding of the issues involved to take one step back and
consider first the fundamental questions: Should new gTLDs be added to the root zone and, if so, how
many and how fast? If new gTLDs are to be added, what types should they be, and how should they and
their operators be selected?

5.4.1 Should New gTLDs Be Added? If So, How Many New gTLDs, and How Fast?

An increase in the number of gTLDs would have technical, operational, economic, and service
consequences that would affect domain name registrants, registries, registrars, and Internet users
generally. Thus, responsible decisions about gTLD additions should take into account the different
potential effects on the several constituencies. In contrast, the public discourse and controversy have often
been framed narrowly--sometimes considering only one type of effect on one constituency. Furthermore,
the arguments are often based on assumptions, rather than evidence, and the assertions of one side are
often vigorously disputed by the other side.
The issue is addressed here broadly in terms of effects as well as constituencies affected, but for
simplicity, the multidimensional arguments for and against new gTLDs are clustered into two groups: (1)
technical and operational performance issues and (2) user needs and economic issues. Where relevant
evidence is available to support or contradict an argument, it is reported.
(Note: The amount of space devoted to each argument does not reflect the committee's judgment
either of its importance or of its credibility. Rather, it is a consequence of the material available and the
space required to explain it.)

Technical and Operational Performance Issues

Arguments in Favor
As noted in Chapter 3, the root zone file is very small, comprising data for 258 TLDs and the 13
root name servers--just over 78 kilobytes in total. It is searched on the root name servers several billion
times per day. The committee did not find any purely technical reasons that the root name servers could
not provide the same level of response with a much larger root zone file. Indeed, the ability of the .com
name servers to respond to billions of queries a day against the .com zone file, with more than 30 million
entries, is a demonstration of the technical capacity that could be applied to the root zone, if necessary.
Nor are there any fixed limits in the design of the DNS on the size or the rate of addition of
domains at any level in the DNS hierarchy. Any such limits would arise from practical matters of
implementation and operation.
Moreover, additions have already been made to the root zone--both the seven gTLDs added in
2000 and the numerous ccTLDs added during the mid-1990s, with no noticeable degradation of root name
server performance.

list of studies for a comprehensive evaluation. It is available at <

Prepublication Copy--Subject to Further Editorial Correction
Arguments Against
However, there are operational and administrative issues that suggest practical constraints on, at
least, the rate of addition of new TLDs to the root zone file, and potentially as well, on its total size. Of
concern to some members of the technical community are the number and the rate of changes to the root
zone file--both at the time of creation of a TLD (as a new entry into the root zone file) and in support of
subsequent changes (as a modification of an existing entry to accommodate a changed IP address, for
example). Errors or corrupted entries in the root zone file pose a greater risk of harmful consequences for
the DNS and Internet than do analogous mistakes made elsewhere in the DNS hierarchy. An increase in
the number of root zone file updates increases the probability of inadvertent errors and makes it more
difficult to detect them in a timely manner.
Yet, as discussed in Chapter 3, the design of the DNS makes heavy use of address caching. This
means that errors at the root will not affect a significant number of users immediately, but rather will
gradually be disseminated as the caches time out. Errors at the root zone, while certainly undesirable,
should not have catastrophic consequences and should be able to be caught before they do much
significant damage.
The administrative procedures for approving additions of new gTLDs (see "Selecting New
TLDs" in Section 3.4.3) have been much more intensive and extended than those, for example, for adding
a new second-level domain to one of the gTLDs. This has been for technical reasons--to ensure that the
gTLD's name server operations will meet Internet standards; for consumer protection reasons--to ensure
that registrants in the new gTLD will have reasonable assurance of a competent, reliable, and continuing
service; and because of the need to deal with a variety of contending legal and commercial interests.
Should a high level of scrutiny be required for approving additions of all new gTLDs and the ongoing
workload increase as a consequence of the additional gTLDs, then the rate of buildup of an adequate
administrative staff would also set a bound on the rate of addition of new gTLDs.

Committee View
In light of these considerations, several members of the committee hold the view that an
extremely cautious approach should be taken toward additions to the root zone file. Their preference
would be for no additions at all. And they would certainly limit new gTLDs to those that can be shown to
meet an important, unsatisfied need. Furthermore, they think that a process for monitoring root zone
operations capable of detecting signs of degradation or instability and of acting to correct their causes
must accompany any process for regular addition of new gTLDs.
Taking into account those views, the committee agreed that if additions are to be made to the root
zone, it would be prudent to limit their rate. After balancing the various considerations and the differing
views of its members, the committee concluded that the addition of tens of new gTLDs per year for
several years would be unlikely to jeopardize the technical or operational stability of the DNS. The
committee accepted that additions at a faster rate would unacceptably increase the risk at present.
However, further refinement of the practices for making and distributing root zone file changes (as
discussed in "Maintaining the Root Zone File--Verisign" in Section 3.3.3), the addition of administrative
capacity to ICANN, and further closely monitored experience with gTLD additions could provide the
basis for larger-scale annual increases in the future, should the demand be shown to exist.

Conclusion: Considering technical and operational performance alone, the addition of tens of
gTLDs per year for several years would pose minimal risk to the stability of the root.

User Needs and Economic Issues

Arguments in Favor
A principal argument for adding new gTLDs is that only by opening the market for gTLD registry
services would a true identification of domain name registrant (user) needs be possible, because only then

Prepublication Copy--Subject to Further Editorial Correction
would entrepreneurs and other innovators have the opportunity to offer a range of gTLDs to potential
registrants. Two indicators of the potential demand for new gTLDs are the more than 3 million registrants
in .info, the million or so registrants in .biz, and the more than 100,000 registrants in alternate TLDs
offered by, which are readily accessible from only 25 percent of the Internet (see "Unique
Characteristics" in Section 3.3.1).
While the advocates of addition do not claim to know when or if the user demand for new gTLDs
would be completely filled, they accept the inevitability of some new gTLDs failing. In their view, such
failures would be a reasonable price to pay for clarifying and meeting user needs, many of which may be
latent--unrecognized even by the potential beneficiaries. Advocates of this position, however, generally
acknowledge that to protect domain name registrants in new gTLDs, registry contracts should require
zone file escrow and agreements with other registries to assume responsibility in case a gTLD were to

One question that arises in this context is whether new gTLDs will provide new users
opportunities to register their desired (second-level) domain names or whether gTLD proliferation will
simply result in existing users registering similar domain names in all gTLDs for which they are eligible.
In the latter case, for example, .biz would simply replicate .com, duplicating registration and
administrative costs without any associated benefits.63

At least as judged by the experience of the operation of the .biz domain, there is only a small
amount of duplication. To begin with, no more than a very small fraction of .com registrants have chosen
to register the same domain name in .biz, since in February 2005 there were more than 33 million names
registered in .com but just over 1 million names registered in .biz.64 Moreover, even when the same
name has been registered, the identity of the .biz and .com registrants is the same in only 25 percent of
those cases, according to an earlier study.65 Thus, the .biz gTLD has, in fact, offered a substantial
number of new registrants opportunities to register their desired domain names, with proportionately little
duplication of .com domain names.
A demand for additional gTLDs could arise from the implementation of international domain
names that use non-Roman scripts (see Section 4.4). Although much of the early activity has been in
registering second-level domain names in non-Roman scripts in existing TLDs, that usage could develop
into a demand for gTLDs in non-Roman scripts. This demand could be quite large, given that the number
of non-English-speaking users of the Internet is already very large and is growing rapidly.
A new gTLD (e.g., .health) might induce additional value-added services at the second level
(e.g., that would be more accessible than the currently possible third-level services
(e.g., ) or second-level services (e.g.,
Another reason for favoring the addition of gTLDs is economic. First, it would open the
opportunity for new participants to enter the registry services market, promoting both price and non-price
competition (which includes diversity in naming methods) among registries. The August 2004 ICANN
report on the policy and legal issues of new gTLDs noted that "the ICANN community now has several
registry operators, as opposed to just one provider, which is [sic] able to operate a global registry of
significant scale. . . . Today the three companies [VeriSign, Afilias, and NeuLevel] compete for the ability
to provide registry services to new and existing TLDs, both in the gTLD and ccTLD markets."66 Second,

63 Indeed, there may be significant costs in addition to those associated with registering in multiple domains as
cybersquatters expend resources in order to acquire names that they may later sell for prices that substantially
exceed the costs of registration.
64 See <>, accessed on June 22, 2004.
65See Jonathan Zittrain and Benjamin Edelman, "Survey of Usage of the .BIZ TLD,"
<>, accessed July 22, 2002. This estimate is based on a random sample of 823
.biz domain names, where the identities of registrants were matched using zip codes, name server second-level
domains, e-mail addresses, or some combination thereof.
66 Summit Strategies International, 2004, "Evaluation of New gTLDs," p. 111.

Prepublication Copy--Subject to Further Editorial Correction
it would encourage innovation. According to the report, "the new gTLDs have had an impact beyond their
size or market share . . . in terms of innovation."67
Furthermore, a steady and foreseeable supply of new gTLDs could reduce the incentive to
cybersquat by lowering the scarcity value of specific names and making it more costly to hoard names.
Thus, the potential victims of cybersquatters may save money should new gTLDs be added.
New gTLDs would also enable the establishment of additional restricted name spaces, whose
registrants would be authenticated by the registry to be members of a specific class and who might also
have to accept specified rules of conduct. These would be like the existing sites .edu (for accredited
institutions of higher education), .pro (for professionals), and .museum (for the museum community).
Possibilities include .kids (for kid-safe sites), .xxx (for adults-only sites), .mail (for non-spam
mailers), and .health (for qualified health care providers). In addition, many organizations, especially
medium and large ones, may want their own gTLDs, enabling their computer services to assume
responsibility directly for functions they currently must acquire and pay for from TLD registries. For
example, IBM might prefer to own and operate the .ibm TLD, rather than rely on the .com registry for
service for

Arguments Against
There are strongly held opinions against opening the market in gTLDs as well.68
On the question of user needs, opponents of gTLD addition contend that the existing selection of
gTLDs (and ccTLDs that act like them--see "Recharacterizing TLDs" in Section 3.4.1) have met
registrant needs quite well and are capable of continuing to do so. They perceive few signs that many
prospective domain name registrants have an unmet need for new gTLDs that could not be satisfied in
one of the available second-level domains. Whatever truly unmet needs there might be are not sufficient,
in their view, to justify even a small threat to the reliable operation of the DNS. The August 2004 ICANN
report on the policy and legal issues of new gTLDs notes that, despite the entry of the new gTLDs, .com,
.net, and .org still have more than 93 percent of the registrations in all gTLDs.69 It concludes that
".com remains the TLD of first choice for a majority of gTLD registrants, including new registrants"70
and that a "choice of TLDs has thus far been unable to overcome the advantages the .com TLD
enjoys."71 Nonetheless, the report concludes that "the new gTLDs presented registrants with significantly
greater choice, at least in terms of initial registration."72 Further, the report concludes that "there are
indications that a respectable number of the new gTLD registrations have attracted new users to the DNS,
and that these new registrations are being actively used."73
From the perspective of the user seeking a site through guessing or relying on a possibly faulty
memory, the addition of gTLDs increases the number of possible domain names to try. For example, is it
"" or "" or ""? This is already a problem, but it could
be exacerbated by the addition of new gTLDs.
There are economic counterarguments as well. Trademark holders, for example, fear that they
would face the possible need to defensively register (to preclude cybersquatting) in all unrestricted
gTLDs, incurring an expense that could be substantial, particularly for small organizations and
individuals. However, the August 2004 ICANN report concludes that "the start-up mechanisms proved

67 Summit Strategies International, 2004, "Evaluation of New gTLDs," p .111.
68 See, for example: Tim Berners-Lee. 2004. "New Top Level Domains Considered Harmful." April 30. World
Wide Web Consortium, Design Issues. Available at: <>.
69 Summit Strategies International, 2004, "Evaluation of New gTLDs," p. 96.
70 Summit Strategies International, 2004, "Evaluation of New gTLDs," p. 109.
71 Summit Strategies International, 2004, "Evaluation of New gTLDs," p. 110. Nonetheless, the report notes the
views of some members of the "non-commercial community" who believe that, although overcoming the advantages
of .com would be difficult, it would not be impossible.
72 Summit Strategies International, 2004, "Evaluation of New gTLDs," p. 97.
73 Summit Strategies International, 2004, "Evaluation of New gTLDs," p. 99.

Prepublication Copy--Subject to Further Editorial Correction
generally effective in protecting legitimate trademark owners against cybersquatting."74 At the same time,
the report notes "a contradiction between . . . trying to attract new users and uses to the DNS, and
allowing trademark holders to claim priority registration of the same names in new TLDs."75 And it
observes that the number of sunrise registrations76 in the new gTLDs that used such a process "has turned
out to be much smaller than anticipated."77
More generally, holders of existing domain names that have built a reputation could find the
value of that reputation reduced as similar sites proliferate. For example, the owner of a discount ticket
Web site might fear that some of his customers would become confused and unable to remember whether
it is "" or "" or "" that is
known for offering the cheapest tickets.
Furthermore, current registry services stand to lose some of their economic (scarcity) value as a
result of the entry of new TLD registry service providers. So trademark holders, the owners of existing
sites, and current registry services all fear possible additional costs or losses of value if new gTLDs are
Opponents are also concerned that each increment of new gTLDs could be followed by a period
of administrative instability, set off by a race to register potentially valuable domain names and the flurry
of administrative and dispute resolution activity that it induces.
In the extreme, if unlimited additions were permitted (say, for example, .ibm, .ge, .sony,
.siemens, and so on), the root zone could become comparable to the .com gTLD, which would seem
to have uncertain benefits, while threatening considerable upheaval in the technical operations of the DNS
and administrative processes of ICANN. However, the imposition of a restriction on the allowable
number and rate of additions, as suggested by technical and operational considerations, would alleviate
this risk.

Committee View
After taking into account the user needs and economic benefits arguments for and against
potential gTLD additions, the committee remained divided on the addition of new gTLDs. Some
members felt strongly that on balance those needs and benefits were not great enough to risk the technical
and operational stability and reliability of the DNS. Other members believed that within the limits of tens
of new gTLDs per year it was worthwhile to enable some form of market process to be used (with proper
safeguards, discussed below) to determine how extensive the need might be.

Conclusion: Neither the user needs and economic benefits arguments in favor of nor those
against additional gTLDs are conclusive.

In addition to the question of whether and how many gTLDs should be added is the question of
how often. Thus far, additions have occurred at an irregular and unpredictable pace: the initial group of 8,
7 more in 2000, and up to 10 more in 2004. That uncertainty makes it difficult for current and potential
gTLD registries to develop and operate according to reasonable business plans and has the effect of
overvaluing new gTLDs (because of the uncertainty of whether and when there will be any further
additions). A regular schedule would enable those who lose out one year to anticipate additional chances
in the next year, reducing the price they would be willing to pay (in a gTLD auction, for example) were
there no certainty about the addition of gTLDs in the near future.

74 Summit Strategies International, 2004, "Evaluation of New gTLDs," p. 79.
75 Summit Strategies International, 2004, "Evaluation of New gTLDs," p. 80.
76 See "Possible Remedies to Conflicts over Names" in Section 3.5.2.
77 Summit Strategies International, 2004, "Evaluation of New gTLDs," p. 81.

Prepublication Copy--Subject to Further Editorial Correction

Recommendation: If new gTLDs are added, they should be added on a regular schedule that
establishes the maximum number of gTLDs (on the order of tens per year) that could be added each time
and the interval between additions.

Recommendation: Since the effect of continuing increments of new gTLDs on the performance
and stability of the root zone is not known, and the consequences of reduced performance and instability
can be great, it would be prudent to accompany any process of addition with a process for monitoring and
identifying any technical or operational problems the new gTLDs may cause.

Recommendation: A mechanism to suspend the addition of gTLDs in the event that severe
technical or operational problems arise should accompany a schedule of additions. It should explicitly
specify who has the authority to suspend additions and under what conditions.

Recommendation: A neutral, disinterested party should conduct an evaluation of new gTLDs
approximately 1 or 2 years after each set of new gTLDs is operational to make recommendations for
improving the process for selecting and adding gTLDs.

As noted in the introduction to Section 5.4, ICANN has included as part of its strategic initiative
the need for an evaluation of the additions made in 2000. The committee believes that ICANN should
contract with neutral, disinterested parties to conduct such evaluations for every subsequent addition of

5.4.2 If New gTLDs Are to Be Added, What Types Should They Be, and How Should They and
Their Operators Be Selected?

If new gTLDs are to be added to the root zone, there remain the questions of deciding which
types of gTLDs should be added and which entities should be selected to operate the associated registries.

In some selection processes used in comparable circumstances, what should be added and who
should be their operators are considered completely separately.78 In other cases, what should be allocated
and to whom it should be assigned have been determined in the same proceeding. In the case of the DNS,
both types of approach have been used. When seven new gTLDs were authorized in 2000, ICANN's
selection process combined the choices of the types and the identities of the registries, so that a particular
entity was selected as a registry because ICANN favored the type of service it promised to provide. In
contrast, the decision as to which organization would operate .org upon its transfer from VeriSign was
independent of the legacy decision to have such a gTLD.
Whatever process is used to make such decisions, it should at a minimum satisfy four
fundamental criteria:

1. Fairness. The process should not favor an applicant or class of applicants over others.
2. Transparency. The reasons for the outcomes should be clear to all involved.
3. Efficiency. The process should not place a heavy burden on the applicants or the selection

78 In radio frequency management, this is referred to as the distinction between "allocation," which denotes setting
aside blocks of frequencies in specific geographic areas for a particular use, and "assignment," which denotes the
award of these frequencies to particular users. Of course, it may be possible to dispense with the determination of
the type of new entrant and to limit the process to determining which entrants to allow.

Prepublication Copy--Subject to Further Editorial Correction
4. Economy. The process should not impose undue costs on the applicants or the administrator
of the selection process.

The discussion that follows provides a context against which to evaluate ICANN's current round
of selections and the process that it is following.

Which Types of gTLDs Should Be Added?

There are three approaches to determining the types of gTLDs. They derive from different views
about the desirable structure of the gTLD name space.


The first approach holds that the gTLDs should adhere, as much as possible, to a taxonomic
structure. Its adherents believe that such a structure can assist Internet navigation by serving as a high-
order directory and that it can channel and bound the addition of gTLDs. In an ideal taxonomic structure,
each gTLD would be restricted to members of a specific class, there would be a gTLD for each
appropriate class of members, and each possible member would fit into one and only one class.

The current structure of gTLDs is already far from an ideal taxonomy. Although the names of the
three largest gTLDs, .com, .net, and .org, imply a restriction to commercial, networking, and non-
commercial registrants, respectively, in practice the registries have not enforced such restrictions.
Moreover, there is nothing to exclude a Web site having, say, both .org and .edu domain names, even
if both were strictly restricted to appropriate registrants. So the best that an adherent to this view could
hope for would be that each new gTLD would be restricted to a specific class whose membership could
be inferred from the domain name (e.g., names such as .travel, .xxx, .library, and . health).

The "taxonomists" favor an extension of the gTLDs only in the restricted form.79 In this case,
each new gTLD would need a registry willing and able to enforce the specific restrictions, which can be
arduous and severely limit the rate of registration. For example, the .pro registry, which was formed in
2000, began registration only in the spring of 2004. It is limiting its rollout to the rate at which it can gain
access to records of registered professionals so that it can identify those that it will permit to have a .pro

Although it is no longer feasible to impose a full taxonomy on the DNS, this approach can
strengthen its implicit directory function by ensuring that all new gTLD names clearly identify a restricted
class of second-level domains. This is the approach taken by ICANN in adding new gTLDs in 2004 as
described below.

If there are benefits to using a tightly controlled taxonomy, there is nothing to stop this from
occurring at the second level. For example, there could be a gTLD called .industry, with second-
level domains describing various instances of industries: travel.industry,
health.industry, automobile.industry, and so on.


The second approach favors allowing the processes of supply and demand to determine how
many and which domains are offered, first by the willingness of an operator to offer it (supply) and
second by the willingness of those desiring domain names to register in it (demand). Under this
approach, it would be up to the gTLD registry to decide whether or not it would restrict registrations to
members of a certain group.

79 See, for example: Business Constituency. 2002. "A Differentiated Expansion of the Names Space." ICANN
Position Paper. December. Available at <>.

Prepublication Copy--Subject to Further Editorial Correction

To the advocates of this approach, opening a new gTLD and then closing it down because of an
insufficient number of registrations would be an acceptable outcome, assuming (as noted above) that zone
file escrow and alternate registry provisions could be made to protect domain name holders for some
period of time after the closure.

In contrast to the taxonomic/restricted view, the market-determined approach holds that
attempting to structure or control the gTLD name space provides little navigational value in a world of
search engines and other navigational aids. Many also doubt the taxonomic/restricted approach's
practicality. They believe that a market-driven name space is more flexible and reflective of users'
desires than is adherence to a limiting structure imposed by some authority.


The third approach takes a position intermediate between the first two. It would add new gTLDs,
both restricted and not, based on the qualifications and justifications presented by those who propose to
run them, on a case-by-case basis. An administrative body, presumably ICANN, would exercise judgment
to select which new gTLDs were added and who would operate them. This is effectively the approach
taken by ICANN in 2000 when it added seven new gTLDs, selected from the more than 40 applicants
because ICANN favored the TLD name and the type of service proposed, as described in "New gTLDs"
in Section 3.4.3.

How Should the Operators of gTLDs Be Selected?

The necessity for selection arises when the number of candidates for a resource exceeds the
quantity of the resource that is available, when candidates' qualifications to receive the resource must be
validated, or both. Three common forms for the selection process in situations comparable to selection of
new gTLDs and their operators are comparative hearings, auctions, and lotteries.

Comparative Hearing
A comparative hearing is an administrative process in which would-be entrants attempt to
convince the decision body that they, and their proposed service, are qualified for selection in competition
with other candidates. It is the form that has been and is being used by ICANN. Comparative hearings
typically feature discretionary entry, merit assignment, and heavy regulation. Such hearings can take into
account a large number of factors, draw on a wide range of expertise, offer opportunities to learn from
experience, and enable judgment to be employed. Through use of a non-refundable application fee, such
as ICANN has required, they can reduce speculative applications and be self-funding.
There are downsides to comparative hearings, however, as they can be subject to capture by
interest groups. Furthermore, the real decision process may not be transparent, and thus may be perceived
as arbitrary and unfair, especially by the losers. Also, decision quality is highly dependent on staff and
decision maker competence, while the negotiation and oversight of contracts can be time-consuming and
expensive. In addition, depending on the nature of the process and the stakes involved, application costs
can be very high.

In the basic auction model, the highest bidder wins, although participation in the auction might be
limited to those meeting a minimum set of qualifications. The advantages of auctions are that they are fair
and transparent processes that reflect the economic value perceived by applicants, they may be designed
to self-fund the process, and they discourage an excessive number of speculative applications. However,
the tradeoffs with the auction model are that it tends to favor well-funded applicants (which could be
corporations or non-profits) and thus does not necessarily reflect societal value, only economic value.
In principle, auctions could be designed to choose winners on the basis of who can best minimize
prices (say, for example, the price of a domain name registration). However, in such cases, it may be

Prepublication Copy--Subject to Further Editorial Correction
impossible to prevent the winning bidder from later claiming (successfully) that changing conditions
justify a higher price.
To the extent that in a specific situation social welfare extends beyond economic considerations,
auctions, at least in their pure form, become less appealing.

Winners in the lottery selection model are determined through a random choice from all entrants,
each of which may be required to meet some minimum qualifications in order to participate and would (in
contrast to a fund-raising lottery) be restricted to one entry per entity. The biggest advantage of lotteries is
their transparency: they are fair and transparent, giving all entrants an equal chance regardless of means.
Lotteries can also be self-funding if a charge is made for each "ticket." Although the cost to a given
entrant may be quite small, lotteries can be expensive to society in the aggregate because they can attract
a very large number of participants.
Because it can be difficult to determine whether an entry to the lottery is from a distinct qualified
entity, a lottery can be gamed by one participant obtaining multiple entries under different "legitimate"
guises. Moreover, unless the qualification criteria are very restrictive and carefully vetted, participants
may not even have the financial or technical capacity to operate a service if they win, and the winners
may simply sell the service to another entity for a significantly greater price. In the latter case, the lottery
is turned into an auction, with the proceeds going to the lottery winner rather than the organization
holding the lottery.

Combination Methods
It is also possible to integrate two or three of the basic methods into a variety of combination
methods, some of which are described below.

What Selection Process Should Be Used?

There is a widely held view that the process employed by ICANN in 2000 to add new gTLDs was
faulty. There is much less agreement about how to improve it, although the elements from which an
alternative can be developed are described above. In July 2004, the Organization for Economic
Cooperation and Development (OECD) issued a report that describes and compares two of those
elements, comparative selection (hearing) and auctions, as means for allocation of gTLDs.80 What follows
is a description of several of the alternatives that have been specifically suggested, beginning with the
process used in 2004 by ICANN for the next round of additions.

Alternative A: Sponsored gTLDs Selected by Comparative Hearing (2004)

Description. ICANN is using a modified comparative hearing approach to select new sponsored gTLDs.
Each new gTLD will be restricted to registrants from a well-defined and limited community and managed
by a sponsoring organization with ongoing policy-formulation responsibility for the gTLD. The
sponsoring organization will select the registry operator and, to some degree, establish the roles played by
the registrars and their relationships to the registry operator.81

80 Working Party on Telecommunication and Information Services Policies. 2004. "Generic Top Level Domain
Names: Market Development and Allocation Issues." July 13. OECD Directorate for Science, Technology and
Industry. Available at <>.
81 For full details, see ICANN, "New sTLD Application. Part A. Explanatory Notes," December 15, 2003. Available
at <>.

Prepublication Copy--Subject to Further Editorial Correction

The request for proposal, evaluation, and selection processes are modifications of those used in
2000 to select seven new gTLDs. Non-refundable application fees of $45,000 have been charged to cover
the costs of the selection process.

The eventual registry agreement will be similar to those entered into by the .museum, .coop,
and .aero registries. However, the August 2004 report published by ICANN that evaluated the policy
and legal issues of new gTLDs concluded that "while it is understandable for ICANN to have wished to
err on the side of caution as it undertook gTLD expansion . . . the resulting legal framework is
cumbersome."82 It concludes that "the number and length of appendices [in agreements between gTLDs
and ICANN] could be reduced in a future round. A streamlined base agreement with perhaps a few
appendices could provide a more workable format that also preserves the critical elements of registry
performance and mandates compliance with ICANN policies."83

Selection is to be determined by the degree to which independent evaluators judge the applicant
to have met ICANN's requirements in four major categories, which are divided into 15 subsidiary

1. Sponsorship
a. Definition of sponsored TLD community
b. Evidence of support from the sponsoring organization
c. Appropriateness of the sponsoring organization and the policy formulation
d. Level of support from the community
2. Business Plan Information
a. Business plan
b. Financial model
3. Technical Standards
a. Evidence of ability to ensure stable registry operation
b. Evidence of ability to ensure that the registry conforms with best-practices
technical standards for registry operations
c. Evidence of a full range of registry services
d. Assurance of continuity of registry operation in the event of business failure
of the proposed registry
4. Community Value
a. Addition of new value to the Internet name space
b. Protection of the rights of others
c. Assurance of charter-compliant registrations and avoidance of abusive
registration practices
d. Assurance of adequate dispute-resolution mechanisms
e. Provision of ICANN-policy-compliant Whois service

In light of the prior discussions of the types of gTLDs to be added and their potential value, it is
useful to examine the specifics under topic 4a, addition of new value to the Internet name space. The
subtopics are:

1. Name value. The proposed name must be of broad significance and establish clear and lasting
value. It should categorize a broad and lasting field of human, institutional, or social endeavor or activity.
It should represent an endeavor or activity that has importance across multiple geographic regions.
2. Enhanced diversity of the Internet name space. The TLD must create a new and clearly
differentiated space and satisfy needs that cannot be readily met through existing TLDs. The proposed

82 Summit Strategies International, 2004, "Evaluation of New gTLDs," pp. 130-131.
83 Summit Strategies International, 2004, "Evaluation of New gTLDs," p. 131.

Prepublication Copy--Subject to Further Editorial Correction
TLD should enhance competition in registry services and should attract new suppliers and user
communities to the Internet.
3. Enrichment of broad global communities. The TLD should have broad geographic and
demographic impact. "Significant consideration" will be given to those gTLDs that serve larger user
communities and attract greater numbers of registrants. "Consideration" will be given to those gTLDs
whose charters have relatively broader functional scope.

These specifics, together with the sponsorship requirement, indicate that ICANN is adhering to a
taxonomic/restricted approach in this selection, not allowing the TLD applicant to simply let the market
decide which TLD names will succeed, but requiring that a gTLD name should categorize a "broad and
lasting field of human, institutional, or social endeavor or activity . . . that has importance across multiple
geographic regions."

Ten proposals84 were submitted by the March 2004 deadline. Evaluation of the proposals was
carried out by teams external to ICANN and, therefore, not involved in ICANN activities or subject to
ICANN political pressures. Each evaluation team comprised three members: a technical, a financial, and a
sponsorship-and-other-issues evaluator. The evaluators, whose identities were kept secret during the
process, were selected and managed by an outside firm.85 The evaluation teams made recommendations
about the preferred applications from among those that were successful in meeting the selection criteria.
Some proposed domains met all of the criteria and entered a contract negotiation with ICANN staff. Other
proposals did not meet all the criteria, were sent back to the sponsors with suggestions for improvement,
and were resubmitted. The result is that new gTLDs will be announced one at a time and not all at once as
in 2000. In October 2004, ICANN announced that it was negotiating with two prospective new gTLDs:
.post and .travel; in December 2004, it began negotiations with .mobi and .jobs. None of the
negotiations had concluded by the beginning of March 2005.

Evaluation. Although ICANN adhered to the comparative hearing approach in this selection process, it
significantly reduced the use of staff and board judgment and relied instead on the judgment of
independent outside evaluators using an explicitly defined set of criteria. By doing so and by giving
applicants the opportunity to revise and resubmit their applications, ICANN increased the apparent
transparency (although keeping evaluators anonymous) and apparent objectivity of the process (although
using criteria subject to a wide range of discretion), and reduced the potential for disappointed applicants
to challenge the scores awarded by the evaluators.

Alternative B: Unlimited gTLDs Awarded First-come, First-served to Qualified Sponsors (2003)

Description. Almost at the opposite extreme from the preceding approach lies a proposal by Ross Wm.
Rader in which an undetermined number of new gTLDs would be awarded on a first-come, first-served
basis to approved "delegants" that would in turn contract with ICANN-accredited "operators" for the day-
to-day operation of the gTLD.86

The delegant would be the "policy coordinator for the gTLD that ensures that the registry
operates in a manner that benefits its target community." A delegant would be approved by ICANN on
the basis of four criteria: (1) the requested gTLD name is not confusingly similar to an existing gTLD
name; (2) the delegant has a satisfactory plan specifying all significant operational policies; (3) an

84 The list is available at <>.
85 Summit Strategies International in Washington, D.C. See Sarah Lai Stirland. 2004. "Domain-Name Registry
Expansion Process Underway." National Journal Tech Daily. May 26.
86 Ross Wm. Rader. 2003. "A Sustainable Framework for the Deployment of New gTLDs Part I." CircleID.
February 26. Available at <>; and Part II. March 26. Available at

Prepublication Copy--Subject to Further Editorial Correction
accredited operator is willing to manage the gTLD; and (4) the delegant yields rights in the gTLD name
so that it can be transferred by ICANN to a new sponsor if required.

The operator would perform the registry functions. To be accredited by ICANN, the operator
would have to satisfy the minimum standards for technically operating a registry. There would be no
limitation on the number of gTLD registries an accredited operator could contract to run.

Evaluation. Alternative B lies squarely in the market determination camp, both with respect to
determining how many gTLDs there should be and who should operate them and with respect to the
organization of gTLD names. In fact, it essentially allows there to be an unlimited number of gTLDs,
simply determined by the number of delegants who qualify. It does not even specify a limit to the rate of
addition of new gTLDs.

This process could be combined with--for example--a policy of adding X new gTLDs per year
by proceeding down the list of first-come, first-served requests and filling them as slots become available.
But that raises the question of how to actually implement a first-come, first-served process. How would
all the essentially simultaneous electronic entries submitted at the opening instant be prioritized? If a
random selection were to be used, for example, then this process would become--at least initially--a

Nor does the proposal specify a process for resolving conflicts that might arise with trademark
holders if, for example, someone other than Sony wished to register .sony as a gTLD. Presumably,
however, some variant of the sunrise procedures used to protect trademark holders when the .biz and
.info domains were introduced could be used in this instance.

Alternative C: Fixed Number of gTLDs Annually Awarded by Auction and Lottery (2003)

Description. Mueller and McKnight87 have proposed an auction/lottery process to award a limited number
of new gTLDs each year. 88 No structure would be imposed on the specific gTLDs awarded, resulting in a
supply-and-demand-driven approach to the types of gTLD names accepted.

Forty new gTLDs would be made available for award each year. That number was chosen based
on advice from members of the technical community with the goals of retaining the hierarchical structure
of the DNS and avoiding the introduction of errors into the root zone through too many changes, made
too fast. Although 40 was the number chosen for the proposal, any number up to about 80 would be
compatible with the authors' understanding of the advice they received.

The 40 new gTLDs would be divided into two tranches: 30 would be targeted for commercial
applicants (but open to non-commercial as well), which could apply for any number of TLD names but be
awarded no more than 2; 10 would be reserved for non-commercial and lesser developed country (LDC)
applicants, which could apply for and be awarded only 1 name. Each applicant would pay a modest
application fee, estimated by Mueller and McKnight at $1000, to cover ICANN's costs of the selection
process. The authors leave open the question of whether applications should include a fitness disclosure
and statement of financial capability to operate a registry or association with an already accredited

The 10 noncommercial gTLDs would be allocated first, if necessary, by a random selection
process, such as a lottery. To avoid possible abuse, the resultant TLD allocations could not be sold or
transferred to commercial entities.

87 Milton L. Mueller and Lee W. McKnight. 2003. "The Internet: Toward Regular and Objective
Procedures for Internet Governance." Syracuse University. Syracuse, New York. August. Available at
88 A similar auction model has been proposed in: Karl Manheim and Lawrence Solum. 2003. "The Case for gTLD
Auctions: A Framework for Evaluating Domain Name Policy." Research Paper No. 2003-11. Loyola Law School,
Los Angeles, Calif. March. Available at <>.

Prepublication Copy--Subject to Further Editorial Correction

The selection of commercial gTLDs would follow. If there were more than 30 commercial (and
other) applicants, the selection would be made by a simultaneous multiple-round Web-based auction, with
each bidder knowing whether it is in the top 30 or not at every time, just as with e-Bay. When the auction
period ended, the winning bidders would pay the amount of the lowest successful bid in order to keep
auction prices at a reasonable level. However, should there be multiple bidders for the same gTLD, such
as .sex, the high bidder would receive the gTLD and pay the amount bid by the second-highest bidder.
Having a predictable annual increment in gTLDs would also be expected to relieve some of the pressure
on the auction prices.

If there were 30 or fewer eligible commercial applicants, each would pay ICANN's reserve price,
which would be fixed to cover the costs of the auction and of adding new names to the root. (Note that if,
as some claim, there is only a limited demand for new gTLDs, this would be the outcome in the first, and
possibly only, auction.)

Mueller and McKnight suggest that the proceeds of the auction (or the reserve price) go to
maintaining and managing the root, with a portion reserved for the root name server operators. (See
Section 5.3.2.)

After the award of the gTLDs, there would be a period during which the gTLD names could be
challenged on intellectual property grounds or because of possible confusion with another gTLD. The
UDRP-like process would be used to resolve such challenges.

The winning gTLD applicants would enter into standardized and uniform registry agreements
with ICANN that would require adherence to a minimal set of ICANN-defined technical specifications
and conformity to existing ICANN policies. They would be required to meet standards for transferring a
zone file that would allow the domain to be maintained if the registry failed.

Evaluation. The result of the Mueller and McKnight proposal would be a market-determined structure of
the DNS name space, presumably responding--at least for commercial TLDs--directly to the supplier-
perceived demand for new TLDs. And except for the limitation on the rate of additions and the division
between commercial and non-commercial/LDC awards, this is also a clear expression of the market-based
approach to gTLD operator selection.

Those who favor a purer market-based approach could argue against the commercial/non-
commercial/LDC division on the grounds that LDCs already have country-code TLDs and that just as
with the allocation of other goods (e.g., office space, equipment, personnel) there is no need to favor non-
commercial organizations, which should be capable of raising funds to acquire a gTLD if they need one.

Consistent with its "let the market decide" approach, this proposal does not specify any
restrictions on the registration practices of the awarded TLDs except that non-commercial/LDC TLDs
cannot become commercial.

Alternative D: Differentiated Expansion of gTLDs with Selection by Comparative Hearing (2002)

In 2002, the Business Constituency (BC) of ICANN proposed that all future expansion of
gTLDs should occur within the framework of a previously agreed set of principles.89 In its view:

Given that there is pressure on ICANN to introduce additional names, the BC supports the
development of a logical expansion, which will result in a name space with added value, rather
than the cloning of the existing space. Such a value-added space will create differentiation and
reduce the need for entities to defensively register.

Users--regardless whether they are businesses, non-profit organizations or individuals--want
certainty. Spending time searching is not cost-effective. The user community needs a certain

89 Business Constituency, 2002, "A Differentiated Expansion of the Name Space," p. 1.

Prepublication Copy--Subject to Further Editorial Correction
process for identifying prospective names and a certain process for selecting sponsors/registries to
operate those names.90

Under this approach all new domain names would have to satisfy six principles:

1. Differentiation: be clearly differentiated from other gTLDs,
2. Certainty: give the user confidence that the name stands for what it purports to stand for,
3. Honesty: avoid increasing opportunities for bad faith entities that wish to defraud users,
4. Competition: create value-added competition,
5. Diversity: serve commercial or non-commercial users, and
6. Meaning: have meaning to its relevant population of users.91

In the view of the BC, "the principles . . . determine a taxonomized or directory-style domain
name structure. . . . The structure does not imply a rapid expansion. The choice of one name will
preclude future non-differentiated choices."92 The BC also argues that all new gTLDs should be both
sponsored and restricted. Only bona fide members of the target group would be able to register in a
gTLD. The sponsor would be responsible for ensuring that the registered names are appropriate for the
registrants and do not infringe on others' intellectual property. The BC believes that this approach
"simultaneously solves three intellectual property issues. Cyber-pirates will not be able to obtain the
names of others. There will therefore typically be no need for costly defensive registration. New Whois
databases will be verified and therefore accurate."93

The BC proposal recommends separation of the functions of sponsor and registry, similar to but
less specific than the Rader proposal (alternative B) described above. The goal is to create a set of
qualified registries that can operate any number of gTLDs under contract with the gTLD sponsor. Should
a registry fail, it could readily be replaced by another qualified registry.

Evaluation. One goal of the BC proposal is to enhance the role of the DNS as a directory service for the
Internet. (See Chapters 6 and 7 for a discussion of the role of the DNS as an Internet navigation aid.) Its
principle of differentiation is intended to avoid the Internet users' navigation confusion that might result
from overlap among gTLDs. But it would have precluded the creation of .biz as a way of giving
Internet providers another chance at a preferred second-level domain name that had been registered by
someone else in .com. Moreover, by emphasizing differentiation, it would limit price competition in favor
of non-price competition between gTLDs.

The BC proposal is not specific on the rate of addition of new names, other than that it is not
necessarily a rapid expansion. Nor is it specific on the process by which sponsors and registries for the
new names would be selected. It is silent on the fact that, as noted above, many of the existing names
would not belong in a strictly taxonomic structure.

By requiring that all TLDs be sponsored and restricted, this alternative places on the sponsors and
registries the responsibility of enforcing intellectual property rights and qualifying an organization's or
individual's right to use a specific domain name. By doing so, it moves from ex post to ex ante
enforcement of those rights and from consideration of name assignments when an issue arises to
examination of every assignment in advance. This approach can be expected to raise the costs of running
a registry and, consequently, to increase the likely registration fees.

90 Business Constituency, 2002, "A Differentiated Expansion of the Name Space," p. 1.
91 Business Constituency, 2002, "A Differentiated Expansion of the Name Space," p. 1.
92 Business Constituency, 2002, "A Differentiated Expansion of the Name Space," p. 2.
93 Business Constituency, 2002, "A Differentiated Expansion of the Name Space," p. 2.

Prepublication Copy--Subject to Further Editorial Correction
Comparison of the Four Alternatives

The four alternatives presented above for adding new gTLDs are compared in Table 5.3 on two
dimensions: desirable structure of the name space and the means of selecting gTLD operators. Both the
BC and the ICANN alternatives anticipate only sponsored and restricted gTLDs, whereas the Mueller and
McKnight and the Rader proposals allow, but do not require, them.

TABLE 5.3 Alternatives for Adding Generic Top-Level Domains

Means of Selecting gTLD Operators
Desirable Structure of
Name Space
First-come, First-served
Comparative Hearing
Auction (and Lottery)
Taxonomic / restricted

D. Business constituency

A. ICANN 2004 addition

ICANN 2000 Addition

B. Rader

C. Mueller and

5.4.3 Recommendations

ICANN's 2004 modification of its comparative hearing process so as to eliminate (or at least
significantly reduce) subjective judgments by its staff and board and increase the process's transparency
and objectivity has reduced the potential sources of dissatisfaction with the resultant selections. However,
the question still remains open as to whether it is really necessary for ICANN to qualify new gTLDs on
such matters as sponsorship by a community, business and financial plans, and addition of new value to
the name space.
An alternative approach would be qualification of applicants only on technical capability, basic
financial viability, and adherence to registrant protection standards and ICANN policies. Then a market-
based selection process--essentially an auction--could be used to select among qualified applicants when
their number exceeds the number of available slots.

Recommendation: If new gTLDs are to be created, the currently employed comparative hearing
or expert evaluation processes should not be assumed to be the only processes for selecting their
operators. ICANN should consider alternate processes that are less reliant on expert, staff, or board

Recommendation: Any gTLD auction selection process should be designed with reference to the
substantial literature on auction design.94 The following principal matters will have to be decided:

What is being auctioned. Is it a "slot" that the winner can use for a TLD with its choice of
name and operating policy, or the "right to operate" one of a prespecified list of TLD names and policies,
each of which is subject to a separate auction? Is it an outright sale or the license to operate for a fixed,
potentially renewable term?

94 See, for example: Paul Klemperer. 2004. Auctions: Theory and Practice. Princeton, N.J.: Princeton University

Prepublication Copy--Subject to Further Editorial Correction
How the winner is chosen. Does the high bid win, or the low bid? In the latter case the bid
would be the maximum amount the winner would charge for a basic registration in the TLD. (This latter
case raises the issue of tightly specifying levels of service so that the bids can be comparable.)
What is done with the proceeds of a "high bid wins" auction. Would the proceeds go to cover
ICANN operating costs or be allocated to other organizations, such as the root name server operators or
the IAB/IETF in support of technical activities related to ICANN responsibilities?
How commercial and non-commercial bidders are treated. Would they participate equally in
the same auctions, or would there be separate selection processes for each?
What the auction mechanism is. Would it be open or closed; single or multiple iterations;
with a reserve price or not; English or Dutch;95 and so on?
How the bidders are qualified. Would it be "no restrictions" or through some combination of
technical capability, financial strength, continuity provisions, and so on?
How many "slots" or "rights" a bidder could win. Would it be just one or two or as many as
desired? How post-auction actions by winners are restricted. Would it be at all; no resale for X years;
no change in policies; and so on?
How various failure modes are dealt with. What happens if the winner doesn't pay, or doesn't
proceed in a timely manner to set up the domain? What if the winner operates the domain poorly from a
technical or ethical point of view?

Because of the many choices for each of these matters and their many possible combinations,
there can be many different kinds of TLD auctions with different goals and quite different processes.
Furthermore, as the previously cited OECD report96 suggests, there can be various combinations of
comparative hearings (say, for qualification of prospective registries) and auctions (where demand for
gTLDs exceeds the number made available). The exploration of one or more such designs should be
included in ICANN's evaluation of alternative gTLD selection processes.


Issues: Does ICANN's responsibility for the root require that it work to increase its oversight of
and authority over the ccTLDs? If so, what form should its increased authority take, and how can it be
implemented? And to what degree should the ccTLDs accept ICANN's authority and participate in its

Although it does not have much visibility in the United States, the issue of who controls the
delegation and redelegation of ccTLDs is highly sensitive in many parts of the world. On the one hand,
some nations97 have expressed concern that the U.S. government, acting through ICANN, could
unilaterally remove their ccTLDs from the root zone file and, therefore, cut them off from the Internet. On
the other hand, some non-governmental ccTLD registries fear that their governments could take over
control of the ccTLDs through the use of ICANN's redelegation responsibility. In some sense, these fears
are anomalous since although ICANN has been given responsibility for management of the root, the vast
majority of the TLDs in the root--the ccTLDs--have, for the most part, eluded its direct authority. Fewer

95 In an English auction, participants bid openly against one another, with each bid higher than the previous one. By
contrast, in a Dutch auction, the auctioneer begins with a high asking price that is lowered until someone accepts the
auctioneer's price.
96 Working Party on Telecommunication and Information Services Policies, 2004, "Generic Top Level Domain
Names," p. 51.
97 For example, a Brazilian government representative at the February 2004 U.N. meeting on Internet governance
expressed this concern.

Prepublication Copy--Subject to Further Editorial Correction
than half of them (though they include some of the larger ones) contribute to ICANN's budget. Although
many participate in the ICANN meetings (and the special ccTLD meetings that have occurred at the same
time), they have not fully participated in ICANN's decisions. And few have signed agreements with
ICANN. Yet, through its control of the root zone file, ICANN does have the sole responsibility for
recommending delegations and redelegations of ccTLDs to the DOC. So the policy issues that face
ICANN are whether its responsibility for the root requires that it work to increase its oversight and
authority over the ccTLDs and, if so, what form such oversight should take and how it can be
implemented. And the complementary issue for the ccTLDs is the degree to which they should accept
ICANN's authority and participate in its activities. This latter issue is, of course, complicated by the fact
that the 243 ccTLDs do not act in concert98 and, as described in "ccTLDs" in Section 3.4.1, represent a
very wide range of relationships between government and registry and a very wide range of registry

5.5.1 Current Situation

As the Internet has grown and matured, the role and importance of ccTLDs has grown and
changed as well, as have their relationships with their governments, with their local communities, and
with ICANN.

A study in 2002 of the ccTLDs of 45 countries revealed a wide diversity of relationships between
ccTLDs and the countries' governments.99 The ccTLDs of 10 of the 45 countries surveyed were operated
by government agencies or departments, 9 by private commercial enterprises, 20 by non-profit
organizations, 5 by academic institutions, and 1 by an individual. Of the 35 non-governmental sponsoring
organizations, only 9 had a formal contractual relationship with their governments and 13 had informal
relationships, of which 3 were awaiting formalization. Three of the ccTLDs operators were battling
government attempts to take over management of the ccTLD. Altogether, only half of the studied ccTLDs
had formal or soon-to-be formal relationships with their governments. Their formal relationships with
ICANN were also weak. By early 2005, only 12 of the 243 ccTLDs had entered into formal agreements
with ICANN.

The consequence of this evolution is that the ccTLDs operate in a space that is only in part under
the oversight of any higher authority. A few ccTLDs are overseen by their national governments; some
have established representative non-governmental bodies to represent the local Internet community and
exercise varying degrees of oversight; some are completely autonomous non-profit bodies that operate
voluntarily to meet local Internet community interests; and some are commercial bodies with some
linkage to the national government.

The only body that currently has an opportunity to exercise oversight over all the ccTLDs is
ICANN. According to ICANN, it does so on the basis of RFC 1591 in conjunction with "ICP-1: Internet
Domain Name System Structure and Delegation (ccTLD Administration and Delegation)"100 and the
"Principles for Delegation and Administration of ccTLDs Presented by Governmental Advisory

98 There are, however, a number of regional ccTLD associations such as CENTR, the Council of European National
Top-Level Domain Registries, which has 39 member registries, not all of which are European. See
<>. CENTR has expressed strong positions on ICANN matters, especially its relationships with the
99 Michael A. Geist. 2003. ccTLD Governance Project. Available at <>. Data
for this project was obtained from ccTLD Web sites, ccTLD contacts and, GAC representatives between June and
September 2002. It was not claimed to be a representative sample.
100 Available at <>.
101 Available at <>. In 2004 this was to be replaced
by a new statement.

Prepublication Copy--Subject to Further Editorial Correction

The delegations in the early days of the DNS placed highest priority on the responsibility of the
manager and less on notions of ownership. However, with the current economic, political, and social
importance of the Internet to all nations, matters of accountability to the local government, local Internet
community, and the global Internet community have assumed much greater significance. Thus, matters of
delegation for new ccTLDs and redelegations of responsibility for existing ccTLDs have become much
more important and, when combined with the many different ccTLD-government relationships, have also
become much more complex.

The relationship between ccTLDs and ICANN has been difficult from the beginning of ICANN.
First, a large number of the ccTLDs felt no need to contribute to ICANN's budget, since they did not
think they received any corresponding benefits. Whether or not true,102 many ccTLDs believed that 90
percent of ICANN's resources were devoted to gTLD issues, while they were asked to provide 35 percent
of its budget.103 Second, many of them resented ICANN's major role in deciding on delegations and
redelegations--essentially a policy role--that they felt would be better performed locally. They also
believed that their position as one constituency within ICANN's Names Supporting Organization, whose
other constituencies primarily addressed gTLD issues, did not adequately reflect their importance as 243
of the 258 TLDs.

Under its 2003 reorganization (see Section 5.2.4, alternative F), ICANN has responded to that
concern by replacing the Names Supporting Organization with two supporting organizations, the Generic
Names Supporting Organization (GNSO) and the Country-Code Names Supporting Organization
(ccNSO), which formally came into being in March 2004. ICANN hopes, thereby, to draw the ccTLDs
more actively into its operations and build a stronger basis for their support. Furthermore, the
Governmental Advisory Committee was (in August 2004) working on a revision of its Principles for
Delegation and Administration of ccTLDs that addresses many of the concerns of the ccTLDs and

Although ICANN's actions may have the desired effect, it is useful to lay out the alternative
approaches to providing reasonable oversight of the ccTLDs that have been suggested.

5.5.2 Alternatives

ICANN appears to have four goals for the ccTLDs. First, that they operate according to
standards, protocols, and practices that are consistent with the reliable and stable operation of the DNS
and that allow open connectivity to and from their registrants and the rest of the Internet. Second, that
they contribute proportionately105 to the overall costs of maintaining the DNS root zone file and ccTLD
database in which they are listed. Third, that they accept ICANN's authority to decide on delegations and
redelegations when controversies arise. And fourth, that they formalize these expectations by means of an
agreement with ICANN. However, not all of the ccTLDs accept these as appropriate goals for ICANN's
relationship with them. What is the current situation?

102 However, according to a reviewer with knowledge of ICANN's activities: "Close to 50 percent of ICANN
time/resources has been devoted to ccTLD and international issues. . . . much of that time is spent on small ccTLDs
(with) complex redelegation issues. A considerable amount of time is also spent liaising with regional organizations
and governments . . . participating in international fora."
103 Peter de Blanc. 2001. "ccTLD Briefing Document." February 19. Available at
104 Although the text of the draft revision is not available on the ICANN Web site, CENTR has published its
response to a version circulated in January 2004. See <
105 In 1999, the Task Force on Funding recommended that the ccTLDs contribute a 35 percent share of ICANN's
continuing revenue requirements. See <>.

Prepublication Copy--Subject to Further Editorial Correction
The absence of significant and sustained operational problems suggests that the ccTLDs generally
meet the first of ICANN's goals, supporting reliable, stable, and open operation of the DNS and
However, there is only a modest contribution, at present, from a minority of the ccTLDs to
ICANN's budget. (In 2004-2005, the 243 ccTLDs' fixed contributions are budgeted at $1.02 million,
whereas 10 gTLDs are expected to provide $1.45 million. An additional $14 million is expected from the
accredited registrars and gTLD registries in proportion to actual registrations.) So they do not appear to
satisfy its second goal.
The ccTLDs represented by the European association of ccTLD registries, the Council of
European National Top-level Domain Registries (CENTR), favor a much more limited role for ICANN.
In their view, ICANN should restrict its operational responsibilities to maintaining the ccTLD database
(containing Whois information about the ccTLDs) and the corresponding root zone file entries and
ensuring that the ccTLDs satisfy minimal technical requirements to function within the global DNS. In
their view, delegations and redelegations should be made only by ICANN upon verified instruction from
the current manager of the ccTLD. Where there is controversy, it should be referred to a third party for
resolution. (This approach would be consistent with the proposals noted earlier to increase ICANN's
legitimacy by narrowing its range of authority and by delegating sensitive decisions to appropriate third
parties wherever possible.)
An even stronger view, held by a number of ccTLD managers, is that the maintenance of the
ccTLD database should be the responsibility of a ccTLD-sponsored organization independent of ICANN.
This organization would receive entries from the ccTLDs and have full responsibility for updating the
database to reflect the latest information. Through a contract with ICANN, it would provide the relevant
ccTLD entries for inclusion in the root zone file, but ICANN would play no part in deciding what the
entries would be. Participation in the independent database would be voluntary for each ccTLD. If a
ccTLD desired, it could continue to submit information directly to ICANN. As a result, each ccTLD
would have a "vote" on whose management of the ccTLD database it favored.
Thus, although ICANN's delegation and redelegation authority is accepted de facto--because it
currently has sole authority to recommend changes in the root zone file to the DOC--there are many
ccTLDs that dispute either the way that authority is exercised or that it should be an ICANN function at
all. Consequently, its third goal has not been satisfied.
Finally, as noted above, only 12 agreements have been signed between ICANN and ccTLDs. So
ICANN's fourth goal has not been satisfied.
It should be observed that a comparable list of the goals of the 243 ccTLDs cannot be provided,
since they are consistent neither with each other nor, necessarily, with ICANN. As efforts are made to
increase the ccTLDs' role in ICANN and ICANN's influence over them, many independent ccTLDs will
straddle the fence, playing the various forces off against each other until the ambiguities and uncertain
power relationships among the U.S. government, their own national governments, ICANN, and the
international community are worked out. Because of the diversity of ccTLD models, histories, and
relationships to national governments, it is unlikely that any proposal will satisfy everyone; but some
aggressive or lopsided proposals are likely to antagonize all of them.
How then can ICANN best interact with the ccTLDs? The differing goals of ICANN and of
some of the ccTLDs have led to four proposed models for ICANN's "oversight" of the ccTLDs.

Alternative A: "Thick" ICANN

106 On rare occasions, a ccTLD may temporarily disappear from the Internet as the Libyan ccTLD did in April 2004.
This is often the result of political or institutional disputes within a nation over responsibility for the ccTLD, with
which ICANN has to deal. See "Who Runs the Dot LY?," Libyan Jamahiriya Broadcasting Corporation. 2004.
Available at <>.

Prepublication Copy--Subject to Further Editorial Correction
The model initially implemented by ICANN has had ICANN attempting to achieve its goals by
playing a strong role in the oversight of the ccTLDs: maintaining the ccTLD database and ccTLD entries
in the root zone file, making recommendations to the DOC concerning delegations and redelegations,
establishing standards for ccTLD performance, and inducing compliance through MoUs with the ccTLD
managers or the relevant governments. Currently, it is seeking through the newly established ccNSO to
engage the more active participation of the ccTLDs in ICANN activities.

By elevating the ccTLDs to the status of a supporting organization, similar to the GNSO, ICANN
has both raised their profile in the organization and given them influence over board membership and
ICANN policy. By engaging them more directly in its governance, it evidently hopes to gain their support
for its role in ccTLD oversight. At the same time, the revised ICANN bylaws107 give the ccNSO the
principal role in establishing policy for entry of data concerning ccTLDs into the root zone file. Over
time, this may lead to a ccNSO-led redefinition of the delegation/redelegation policies and practices of
However, some of the most important ccTLDs were not participants in the ccNSO upon its
formation. Most notably, only four of the European ccTLDs (from the Netherlands, the Czech Republic,
Gibraltar, and the Cayman Islands) were among the 38 founding members. In April 2004, the European
Community (EC) member responsible for the Internet said that the EC will stand by ICANN as long as it
continues to make changes and that unless the EC ccTLDs come to agreement with ICANN, the EC will
lose patience and the governments will step in, possibly turning ICANN's ccTLD role over to the ITU.108
But the distance that remains between ICANN and the European ccTLDs was clearly shown by the
response of CENTR to ICANN's proposed 2004-2005 budget. In a May 2004 letter to ICANN, CENTR
accused ICANN of a "lack of financial prudence" and refused to support it "financially or otherwise" in
its "unrealistic political and operational targets."109 Specifically, it said: "ICANN/IANA should focus on
doing a few administrative tasks well and not seek to make decisions--decisions are best handled

Alternative B: "Thin" ICANN


Many ccTLDs would favor a much more limited role for ICANN, essentially reducing it to
performance of a technical coordination function and eliminating all policy functions. Under this
approach, ICANN would continue to run the IANA function, maintaining the database of ccTLDs and the
ccTLD entries in the root zone file. However, decisions about the delegation of responsibility for a ccTLD
when it is a subject of dispute would not be made by ICANN. Rather, they would be made, in the first
instance, by the local Internet community relying on national laws and processes as necessary, and if that
failed, by a process established by the ccTLD community. In this model, the ccTLDs would agree to pay
the cost incurred by ICANN in maintaining the database of ccTLDs and ccTLD entries in the root zone
file through a fee based on the size of each ccTLD's membership.

107 The changes in ICANN bylaws that establish the ccNSO and define its roles are in: ICANN. 2003. "Appendix A
to Minutes of Regular Meeting of ICANN Board." June 26. Available at <
108 Speech by EC Commissioner Erkki Liikanen quoted by Kieren McCarthy. 2004. "EC Tells Europe and ICANN
to Make Peace." The Register. April 28. Available at
109 Paul M. Kane, letter to Paul Twomey, May 26, 2004. Available at

Prepublication Copy--Subject to Further Editorial Correction

Alternative B would reduce ICANN to performance of a technical/administrative root registry
function, eliminating its role in determining who among alternative claimants should have the right to be
the registry for a specific ccTLD. In that sense, it is compatible with the approaches to gTLD selection
that favor the use of auctions to determine which gTLDs should enter the root zone file. A combination of
the two approaches would virtually eliminate ICANN's role as a gatekeeper to the root and leave it
primarily as the record keeper and, presumably neutral, validator of the technical qualifications of

Alternative C: International Oversight


In the third model, a major part of the IANA function would be removed from ICANN and turned
over to a third-party organization established by the ccTLDs. That organization would be responsible for
maintaining an up-to-date ccTLD database and sending the appropriate information to ICANN for entry
into the root zone file. Presumably, there would be a corresponding agreement with the DOC (as long as it
retained its stewardship role) to accept the information from the third-party organization as authoritative.
If that organization had broad international participation in its governance and activities, this might
alleviate some of the discontent with the U.S. government's current central role as steward of the DNS.
One possibility would be, for example, to turn the ccTLD database and delegation responsibilities over to
the ITU. However, the ITU is an intergovernmental organization, responsible primarily to the
telecommunication ministries of governments. Yet, as noted earlier, many ccTLDs are either simply
independent from their governments or operating in delicate balance with them. They would probably not
like a process whose only recourse is to the governments.


This alternative takes alternative B a step further by eliminating even the record-keeping function
of ICANN with respect to the ccTLDs. Its role would simply be to pass the appropriate entries to the
organization responsible for distributing the root zone file (currently VeriSign). In fact, this amounts to
establishing an "ICANN" for the ccTLDs, leaving the present ICANN with responsibility only for the
gTLDs. Although alternative C would certainly eliminate any ccTLD discontent with ICANN, it might
simply shift the focus of concern to the new organization, depending on how the issue of
delegation/redelegation decision making was decided. This possibility suggests that finding a solution to
the delegation/redelegation decision process within the existing ICANN that is satisfactory to the ccTLDs
and their governments would preclude ccTLD support for alternative C.

Alternative D: Self-governing Root Management Organization


An additional possibility would be an ICANN focused solely on root management responsibilities
whose members are limited to those groups, including the ccTLDs, having a direct interest in the root.
(See "Alternative C: ICANN as Registry for the Root" in Section 5.2.4.) This model assumes that the
ccTLDs would take an active role in ICANN's governance and would, therefore, be more willing to see
ICANN play an active, "thick" role in ccTLD oversight. By creating a ccNSO and giving it greater
influence on board composition and enabling it to submit recommendations to the board for its
unmodified approval or rejection, ICANN has taken a step in this direction.


Prepublication Copy--Subject to Further Editorial Correction
Alternative D, discussed in greater detail as one of the ICANN alternatives, could achieve the
goal of bringing the ccTLDs fully into the management of an organization whose authority would be
strictly limited to management of the root. Once again, however, the issue would devolve to the sensitive
one of how delegation/redelegation decisions are made. This alternative might be more attractive to the
ccTLDs because it would presumably increase the strength of their influence over ICANN's operations
and decision-making processes. However, the issue still remains the degree to which they, a highly
diverse group, would all be comfortable with the policies and practices that would determine which
registry is delegated or redelegated responsibility for a ccTLD.

Comparison of the Four Alternatives

The four alternative models described above for oversight of ccTLDs are compared in Table 5.4
on two dimensions: who maintains the ccTLD database and who recommends redelegations (to DOC).
Three models assume that ICANN will continue to manage the ccTLD database. Only one posits an
independent manager of the database. Two models foresee the decision/policy responsibility for
delegations and redelegations being removed from ICANN.

TABLE 5.4 Alternatives for Oversight of the Country Code Top-Level Domains

Maintainer of ccTLD Database
Recommender of Redelegations
Third Party
A. Thick ICANN

D. Self-governing root
management organization
Third Party
C. International oversight

5.5.3 Conclusions

Resolution of ICANN's role vis-à-vis the ccTLDs is one of the critical steps on the
path to establishing an ICANN that is viewed as a legitimate and appropriate steward for the DNS.

The creation of the ccNSO represents progress in that direction whose success will depend on the
ccNSO's ability to attract an increasing number of members, both from the large ccTLDs that are needed
for financial and other support of ICANN and the smaller ccTLDs that can benefit from the support that
ICANN could offer them. Even more critical is the refinement of the principles and processes for
delegation and redelegation of ccTLD registries and their acceptance by most of the ccTLDs.

If the creation of the ccNSO does not result in increased participation by the
ccTLDs in ICANN policy making, then ICANN may find itself subject to increasing pressures to
constrain its role to that of gTLD management and root zone file record keeping.


Issue: Does the UDRP need to be improved? If so, how should it be improved?


Prepublication Copy--Subject to Further Editorial Correction
Administrative processes, such as the Uniform Domain-name Dispute Resolution Policy (UDRP),
are playing an important role in helping to resolve certain private-party disputes related to the use of
domain names, without requiring ISPs, registries, registrars, registrants, or other parties to appear in court
to provide evidence or to protect their interests. Such processes contribute to the smooth operation of the
economic and legal framework associated with the DNS. At the same time, while the highly simplified
rules created by the UDRP and similar dispute resolution procedures (e.g., those adopted by some
ccTLDs) and dispute avoidance efforts (e.g., sunrise provisions) make it possible for some actual or
potential disputes to be resolved quickly at relatively low cost and without requiring the parties to be
represented by legal counsel, they can also undermine the potential for fair outcomes. The UDRP is the
primary subject of this section, since most attention has been paid to issues concerning it.
During 2003, the ICANN staff carried out a review of the UDRP. It produced an issues report in
August 2003.110 The report cataloged and identified the pros and cons of proposed solutions to both
procedural and substantive issues and concluded that "while there are some areas where improvements
may be possible, there does not appear to be an urgent need for revision." Furthermore, it noted that
"revision of the UDRP is likely to be contentious; there are not many (if any) areas that are obviously
amenable to achieving consensus." (Since the UDRP is a consensus policy, according to provisions of the
ICANN registry and registrar agreements it must be revised by consensus.)
To provide an understanding of the issues that have given rise to those proposals, this section
begins with an assessment of the UDRP and then examines the major proposals for improvement. It
incorporates the committee's conclusions and recommendations. The section concludes with a discussion
of the potential consequences of deployed internationalized domain names (IDNs) for dispute resolution.

5.6.1 Assessment of the UDRP

Many observers believe that the UDRP has functioned well to resolve disputes over domain
names.111 However, there are others who believe that the current system is biased toward the interests of
trademark holders and away from the interests of individuals.112 Notwithstanding its perceived
disadvantages, numerous decisions have been rendered under the UDRP, and other domain name dispute
policies have been modeled after the UDRP as described in "Resolving Domain Name Conflicts" in
Section 3.5.2. Thus, the UDRP has both positive and negative aspects, which differ, however, depending
on whether they are being considered from the perspective of the complainants or of the respondents.

General benefits. The UDRP crosses national boundaries and relies on communication
technology to bring the parties together It is more informal than litigation in national courts and relies on
panelists who are experts in the areas of trademark law and domain name issues. The proceedings are

110 ICANN. 2003. "Staff Manager's Issues Report on UDRP Review." August. Available at
111 One such favorable assessment appears in: Colm Brannigan. 2004. "The UDRP: How Do You Spell Success?"
Digital Technology Law Journal. Vol. 5. Issue 1. July. Available at
<>. A careful assessment of the pros and cons of the
UDRP can be found at: Laurence Helfer and Graeme Dinwoodie. 2001. "Designing Non-national Systems: The
Case of the Uniform Domain Name Dispute Resolution Policy." William and Mary Law Review. Vol. 43. Issue 1.
October. pp. 141-273. Available at <>.
112 See, for example: A. Michael Froomkin. 2002. "ICANN's `Uniform Dispute Resolution Policy'--Causes and
(Partial) Cures." Brooklyn Law Review. Vol. 67. No. 3. pp. 608-718. Available at
<>. In addition, see: Michael Geist. 2001. " An
Examination of the Allegations of Systemic Unfairness in the ICANN UDRP." University of Ottawa, Faculty of
Law. August. Available at <>. See also a follow-up piece by the same
author in March 2002, "Fundamentally An Update on Bias Allegations and the ICANN UDRP."
Available at <>.

Prepublication Copy--Subject to Further Editorial Correction
quasi-in-rem, meaning that even though both parties are included, the action is focused on resolving
which party has rights to the domain name, rather than assessing fault or monetary damages against either
party. Although it is international in scope, it raises no jurisdictional issues, which may be present in court
litigation in some countries, by requiring all domain name registrants to agree to submit to a mandatory
administrative proceeding as part of the registration agreement. The UDRP requires that the proceeding
be conducted in the language of the registration agreement, which eliminates language as a potential
barrier to participation by domain name registrants.
Complainant's benefits. From the complainant's (i.e., trademark holder's) point of view, the
UDRP's positive features include that it provides a quick and relatively inexpensive method of resolving
a domain name dispute and obtaining the transfer of a domain name to the trademark owner. Domain
name disputes brought under the UDRP are generally resolved within 45 to 60 days of the domain name
dispute provider's receipt of the complaint. In addition, the UDRP is not limited to registered trademarks
identical to the domain name, and it allows trademark owners to file a complaint against a registrant of a
domain name that is "confusingly similar" to the owner's mark. Further, owners with common-law rights
in trademarks may also take advantage of the UDRP as there are no trademark registration prerequisites to
commencing a UDRP action.

The UDRP is a cost-effective dispute resolution mechanism overall because it (1) is based
primarily on the pleadings of the parties, (2) only allows in-person hearings under exceptional
circumstances, and (3) only allows additional evidence at the discretion of the panel.113 Furthermore, the
parties are generally not required to travel in order to participate in the proceeding, which is usually
conducted by postal mail, e-mail and/or facsimile. This last point can be seen as a greater advantage for
the complainants, since ordinary court proceedings would occur in the respondents' locales, thus,
requiring the complainants to travel.
Respondent's benefits. From the respondent's perspective, the panel can grant the
complainant the requested remedy (i.e., transfer or cancellation of the domain name registration) only if
the complainant succeeds in showing all three of the UDRP elements (see "Remedies to Conflicts over
Names in the DNS" in Section 3.5.2), even if the respondent did not submit a response. Respondents see
other advantages to the UDRP as well, since it requires trademark owners to comply with a standard
stricter than that of the courts--demonstrating that the respondent had both registered and used the
domain name in bad faith (although it has been argued that not all panelists have adhered to this
requirement). In addition, the UDRP allows a party against which an adverse decision is rendered to take
the decision to the courts.
Moreover, the UDRP provides limited remedies to trademark owners, namely simply transfer or
cancellation of the domain name. To receive monetary damages or an injunction, a trademark owner
would have to proceed to litigation.
Complainant's disadvantages. From the complainant's perspective, the preparation of a
proper complaint and necessary appendices can be time-consuming and costly, and although much less
costly than preparing for litigation, it is still viewed by many complainants as excessive.
In addition, the only avenue, at present, for correcting what the complainant views as an improper
decision by the panel is to litigate the same matter before a court having in personam jurisdiction over
the respondent, or in some cases in in rem jurisdiction over the domain name in dispute. This is not
perceived as an advantage by all complainants, especially if the respondent is not located in the same
location as the complainant and the domain name was not registered in the United States, where in rem
jurisdiction is possible in some cases.
Moreover, complainants see disadvantages in the limitations on remedies (no potential damage
recovery no matter how egregious the respondent) and in panels' inconsistent definitions of critical terms,
such as "confusing similarity," "use," and "bad faith."
Respondent's disadvantages. There are also, however, disadvantages that fall primarily on the
respondents. Although the UDRP allows administrative proceedings to be conducted in languages other

113 ICANN. Rules for Uniform Domain Name Dispute Resolution Policy, October 24, 1999, Paragraph 12.

Prepublication Copy--Subject to Further Editorial Correction
than English, the UDRP itself is written in English. Many non-native-English speakers who register
domain names with registrars that do not provide translated versions of their registration agreements or
the UDRP may not be aware that they are subject to the provisions of the UDRP or that they should avoid
selecting a domain name that violates the trademark rights of other parties.114 Additionally, since the
complainant selects the dispute resolution service provider, it is possible for complainants to "forum
shop" (i.e., to select a provider more likely to favor the complainant, or which has been more sympathetic
to similar complaints in the past).115
Some critics have also alleged that providers, seeking to increase their chances of being selected
by future complainants, purposely choose arbitrators who are more likely to favor complainants, but little
concrete evidence supporting this allegation has been provided. Nevertheless, arbitrator selection bias
would be a serious issue were it to occur, and service providers should be reviewed on a periodic basis to
make sure such bias does not exist.
Once a decision is rendered, the respondent's only recourse for dealing with a decision
transferring or canceling the domain name is to proceed to court.
General deficiencies. Critics of the current UDRP116 have pointed to a number of perceived
deficiencies. Among them are that some panelists do not apply the precedents of previous arbitrations
appropriately, or in some cases consistently; some panelists (and many respondents) are not well-enough
educated in either the operations of the DNS or the policies and rules applicable to domain name disputes;
the charges for a UDRP proceeding and the ways in which panelist are compensated can lead to undesired
consequences; and there is no appeals process with the UDRP itself--the only appeal is a de novo action
before a court.
Each of these perceived deficiencies is described and proposals for remedying them are addressed
in the next section.

Conclusion: The UDRP has generally satisfied the need for an effective and cost-efficient means
of resolving disputes concerning domain names; however, it has weaknesses for which remedies have
been proposed.

5.6.2 Proposed Improvements to the UDRP

In response to the perceived deficiencies of the current UDRP, a number of improvements have
been proposed: a better, more consistent application of arbitral precedents; an appeals process; required
use of three-member panels; improved training and self-help tools; and revised funding and compensation

Better application of precedents. Some believe that more consistent application of arbitral
precedents in UDRP proceedings is needed, so that similar issues can be addressed in a more predictable
manner that also supports case-by-case knowledge building. In addition, greater consideration of

114 As noted by one reviewer: ". . . many countries have consumer protection laws that require all consumer
contracts concluded within the jurisdiction to be in the local language in order to be valid and enforceable. This
condition is not satisfied by the UDRP's requirement that the proceedings be conducted in the language of the
registration agreement." See Holger P. Hestermeyer. 2002. "The Invalidity of ICANN's UDRP Under National
Law," Minnesota Intellectual Property Review. Vol. 3. Issue 1.
115 An early analysis of the UDRP that asserted that "forum shopping" was a source of bias favoring the complainant
was: Milton Mueller. 2000. "Rough Justice: An Analysis of ICANN's Uniform Dispute Resolution Policy."
Convergence Center. Syracuse University School of Information Studies. Available at
<>. The INTA study, "UDRP-A Success Story" (2002), is a rebuttal
to the Mueller article. Michael Geist's reports, "" (2001) and "Fundamentally" (2002), both
provide further data on the asserted complainant-beneficial effects of forum shopping.
116 See, for example, the Froomkin, Mueller, and Geist articles cited above.

Prepublication Copy--Subject to Further Editorial Correction
international legal issues is needed, given that laws vary from country to country. Because the panelists in
UDRP proceedings tend to be most knowledgeable about their home country's laws, new issues in
disputes tend to be examined through the legal lens of a particular panelist's country. When these
decisions are relied on in later cases, panelists unfamiliar with the legal context of the original decision
will often assess them--inviting misinterpretation, however well-intentioned a panelist may be. One way
to encourage better and more consistent use of arbitral precedents is to have an internal appeals process,
whose panelists would be in a better position to require and make use of precedents.
Appeals process. An appeals process could serve the purpose of reversing decisions that were
clearly faulty or that covered a situation or issue for which competing bodies of precedent exist.117 To
remain consistent with the original purpose of the UDRP, the intent of such a process would be to re-
examine only a small percentage of decisions, so as to provide an inexpensive mechanism (as compared
to a court case) for resolving relatively straightforward cases. As noted above, an appeals process would
have the likely effect of encouraging better and more consistent use of arbitral precedents. But support for
an appeals process has been limited, with the emphasis being placed on resolving such issues through a
national court proceeding--rather than creating another layer to a relatively quick and inexpensive
dispute resolution process. Those who hold that view emphasize that the UDRP, with or without an
appeals process, is not intended to serve as a full substitute for national and international law or courts,
but simply to provide a quicker and less costly process for the majority of disputes, whose resolution is
often obvious. With careful design and restriction to very specific situations, the proposal for a limited
appeals process could be consistent with that intent.
Three-member panels. Analyses conducted during 2001 and early 2002 of UDRP
proceedings indicated a significant difference in their outcomes, depending on whether they were heard
by one-member or three-member panels.118 (The choice is made by the complainant in the first instance,
but the respondent can request a three-member panel.) Three-member panels found for the complainant in
a smaller percentage of the cases. Critics have taken this as an indicator of greater bias toward
complainants by the one-member panels and have recommended, therefore, that all panels have three
members.119 (In three-member panels, the complainant, the respondent, and the provider each provide lists
from which one of the panelists is chosen.) Others have argued that the impression of bias is due to other
factors.120 Both sides agree that three-member panels are more expensive and, therefore, that they erode
the benefits of the UDRP as a relatively inexpensive means of resolving domain name disputes. Those
who favor them argue that the increase in fairness is worth the cost. In addition, they have suggested the
payment of a bond by the complainant that would be used to cover the respondent's costs of the
proceeding if the complainant lost and would be refunded if the complainant won.121
Improved training and self-help tools. In many UDRP proceedings the focus is on the use of
domain names in Web addresses (URIs). The role of domain names in e-mail addresses and in other
applications is often ignored by the panelists, even though the effects can be different from those in Web
addresses. One possible reason for this oversight is that some panelists may lack a technical
understanding of how the DNS and the Internet operate. Thus, some believe, UDRP proceedings could be

117 For a specific proposal, see: Patrick Kelley. 2001-2002. "Emerging Patterns in Arbitration Under the Uniform
Domain-Name Dispute-Resolution Policy." Law and Technology Writing Workshop, Annual Review of Examplar
Papers, Law School, University of California, Berkeley. Available at
<>. Another proposal for an appeals
process appears in: M. Scott Donahey. 2002. "Divergence in the UDRP and the Need for Appellate Review."
Available at <>.
118 Michael Geist, 2001, " An Examination of the Allegations of Systemic Unfairness in the ICANN
UDRP," and Michael Geist, 2002, "Fundamentally An Update on Bias Allegations and the ICANN
119 Michael Geist, 2001, ""
120 INTA, 2002, "The UDRP by All Accounts Works Effectively."
121 Milton Mueller. 2002. "Success by Default: A New Profile of Domain Name Trademark Disputes Under
ICANN's UDRP." June 24. Convergence Center. Syracuse University School of Information Studies.

Prepublication Copy--Subject to Further Editorial Correction
improved by enhancing the training requirements for panelists in the technology underlying the DNS, the
manner in which domain names can be used, and the application of the policies and rules applicable to
domain name disputes. Improvements in the process might be developed to help panelists verify the
manner in which domain names are used, either through self-help mechanisms or by changing the rules to
request this information from the respondents. Dispute resolution providers could provide training for
panelists on a regular basis, with such training being a requirement to maintain panelist status with that
Thorough and detailed self-help tools might also be developed to enable respondents to better
understand the UDRP process itself, the timeline involved, and the substance and format of an effective
response to better comprehend and respond to a UDRP action.122
Revised funding and compensation structures. Under the current funding structure, the
revenue for panelists depends on the volume of cases, thereby either creating a disincentive to spend a
sufficient amount of time reviewing the facts in a case and writing a well-thought-out opinion, or creating
an incentive for marketing strategy and tactics to attract cases by defining lucrative niches, which may or
may not correspond to justice in dispute resolution proceedings. Observers assert that such niches exist
and that complainants often forum shop--selecting dispute resolution service providers based on their
past record of favorable (to the complainant's position) rulings.
Since some parties believe that the $1150 to $1500 fee for filing a complaint regarding a single
domain name is already expensive, there is some resistance to any proposed increase. In addition,
increasing the fees paid to resolve or avoid a dispute raises the likelihood that, on the one hand, individual
domain name holders would be discouraged from employing dispute resolution processes, while, on the
other hand, well-financed domain name holders might be discouraged from filing large numbers of not
completely justified complaints.
On the compensation side, the fee paid to panelists, typically $1000 to $1750, is below the level
that highly qualified attorneys and consultants say is needed to attract them to serve or continue to serve
as panelists. While it may never be possible to set the level high enough to attract and retain highly paid
specialists on the basis of compensation alone, it may be worth examining the fee schedule to see whether
a higher level could be established while retaining the low cost of the UDRP.

Arbitral domain name dispute resolution processes, rather than national
courts, should continue to be encouraged as the initial and primary vehicle for resolving most disputes
associated with the rights to domain names.

Recommendation: The feasibility and desirability of five specific UDRP improvements should
be further considered by ICANN: improving consistent use of arbitral precedents, establishing an internal
appeals process, using three-member panels, improving panelist knowledge about the technology
underlying the DNS, and improving the nature and structure of incentives in the process.

5.6.3 Disputes Concerning Internationalized Domain Names

The widespread deployment of internationalized domain names (IDNs)123 may well compound
the difficulty of resolving disputes over domain names by increasing the possibility that domain names
will be created that appear to be the same, but are not.

122 Early in 2005 the World Intellectual Property Organization posted on its Web site an "informal overview of panel
positions on key procedural and substantive issues," including references to decisions supporting each line of
opinion. The "WIPO Overview of WIPO Panel Views on Selected UDRP Questions," which is not binding on the
panelists, is available at <>.
123 See Section 4.4 for discussion of internationalized domain names.

Prepublication Copy--Subject to Further Editorial Correction
The introduction of non-ASCII characters introduces a number of opportunities for conflicts, not
about domain names themselves, but about characters that look alike. As the most trivial of examples, the
upper-case Greek alpha and its Cyrillic equivalent are indistinguishable on the printed page from Roman
upper case "A," but the three have different Unicode code points and strings containing them will
compare differently. There are several similar combinations involving Roman, Greek, and Cyrillic scripts,
but other examples appear in almost all pairings of alphabetic scripts. Another problem occurs because of
the overlap between, for example, Simplified and Traditional Chinese, where the characters look different
but have the same meanings.
These concerns about similar-appearing, or similarly interpreted, domain names are compounded
by the observation that, in some circumstances, a single registered domain name might have dozens, or
even hundreds, of such variations. A cybersquatter could turn such conflicts into a potentially lucrative
business by offering to sell such variant names to the "legitimate" owner at a fee just below the cost of a
UDRP proceeding. Some of the registries and communities that would be most affected have concluded
that it is preferable to shift the problem, to the degree possible, from conflict resolution to conflict
avoidance by imposing restrictions on the registration of domain names that would conflict or otherwise
cause confusion. ICANN has reinforced this approach by creating a guideline that requires that an IDN
must be registered only with regard to a specified language, which eliminates some of the difficulties
encountered with mixed scripts.124 This approach is discussed in more detail in Section 4.4
One view is that the potential for confusion in these cases is not really different from that of
existing similar-appearing domain names, for which it has been suggested that UDRP-based name
conflict resolution is adequate and appropriate. But variations among similar-looking domain names are
such as to generate, potentially, hundreds of possible conflicts with a given character string.
The Joint Engineering Team (JET) guideline model (see Section 4.4.3) addresses this problem by
preventing some large fraction of the potential conflicts, rather than devising remedies for them after they
occur. The JET guidelines take the position that IDN packages are atomic, and that there should be no
mechanism for moving domain names in or out of one once it is created (see Section 4.4 for discussion of
IDNs). Under that model, if a domain name conflict arises in the creation of such a package, the
conflicting (already-registered or reserved) domain name is simply not placed in the new package. But if
the conflicting domain name is later deleted, it does not become part of the later IDN package unless the
domain names associated with that package are explicitly deleted and reregistered. That may or may not
be the best possible model, but the alternatives, such as having domain names appear as reserved in two
or more packages, with a priority order, lead to administrative, policy, or database management
But there are constituencies that oppose such systems, some of them on the grounds that dispute
resolution is adequate and others, perhaps, on the more cynical grounds that letting things go to dispute
resolution permits them to collect registrar and registry fees on the names whether they are valid or not
and encourages even more business in defensive registrations.

Conclusion: The deployment of internationalized domain names introduces new sources of
potential conflict over domain name rights. Reduction of such conflicts through guidelines and
registration policies should be encouraged.


Issue: What is the appropriate balance among the various interests in Whois data?

As noted in Chapter 2, the Whois service began as a vehicle for Internet operators to find and
contact those responsible for the operation of an Internet host when, for example, an operational problem

124 Guidelines are available at <>.

Prepublication Copy--Subject to Further Editorial Correction
arose. However, with the commercialization of the Internet, the Whois service has become an important
and valuable tool for intellectual property owners and is often used by trademark owners to determine the
identity of infringers and cybersquatters. In addition, it is used by law enforcement agencies, such as the
Federal Trade Commission in the United States, to track down the sources of fraudulent or other illegal
uses of the Internet. At the same time, there has been concern about its real and potential exploitation by
marketers and others who find the information about domain name owners valuable. These uses have, in
turn, given rise to significant and strongly held privacy concerns. Thus, while the ability to search the
Whois database has always been limited, because of privacy concerns access and searching of Whois
information have become more and more restricted over time.

5.7.1 Assessment of Whois Data Issues

In the early days of the DNS, there were few, if any, concerns about the misuse of Whois data,
just as ensuring the integrity of DNS data was deemed to be unnecessary. However, the population of
users of Whois data has increased markedly in scale and scope, and assumptions about the good intent of
all users have become unfounded. Furthermore, under the UDRP, giving false Whois information and not
responding to requests for information have led to a presumption of bad faith by the respondent.

For example, when Whois was used as a Unix command, trademark owners were able to retrieve
a wide range of information, including contact information of the domain name registrant and a list of all
domain names registered by one particular registrant. Later, and until 2001, Network Solutions, Inc.
(NSI) allowed Internet users to retrieve a list of up to 50 domain names registered by a particular
registrant, but then changed the maximum number to 10 registrations. Currently, none of the registrars
allow Internet users freely to query their Whois databases to determine which domain names a particular
registrant has registered. Many registrars charge a fee for each request for a list of domain names
registered by one of their registrants. In addition, some of the registrars do not provide a domain name
registrant's e-mail address in the contact information, but instead assign each registrant a generic e-mail
address125 that is linked to the e-mail address the registrant provided in registering its domain name.

Data Accuracy

Whois information can be inaccurate, out of date, or false. Indeed, registrants may provide
fictitious names and addresses and fail to update any of their contact information promptly, if ever.
ICANN's Registrar Accreditation Agreement contractually binds each of its accredited registrars to
investigate and correct any reported inaccuracies in contact information for the domain names they

ICANN established in September 2002 the Whois Data Problem Reports System (WDPRS) to
receive public reports of inaccurate or absent Whois data. The sixth amendment to its MoU with the DOC
requires that ICANN publish an annual report containing an analysis of the received reports. According to
its March 2004 report,126 over the 18-month period from September 2002 through February 2004, the
system received about 24,000 confirmed Whois inaccuracy reports, concerning about 16,000 different
domain names. Of these, 82 percent concerned .com; 13 percent, .net; and 5 percent, .org. (An
enhanced version of the system that will cover the new gTLDs as well as the legacy ones has recently
been launched.) The complaints received by each registrar were generally proportional to the number of
names it registered. On average, each registrar received 4.8 complaints per year per 10,000 names

125 This could cause problems in a UDRP proceeding since the generic e-mail address could be interpreted as false
and, consequently, a contributor to the presumption of bad faith on the part of the respondent.
126 ICANN. 2004. "Community Experiences with the InterNIC Whois Data Problem Reports System." March 13.
Available at <>.

Prepublication Copy--Subject to Further Editorial Correction
managed. Somewhat more than a third of these complaints resulted in the correction of data or the
removal of a domain name.

As a further step to improve Whois data accuracy, ICANN adopted the Whois Data Reminder
Policy (WDRP) on March 27, 2003.127 Since November 2003, all ICANN-accredited registrars must
comply with the WDRP with respect to registrations they sponsor in all top-level domains for which they
are accredited. At least annually, a registrar must present the current Whois information to registrants and
remind them that provision of false Whois information can be grounds for cancellation of their domain
name registration. Registrants must review their Whois data and make any corrections.

Data Privacy

ICANN posted a staff manager's issues report on privacy issues related to Whois on May 13,
2003,128 that spelled out a catalog of the issues, the stakeholders, and their apparent positions on the issues.
The issues concerned the data collected, including its quality, handling, disclosure, and use; the
classification of registrants (i.e., political, commercial, individual); and commercial confidentiality and
rights in data.
The various stakeholders were viewed as placing emphasis on different issues. Non-commercial
users were viewed as focusing on privacy, whereas commercial users were seen as concerned with
accessibility to enforce accountability of uses. The intellectual property interests were understood to stress
the importance of ready access to support investigations of intellectual property abuse, while ISPs support it
to facilitate resolution of network problems and identification of the sources of spam. Registrars and
registries view registrant data as an important business asset that should not be made available to
competitors, while at the same time registrars need to access registrant data of competitors to confirm
authorization of transfers. Registrars and registries both bear the expense of providing the services and,
therefore, have strong incentives to reduce the cost of doing so.
As a consequence of these differences in emphasis among stakeholders, the policy issues
surrounding Whois services (as opposed to the Whois protocol) are often framed in adversarial terms. On
the one hand, trademark holders and their representatives want comprehensive and free access to all
Whois data and would like improvements in Whois services, such as higher quality in the Whois data and
the ability to consolidate data across Whois services more easily. They see Whois data as an essential
resource in the pursuit of those who compromise their trademarks in domain names.129 On the other hand,
those who are concerned about individual privacy highlight the problems that could be associated with
unconstrained access to Whois data--from junk mailers and marketers to those who may use such data to
facilitate more serious, illegal activities such as identity theft.
In recognition of the complexity of the issues and interests involved, the ICANN staff manager's
issues report recommended as the next step the formation of a Whois/privacy steering group in the Generic
Names Supporting Organization (GNSO) to conduct a fact-finding and issues definition process. Following
on the work of that steering group, the Names Council of the GNSO in October 2003 launched three
simultaneous task forces on various aspects of Whois privacy. The council intended to align their
recommendations for submission to the ICANN board.

127 ICANN. 2003. "Whois Data Reminder Policy." June 16. Available at <
128 ICANN. 2003. "Staff Manager's Issues Report on Privacy Issues Related to Whois." May 13. Available at
129 In a UDRP proceeding, for example, trademark owners are often required to show a "pattern of conduct" by the
respondent of registering domain names incorporating the trademarks of other parties. Unless a trademark owner can
guess the domain names registered by the respondent, it can incur considerable costs in obtaining this information since
the respondent may have used several different registrars in registering its domain names or provided slightly different
contact details for each registration.

Prepublication Copy--Subject to Further Editorial Correction
Task Force 1 (TF1) was charged with examining what contractual changes (if any) would be
required to allow registrars and registries to protect domain name holder data from data mining130 for
marketing purposes. Task Force 2 (TF2) was asked to address issues concerning the data to be collected
from registrants, their options to restrict access to the data and be informed of its use, and their ability to
remove certain data elements from public access and receive notice if it is accessed. Task Force 3 (TF3)
was tasked with looking at verification of the data collected, considering both errors and deliberate
falsification. The three task forces presented their preliminary reports at the end of May 2004. They have
been posted on the ICANN Web site for comment.131 Among the significant issues and positions
identified were the following.

Local law.132 In some cases, national privacy laws conflict with the provisions of ICANN's
agreement requiring the registrars to collect and make accessible certain data elements about registrants.
The ICANN registry/registrar agreement should be modified to exempt registrars who obey local law
from the conflicting provisions of the agreement.
Data elements. All of the data elements currently collected are considered by at least some
constituencies to be required, although some constituencies dispute the needs for some of them. No
consensus exists on whether new elements are needed and whether some existing elements should be
made voluntary. The issue is less what should be collected by registries/registrars and more what data
should be made available for public access.
Publication of data. Whois data has a wide range of uses (as discussed above.) It is also
subject to abuses--telemarketing, identity theft, spamming, stalking, and abuse and harassment have been
reported, though not quantified. There is a need to achieve a balance between accessibility and privacy.
Possible approaches include tiered access, in which different types of users would have access to different
subsets of the data; proxy registration services that would substitute third-party for registrant data and
control access to the latter; and the ability of registrants to opt out of publication of certain data on a case-
by-case basis. The latter approach has been adopted by some ccTLDs.
Data mining and marketing.133 If only non-sensitive data (generally, technical information)
were to be available via Whois, it would have little value, be unlikely to be data mined, and have little
impact on privacy. However, to the extent that sensitive data (generally, personal contact information) is
publicly available through registry/registrar Whois services, TF1 members agreed that at a minimum the
requestor of Whois information should be required to identify (and authenticate) itself to the Whois
provider together with its reasons for seeking the data. They left open the issue, however, of whether
notice to the registrant of such a request should be required. They also left open the question of whether
and under what conditions automated access to Whois data could be allowed and to whom. Among the
possibilities would be enabling a restricted license that would provide data to approved requestors for
recognized purposes in human-readable format only. Requestors could be approved generally and
centrally (a white list) or locally and specifically (an individual use list).

130 "Data mining" as used here means the use of computerized techniques to extract data about registrants from
registrar Whois files in large quantities. Often these techniques are designed to overcome specific limitations
imposed by registrars on the number of names that may be requested. The lists are then used for unsolicited mailings
(spam) and other possibly illicit (identity theft) purposes.
131 Links to the preliminary reports are available at <>.
132 For this and the next two items, see the report of TF2 available at <
133 See the report of TF1 available at <>.

Prepublication Copy--Subject to Further Editorial Correction
5.7.2 Whois and Internationalized Domain Names134

Just as it does for the UDRP, the introduction of internationalized domain names (IDNs) raises
technical and institutional issues for the Whois service. For the most part, these are issues about the
languages in which Whois queries will be posed and responded to.
At the basic query level, the current Whois service expects to receive ASCII characters only; it
cannot receive queries in Unicode (which is used to encode the many different character sets of
contemporary human languages), and its responses are similarly in ASCII. But in an internationalized
environment, domain names will not all be written in ASCII (although, as explained in Section 4.4, they
will all be mapped into ASCII strings). This raises the first question: What character sets should be
acceptable in a query? The choices include not only Unicode, but also IDNA puny code (see Section 4.4)
and local character sets, or some combination of them.
Similarly, responses to Whois queries are currently provided in ASCII. This raises the second
question: What language should be acceptable in a response, and how should it be encoded? The choices
of language include the language of the nation in which the registrar or the registrant is located or
English. Or one might permit some "international languages," such as English, Chinese, French, Spanish,
Russian, Arabic, and so on. If the response is to be useful to most questioners on the international
Internet, then would it be reasonable to expect them to have to hire translators? Or should the Whois
registrant be required to list its information in some commonly accepted language? If the language is
other than English, then issues about coding arise that are similar to the question regarding queries. For
example, should Unicode be required and, if so, which encoding form of Unicode? Or should local
character encodings, which might be in much more general use with the particular relevant language or
script, but less easily accessible internationally, be permitted?
A third question arises since IDN practices for complex languages actually create packages of
reserved names (see Section 4.4). In such cases, how much information should Whois provide about
other names in the package in response to a query about one of them?
None of these issues had been resolved by September 2004. However, as IDNs are more widely
adopted, the lack of their early resolution will increase the likelihood of problems arising and the
difficulty of introducing the necessary changes.
On the other hand, the work on a new protocol to replace Whois (see Section 5.7.3) has explicitly
addressed some internationalization issues. Although that work does not address all of the issues raised
above, it at least makes it possible to transmit and receive Unicode characters without somehow encoding
them into ASCII form and, if it is desired to support local character encodings, to construct a framework
for identifying and using them.

Recommendation: The IETF and ICANN should address Whois data internationalization issues
with high priority in order to enable their resolution and implementation of the results together with the
widespread introduction of IDNs.

5.7.3 Conclusions and Recommendation

The issues concerning the accuracy of and access to Whois data engage the interests of many
stakeholders with legitimate but sometimes conflicting interests. They entail actual and potential conflicts
with differing national privacy laws. Furthermore, the ICANN agreements with registrars and registries
obligate them to accept only consensus policies. Consequently, the best way to achieve improvements in
the Whois policies and practices appears to be through the consensus policy development process in
which ICANN is engaged. Attempts by individual governments to impose specific requirements on

134 This section draws on: John Klensin. 2003. "`Whois' Internationalization Issues." ICANN Meeting in Carthage.
October. Available at <>.

Prepublication Copy--Subject to Further Editorial Correction
Whois, such as recent legislative initiatives in the U.S. Congress,135 can interfere with these efforts and
have counterproductive consequences by inducing registrants to find ways to hide their identities.

Conclusion: Legislative or technical initiatives that construe Whois narrowly will not be
productive in the long run and serve only to energize those constituencies that perceive their interests as
being compromised.

The committee agrees that access to Whois data should be viewed as a tiered decision, and not as
a binary decision. Gradations should exist, as they do in local telephone directories where entries are
included by default, but where unlisted numbers can be obtained. Moreover, under certain conditions,
law enforcement officials can obtain an individual's information, even if the individual has opted not to
be included in the public directory. Alternatively, individuals can sometimes embellish their generic entry
(for a fee). Thus, changes to the Whois process need to be conceived in a systematic way that accounts
for the varying legitimate perspectives. The example of local telephone directories is offered for
illustrative purposes only. The committee is not recommending this specific model per se, although the
analogy can also be helpful since personal data (name, address, and phone number) are made publicly
available through printed (and now online) directories, just as they are through Whois services.

Recommendation: Future systems that support Whois data management and access should be
designed to allow for gradations in access while maintaining some degree of free access to Whois
information. The Whois protocol will have to be replaced to accommodate the desired gradations in

135 H.R. 3754. 108th Congress. Fraudulent Online Identities Sanctions Act (FOISA).
136 The IETF had, by October 2004, approved as "standards-track" documents (see Box 3.3) several elements of a
proposed replacement protocol, which is called IRIS and defined by the CRISP Working Group, that will implement
this capability. The protocol also addresses most or all of the other perceived deficiencies of the Whois protocol,
including its inability to deal with non-ASCII characters. More detail on those deficiencies is available in the
statement of requirements for the new protocol: A. Newton. "Cross Registry Internet Service Protocol (CRISP)
Requirements." RFC 3707. February 2004. However, in October 2004 it was still unclear how long it would take
for all the elements to be approved, published, and implemented.

Prepublication Copy--Subject to Further Editorial Correction
Internet Navigation: Emergence and Evolution

As the previous chapters show, the Domain Name System has been a foundation for the rapid
development of the Internet. Domain names appear on the signposts designating origins and destinations
linked by the Internet and in the addresses used by the principal applications traversing the Internet--e-
mail and the World Wide Web. And they have been useful for navigating across the Internet: given a
domain name, many Web browsers will automatically expand it into the Uniform Resource Locator
(URL) of a Web site; from a domain name, many e-mail users can guess the e-mail address of an
addressee. For these reasons, memorable domain names may often acquire high value. Their registrants
believe that searchers can more readily find and navigate to their offerings.
However, as the Internet developed in size, scope, and complexity, the Domain Name System
(DNS) was unable to satisfy many Internet users' needs for navigational assistance. How, for example,
can a single Web page be found from among billions when only its subject and not the domain name in its
URL is known? To meet such needs, a number of new types of aids and services for Internet navigation
were developed.1 While, in the end, these generally rely on the Domain Name System to find specific
Internet Protocol (IP) addresses, they greatly expand the range of ways in which searchers can identify the
Internet location of the resource they seek.
These navigational aids and services have, in return, relieved some of the pressure on the Domain
Name System to serve as a de facto directory of the Internet and have somewhat reduced the importance
of the registration of memorable domain names. Because of these tight linkages between the DNS and
Internet navigation, this chapter and the next ones address--at a high level--the development of the
major types of Internet navigational aids and services. This chapter is concerned with their past
development. The next chapter deals with their current state. And the final chapter on Internet navigation
considers the technological prospects and the institutional issues facing them.
After describing the distinctive nature of Internet navigation, this chapter traces the evolution of a
variety of aids and services for Internet navigation. While its primary focus is on navigating the World
Wide Web, it does not cover techniques for navigation within Web sites, which is the subject of
specialized attention by Web site designers, operators, and researchers.2


Navigation across the Internet is sometimes compared to the well-studied problem of readers
navigating through collections of printed material and other physical artifacts in search of specific
documents or specific artifacts. (See the Addendum to this chapter: "Searching the Web Versus Searching
Libraries"). That comparison illustrates the differences in the technical and institutional contexts for
Internet navigation. Some purposes for Internet navigation are very similar to library environments and
rely on the same tools, while navigation for other purposes may be performed quite differently via the
Internet than in a library. The multiple purposes and diverse characteristics listed below combine to make
navigating to a resource across the Internet a much more varied and complex activity than those

1 The difference between a navigation aid and a navigation service is one of degree. A navigation service, such as
the offerings of a search engine service provider, are more elaborate and extensive than those offered by a
navigation aid, such as the bookmark feature of a Web browser.
2 See, for example, Merlyn Holmes, 2002. Web Usability & Navigation: A Beginner's Guide. Berkeley, Calif.:
McGraw-Hill/Osborne; and Louis Rosenfeld and Peter Morville. 2002. Information Architecture for the World
Wide Web: Designing Large Scale Sites,
2nd ed. Sebastopol, Calif.: O'Reilly & Associates.

Prepublication Copy--Subject to Further Editorial Correction
previously encountered. The library examples provide a point of reference and a point of departure for
discussion in subsequent chapters.

6.1.1 Vast and Varied Resources for Multiple Purposes

First, the Internet connects its users to a vast collection of heterogeneous resources that are used
for many purposes, including the dissemination of information; the marketing of products and services;
communication with others; and the delivery of art, entertainment and a wide range of commercial and
public services. The kinds of resources connected to the Internet include:

Documents that differ in language (human and programming), vocabulary (including words,
product numbers, zip codes, latitudes and longitudes, links, symbols, and images), formats (such as the
Hypertext Markup Language (HTML), Portable Document Format (PDF), or Joint Photographic Experts
Group (JPEG) format), character sets, and source (human or machine generated).
Non-textual information, such as audio and video files, and interactive games. The volume of
online content (in terms of the number of bytes) in image, sound, and video formats is much greater than
most library collections and is expanding rapidly.
Transaction services, such as sales of products or services, auctions, tax return preparation,
matchmaking, and travel reservations.
Dynamic information, such as weather forecasts, stock market information, and news, which
can be constantly changing to incorporate the latest developments.
Scientific data generated by instruments such as sensor networks and satellites are
contributing to a "data deluge."3 Many of these data are stored in repositories on the Internet and are
available for research and educational purposes.
Custom information constructed from data in a database (such as product descriptions and
pricing) in response to a specific query (e.g., price comparisons of a product listed for sale on multiple
Web sites).

Consequently, aids or services that support Internet navigation face the daunting problem of finding and
assigning descriptive terms to each of these types of resource so that it can be reliably located. Searchers
face the complementary problem of selecting the aids or services that will best enable them to locate the
information, entertainment, communication link, or service that they are seeking.

6.1.2 Two-sided Process

Second, Internet navigation is two-sided: it must serve the needs both of the searchers who want
to reach resources and of the providers who want their resources to be found by potential users.
From the searcher's perspective, navigating the Internet resembles to some extent the use of the
information retrieval systems that were developed over the last several decades within the library and
information science4 and computer science communities.5 However, library-oriented retrieval systems,

3 See Tony Hey and Anne Trefethen. 2003. The Data Deluge: An e-Science Perspective. Wiley; and Fran Berman,
Geoffrey Fox, and Anthony J. G. Hey (editors). 2003. Grid Computing: Making the Global Infrastructure a
4 For an overview see Elaine Svenonius. 2000. The Intellectual Foundation of Information Organization
Cambridge, Mass.: MIT Press; and Christine L. Borgman. 2000. From Gutenberg to the Global Information
Infrastructure: Access to Information in the Networked World.
Cambridge, Mass.: MIT Press.
5 For an overview see Ricardo Baeza-Yates and Berthier Ribiero-Neto. 1999. Modern Information Retrieval.
Boston: Addison-Wesley Publishing Co. and Karen Sparck Jones and Peter Willett, editors. 1997. Readings in
Information Retrieval.
San Francisco: Morgan Kaufmann Publishers Inc. For typical examples of early work on

Prepublication Copy--Subject to Further Editorial Correction
reflecting the well developed norms of librarians, were designed to describe and organize information so
that the users could readily find exactly what they were looking for. In many cases, the same people and
organizations were responsible both for the design of the retrieval systems and for the processes of
indexing, abstracting, and cataloging information to be retrieved. In this information services world, the
provider's goal was to make description and search as neutral as possible, so that every document relevant
to a topic would have an equal chance of being retrieved.6 While this goal of retrieval neutrality has
carried over to some Internet navigation services and resource providers, it is by no means universal.
Indeed, from the perspective of many resource providers, particularly commercial providers, attracting
users via the Internet requires the application to Internet navigation of non-neutral marketing approaches
deriving from advertising and public relations as developed for newspapers, magazines, radio, television,
and yellow-pages directories.7 Research on neutral, community-based technology for describing Internet
resources is an active area in information and computer science and is a key element of the Semantic
For commercial providers, therefore, the challenge is how to identify and reach--in the complex,
diverse, and global audience accessible via the Internet--potential users who are likely to be interested (or
can be made interested) in the provider's materials. That is done in traditional marketing and public
relations through the identification of media and places (television or radio programs, magazines,
newspapers in specific locations) that an audience with the desired common characteristics (for example,
18- to 24-year-old males) frequents. Similar approaches can be applied on the Internet (see Section 7.2.2),
but unlike the traditional media, the Internet also offers providers the distinctive and extremely valuable
opportunity to capture their specific audience during the navigation process itself, just when they are
searching for what the provider offers; for example, by paying to be listed or featured in a navigation
service's response to specific words or phrases. (See "Monetized Search" in Section 7.1.7.) Marketers
have found ways to use the specific characteristics of the Internet,9 just as they have developed methods

information retrieval systems see, for example George Schecter, editor. 1967. Information Retrieval--A Critical
. Washington, D.C.: Thompson Book Company. For work on retrieval from large databases, see the
proceedings of the annual text retrieval conference (TREC), currently sponsored by National Institute for Standards
and Technology and the Advanced Research and Development Activity, which are available online at
6 See Svenonius (2000).
7 See, for example, John Caples and Fred E. Hahn. 1998. Tested Advertising Methods, 5th edition. New York:
8 Bradley, E., N Collins, WP Kegelmeyer (2001). "Feature characterization in scientific datasets."
Advances in Intelligent Data Analysis. 4th International Conference, IDA 2001. Proceedings (Lecture
Notes in Computer Science Vol.2189). Springer-Verlag. 2001, Berlin, Germany.: 1-12.
Brilhante, V. (1999). "Using formal metadata descriptions for automated ecological modeling."
Environmental Decision Support Systems and Artificial Intelligence. AAAI Workshop. AAAI Press.
1999, Menlo Park, CA, USA.: 90-95.
Hovy, E. (2003). "Using an ontology to simplify data access." Communications of the ACM 46(1): 47-
OWL Web Ontology Language Guide: W3C Recommendation (10 February 2004). 2004: 24 November.
Wariyapola, P., Abrams SL., Robinson AR., Streitlien K., Patrikalakis NM., Elisseeff P., Schmidt H.
(1999). "Ontology and metadata creation for the Poseidon distributed Coastal Zone Management
System." Proceedings IEEE Forum on Research and Technology Advances in Digital Libraries. IEEE
Comput. Soc. 1999, Los Alamitos, CA, USA.: 180-189.
9 See, for example, Joe Cappo. 2003. The Future of Advertising: New Media, New Clients, New Consumers in the
Post-Television Age.
New York: McGraw-Hill; and Barbara Cox and William Koelzer. 2004. Internet Marketing.
New York: Prentice-Hall.

Prepublication Copy--Subject to Further Editorial Correction
appropriate for each new medium.10 This has led, for example, to the establishment of companies that are
devoted to finding ways to manipulate Internet navigation services to increase the ranking of a client's
Web site and, in response, to the development of countermeasures by the services. (See "Search Engine
Marketing and Optimization" in Section 7.1.7.)
For non-commercial resources, the situation is somewhat different, since the providers generally
have fewer resources and may have less incentive to actively seek users, at least to the extent of paying
for Web advertising or search engine marketing. At the same time, the existence of a specific non-
commercial resource may be well known to the community of its potential users. For example, the
members of a scholarly community or a non-profit organization are likely to be aware of the Internet
resources relevant to their concerns. Those new to a community or outside of it are dependent upon
Internet navigation tools to locate these resources.
The consequence of the Internet's two-sided characteristic is that Internet navigation is a complex
interplay of the interests of both the searchers for and the providers of resources. On the Internet, the
librarian's ideal of neutral information retrieval often confronts the reality of self-interested marketing.

6.1.3 Complexity and Diversity of Uses, Users, and Providers

Third, the complexity and diversity of uses of resources on the Internet, of their users, and of their
providers, significantly complicate Internet navigation. It becomes a multi-dimensional activity that
incorporates behaviors ranging from random browsing to highly organized searching and from
discovering a new resource to accessing a previously located resource.11
Studies in information science show that navigation in an information system is simplest and
most effective when the content is homogeneous, the purposes of searching are consistent and clearly
defined, and the searchers have common purposes and similar levels of skills.12 Yet, Internet resources
per se often represent the opposite case in all of these respects. Their content is often highly
heterogeneous; their diverse users' purposes are often greatly varied; the resources the users are seeking
are often poorly described; and the users often have widely varying degrees of skills and knowledge.
Thus, as the resources accessible via the Internet expand in quantity and diversity of content, number and
diversity of users, and variety of applications, the challenges facing Internet navigation become even
more complex.
Indeed, prior to the use of the Internet as a means to access information, many collections of
information resources, whether in a library or an online information system, were accessed by a more
homogenous collection of users. It was generally known when compiling the collection whether the
content should be organized for specialists or lay people, and whether skill in the use of the resource
could be assumed. Thus, navigation aids, such as indexes or catalogs, were readily optimized for their
specific content and for the goals of the people searching them. Health information in a database intended
for searching by physicians could be indexed or cataloged using specific and highly detailed terminology
that assumed expert knowledge. Similarly, databases of case law and statute law assumed a significant
amount of knowledge of the law. In fields such as medicine and law, learning the navigation tools and the
vocabularies of the field is an essential part of professional education. Many such databases are now
accessible by specialists via the Internet and continue to assume a skillful and knowledgeable set of users,

10 It should be noted that the Internet, by making the marginal cost of an e-mail message extremely low, has also
enabled providers to conduct a non-discriminating search for potential users by broadcasting spam e-mail. Although
this might be considered a variant of Internet navigation, where the provider actively advertises its location to a vast
audience whose members may or may not be interested, its one-sided benefits and frequent use for dishonest or
illegal purposes disqualify it for inclusion in this report, which focuses on searcher-beneficial navigation aids and
11 See Shan-ju Chang and Ronald E. Rice, "Browsing: A Multidimensional Framework," Annual Review of
Information Science and Technology
28:231-276 (1993).
12 See Borgman (2000).

Prepublication Copy--Subject to Further Editorial Correction
even though in some cases they are also accessible by the general public. With the growth in Internet use,
however, many more non-specialist users have ready access to the Web and are using it to seek medical,
legal, or other specialized information. The user community for such resources is no longer well defined.
Few assumptions of purpose, skill level, or prior knowledge can be made about the users of a Web
information resource. Consequently, it is general purpose navigation aids and services and less
specialized (and possibly lower quality) information resources that must serve their needs.

6.1.4 Lack of Human Intermediaries

Fourth, the human intermediaries who traditionally linked searchers with specific bodies of
knowledge or services--such as librarians, travel agents, and real estate agents--are often not available to
users as they seek information on the Internet. Instead, users generally navigate to the places they seek
and assess what they find on their own, relying on the aid of digital intermediaries--the Internet's general
navigation aids and services, as well as the specialized sites for shopping, travel, job hunting, and so on.
Human intermediaries' insights and assistance are generally absent during the navigation process.
Human search intermediaries help by selecting, collecting, organizing, conserving, and
prioritizing information resources so that they are available for access.13 They combine their knowledge
of a subject area and of information seeking behavior with their skills in searching databases to assist
people in articulating their needs. For example, travelers have relied upon travel agents to find them the
best prices, best routes, best hotels, and to provide services such as negotiating with hotels, airlines, and
tour companies when things go wrong. Intermediaries often ask their clients about the purposes for which
they want information (e.g., what kind of trip the seekers desire and how they expect to spend their time;
what they value in a home or neighborhood; or what research questions their term paper is trying to
address), and elicit additional details concerning the problem. These intermediaries also may help in
evaluating content retrieved from databases and other sources by offering counsel on what to trust, what
is current, and what is important to consider in the content retrieved.
With the growth of the Internet and the World Wide Web, a profound change in the nature of
professional control over information is taking place. Travel agents and real estate agents previously
maintained tight control over access to fares and schedules and to listings of homes for sale. Until
recently, many of these resources were considered proprietary, especially in travel and real estate, and
consumers were denied direct access to their content. Information seekers had little choice but to delegate
their searches to an expert--a medical professional, librarian, paralegal, records analyst, travel agent, real
estate agent, and so on. Today, travel reservation and real estate information services are posting their
information on the Internet and actively seeking users. Specialized travel sites, such as
and, help the user to search through and evaluate travel options. Similar sites serve
the real estate market. Travel agents that remain in business must get their revenue from other value-
added services, such as planning customized itineraries and tours and negotiating with brokers. Although
house hunters now can do most of their shopping online, in most jurisdictions they still need real estate
agents with access to house keys to show them properties and to guide them in executing the legal
transactions. Libraries have responded to the Web by providing "virtual reference services" in addition to
traditional on-site reference services.14 Other agencies, including the U.S. Department of Education, have
supported the creation of non-library based reference services that use the Internet to connect users with

13 See Chapter 7, "Whither, or Wither, Libraries?" in Borgman (2000).
14 See, for example, the more than 600 references about such services in: Bernie Sloan. "Digital Reference Services
Bibliography." Graduate School of Library and Information Science, University of Illinois at Urbana-Champaign.
November 18, 2003. Available online at <>.

Prepublication Copy--Subject to Further Editorial Correction
"people who can answer questions and support the development of skills" without going through a library

6.1.5 Democratization of Information Access and Provision

Fifth, the Internet has hugely democratized and extended both the offering of and access to
information and services. Barriers to entry, whether cost or credentials, have been substantially reduced.
Anyone with modest skill and not much money can provide almost anything online, and anyone can
access it from almost anywhere. Rarely are credentials required for gaining access to content or services
that are connected to the public Internet, although paid or free registration may be necessary to gain
access to some potentially relevant material. Not only commercial and technical information are openly
accessible, but the full range of political speech, of artistic expression, and of personal opinion are readily
available on the public Internet--even though efforts are continually being made to impose restrictions on
access to some materials by various populations in a number of countries.16

In a great many countries, anyone can set up a Web site--and many people do. The freedom of
the press does not belong just to those who own one, now nearly anyone can have the opportunity to
publish via a virtual press--the World Wide Web.17 Whether or not what they publish will be read is
another matter. That depends on whether they will be found, and, once found, whether they can provide
content worthy of perusal--at least by someone.
For the most part, in many places, provision of or access to content or services is uncensored and
uncontrolled. On the positive side, the Internet enables access to a global information resource of
unprecedented scope and reach. Its potential impact on all aspects of human activity is profound.18 But
the institutions that select, edit, and endorse traditionally published information have no role in
determining much of what is published via the Internet. The large majority of material reachable via the
Internet has never gone through the customary editing or selection processes of professional journals or of
newspapers, magazines and books.19 So in this respect as well, there has been significant
disintermediation, leaving Internet users with relatively few solid reference points as they navigate
through a vast collection of information of varying accuracy and quality. In response to this widely

15 For example, the Virtual Reference Desk is "a project dedicated to the advancement of digital reference." See
<>. This service is sponsored by the U.S. Department of Education.
16 These range from the efforts of parents to prevent their children from accessing age-inappropriate sites to those
made by governments to prevent their citizens from accessing politically-sensitive sites. The current regulations
placed on libraries in the U.S. to filter content are available online at
<>. See also, the increasing sophistication of
filtering efforts in China in David Lee. 2002. "When the Net Goes Dark and Silent," South China Morning Post.
October 2. Available online at <>. For
additional information see, Jonathan Zittrain and Benjamin Edelman. 2002. Empirical Analysis of Internet Filtering
in China
. Berkman Center for Internet & Society, Harvard University. Available online at
17 The rapid growth in the number of blogs (short for Web logs) illustrates this. According to "How Much
Information?" there were 2.9 million active Weblogs in 2003. See Lyman, Peter and Hal R. Varian, "How Much
Information?" 2003. Retrieved from <> on October 27,
18 See Borgman (2000) and Thomas Friedman. 2003. "Is Google God?" New York Times, June 29.
19 However, it must be acknowledged that many "traditional" dissemination outlets (e.g., well-known media
companies) operate Web sites that provide material with editorial review.

Prepublication Copy--Subject to Further Editorial Correction
acknowledged problem, some groups have offered evaluations of materials on the World Wide Web.20
One example is the Librarian's Index to the Internet,21 whose motto is "Information You Can Trust."

6.1.6 Lack of Context or Lack of Skill

Sixth, most general purpose Internet navigation services can assume nothing about the context of
a search beyond what is in the query itself. The circumstances of the person searching the Internet--who
could be anyone, searching for almost any purpose, from almost any place--are generally not known or,
if known, not used.22 While an unskilled user planning a European trip who types "Paris" into a
navigation service may expect to receive back only information on the city of Paris, France, or if studying
the Iliad may expect to find out about the Greek hero, the navigation service has no knowledge of that
context. If it relies on the text alone, it may return information about both, as well as information about
sources for plaster of Paris, the small town in Texas, movies set in Paris, and people with the first or
family name of Paris.
While any general purpose navigation service would be unable to ascertain which of those
specific answers was desired, the person initiating an Internet request always knows its context, and if
experienced or trained, should be able to incorporate some of that information in the query. An
experienced searcher could incorporate context by expanding the query to "Paris France," "Paris and
Iliad," "Paris Texas," or "Plaster of Paris." So the lack of context in many navigation requests is closely
related to the user's level of training or experience in the use of navigation services.23
The incorporation of context is also related to the degree to which the navigation service is itself
specialized. When people searched in traditional information sources, the selection of the source usually
carried information about the context for a search (e.g., seeking a telephone number in the White Pages
for Manhattan in particular, or looking for a home in the multiple listing service for Boston specifically).
Furthermore, there were often intermediaries who could obtain contextual information from the
information seeker and use that to choose the right source and refine the information request. Although
the former approach is also available on the Internet through Internet white pages and Internet real estate
search sites, the latter is just developing through the virtual reference services mentioned earlier.
General purpose navigation services are exploring a variety of mechanisms for incorporating
context into search. For example, both Yahoo! and Google now allow local searches, specified by adding
a location to the search terms. The site-flavored Google Search customizes searches originating at a Web

20 See, for example, The Information Quality WWW Virtual Library at <
InfoQuality.html> and Evaluating Web Sites: Criteria and Tools at
21 "Librarians' Index to the Internet (LII) is a searchable, annotated subject directory of more than 12,000 Internet
resources selected and evaluated by librarians for their usefulness to users of public libraries. LII is used by both
librarians and the general public as a reliable and efficient guide to Internet resources." Quoted from
<>. Visited May 2, 2004.
22An online advertising company, DoubleClick, launched a service in 2000 to track people's Internet usage in order
to serve ads based on personal taste. After considerable controversy, especially from federal regulators and privacy
advocates, and an inability to develop an adequate market, the profiling service was terminated at the end of 2001.
See Stefanie Olsen, 2002, "DoubleClick Turns Away from Ad Profiles," c/net, January 8, available online
at <>. However, Yahoo! indicates in its privacy policy that it does use
information provided by those who register at its site "to customize the advertising and content you see, fulfill your
requests for products and services...".
23 See, for example, Steven Johnson. 2003. "Digging for Googleholes,", July 16. Available online at
<>; Case, Donald O., Looking for Information: A Survey of Research on
Information Seeking, Needs, and Behavior
. 2002, San Diego, Calif.: Academic Press; and Solomon, Paul,
"Discovering Information in Context," pp. 229-264 in Annual Review of Information Science and Technology,
Blaise Cronin, editor. 2002. Medford, N.J.: Information Today, Inc.

Prepublication Copy--Subject to Further Editorial Correction
site to return search results based on a profile of the site's content. Other recent search engines such as
Vivisimo ( return search results in clusters that correspond to different contexts--for
example, a search on the keyword `network' returns one cluster of results describing `cable networks,'
another on `network games,' etc.24
Still, issues of context complicate Internet navigation because the most widely used navigation
services are general purpose--searching the vast array of objects on the Internet with equal attention--
and because many users are not experienced or trained and have no access to intermediaries to assist

6.1.7 Lack of Persistence

Seventh, there is no guarantee of persistence for material at a particular location on the Internet.
While there is no reason to believe that everything made accessible through the Internet should persist
indefinitely, there are a great many materials whose value is such that many users would want to access
them at the same Internet address at indefinite times in the future. For example, throughout this report
there are two kinds of references: those to printed materials--books, journals, newspapers; and those to
digital resources, that are located via Web pages. There is a very high probability that whenever this
report is read, even years after its publication, that every one of the referenced printed materials will be
accessible in unchanged form through (though not necessarily in) any good library (at least in the United
States). There is an equal certainty that whenever this report is read, even in the year of its publication,
some of the referenced Web pages either will no longer be accessible or will no longer contain the precise
material that was referenced. The proportion of missing or changed references will increase over time. To
the extent that it would be valuable to locate the referenced Web material, there is a problem of
persistence (see Box 6.1) of resources on the Internet.25

BOX 6.1
Discovery, Retrieval, and Persistence

Navigation is a two-step process: discovery and then retrieval. The discovery process involves
identifying the location of the desired material on a Web site, an intranet, a database, or otherwise. The
retrieval process involves obtaining the located material at the identified URL. (See Box 6.2.) Discovery
may find dynamically generated pages for which only a transient URL exists, such as driving directions
between two locations or an online order, or a constant URL that has ever changing information, such as a
weather service. It may not be possible or even feasible (last year's weather forecast) to retrieve the same
pages again. Thus, not even an hyperlink to that material can be embedded in other pages nor can the
page be bookmarked for future reference.
Even if the resource is itself static, such as the text of a research report, it may be moved,
invalidating the URL. More problematically, it may be necessary to decide whether what is retrieved a
second time (even if it is in the same place) is the same as what was retrieved previously. Unless the
original object and the (possibly) new one are identical bit-by-bit, the question of whether the two are "the
same," or identical, raises long-standing questions that are typically very subjective and/or contextual: Is
a translation into another language the same as the original? If a document is reformatted, but all of the
words are the same, is it identical to the original? If a document is compressed, or transposed into a

24 See Gaither, Chris. June 19, 2004. "Google Offers Sites its Services," Los Angeles Times, p. c2. Also see
<>. Visited June 18, 2004.
25 One example of the lack of persistence of important information is U.S. Government documents published only
on the Web. The U.S. Government Printing Office has begun an effort to find and archive such electronic
documents. See Florence Olsen. 2004. "A Crisis for Web Preservation." June 21. Federal Computer Week.
Available online at <>.

Prepublication Copy--Subject to Further Editorial Correction
different character code, or adapted to a different version of a viewer program, is it still the same
document? Is a second edition an acceptable match to a request for the first edition? And so on. For
each of these questions, the correct answer is probably "sometimes," but "sometimes" is rarely a
satisfactory answer, especially if it is to be evaluated by a computer system.
If things are abstracted further so that, instead of discovering a URL directly, the user utilizes an
updatable bookmark that incorporates a reference to the discovery processes, the meaning of persistence
becomes even more ambiguous. Such a bookmark, intended to reference a list of restaurants in a
particular city, might upon being used, not only discover a newer version of the same list, but a newer,
more comprehensive, list from a different source and at an entirely different location on the network. So
what is persistent in this case is not the answer, but the question.
These are arguably not new conundrums--analogies occur with editions of books and
bibliographies, but the opportunities on the Internet for faster turnover and for changes of finer
granularity, as well as the ability to update search indexes at very high frequency, make the meaning and
achievement of Internet persistence more complex and ambiguous and difficult to understand for the
casual user.

The problem of persistence arises because material connected to the Internet is not generally
governed by the conventions, standards, and practices of preservation that evolved with regard to printed
material. The material may change in many ways and for many reasons. It may remain the same, but the
location may change because of a change of computers, or a redesign of a Web site, or a reorganization of
files to save space, or a change in Internet service providers (ISPs).26 Or it may be replaced at the same
location by an improved version or by something entirely different but with the same title. It may
disappear completely because the provider stops operating or decides not to provide it any more or it may
be replaced by an adequate substitute, but at a different location. Thus, navigation to an item is often a
one-time event. Incorporating a reference or a link to that item in another document is no guarantee that a
future seeker will find the same item at the same location. Nor is there any guarantee that something once
navigated to will be there when the recorded path is followed once more or that it will be the same as it
was when first found.
The problem is exacerbated for material on the World Wide Web by its structure--a hyperlinked
web of pages. Even if a specific Web page persists indefinitely, it is unlikely that the many pages to
which it links will also persist for the same period of time, and so on along the path of linked pages. So to
the extent that the persistent page relies upon those to which it links, its persistence is tenuous.

6.1.8 Scale

Eighth, the Internet--in particular, the World Wide Web--vastly exceeds even the largest
traditional libraries in the amount of material that it makes accessible.27 In 2003, the World Wide Web
was estimated to contain 170 terabytes of content in its surface, readily accessible--public--sites. This
compares with an estimated 10 terabytes of printed documents in the Library of Congress (which also

26 Thomas Phelps and Robert Wilensky of the University of California at Berkeley have suggested a way to
overcome the difficulties caused by shifting locations. They add a small number of words carefully selected from
the page to its URL. If the URL is no longer valid, the words can be used in a search engine to find the page's new
location. See Thomas Phelps and Robert Wilensky. 2000. "Robust Hyperlinks: Cheap, Everywhere, Now,"
Proceedings of Digital Documents and Electronic Publishing. September. Munich: Springer-
Verlag.RobustHyperlinks: Cheap, Eve
27 Because the Web contains a large amount of "format overhead" and non-reliable information, this comparison
was disputed for the Web of 2000 (25-50 terabytes estimated) by researchers from OCLC. See Edward T. O'Neill, et
al. 2003. "Trends in the Evolution of the Public Web 1998-2002," D-Lib Magazine. April. Vol. 9, No. 4.

Prepublication Copy--Subject to Further Editorial Correction
contains many terabytes of material in other media). In addition, the "dark" Web is estimated to contain
400 to 450 times as much content in its databases and in other forms inaccessible to search engines,28 a
ratio that may be increasing. (See in Section 7.1.7 the discussion titled "The Deep, Dark, and Invisible

6.1.9 The Sum of the Differences

For all the reasons described above, navigation on the Internet is different from navigating through
traditional collections of documents; finding information or services through the use of knowledgeable
intermediaries; locating the television audience for a commercial or political message; or reaching the
readers for an essay. Navigating the Internet is even more than the sum of all those differences over a
much larger extent and variety of resources and a much greater number and diversity of users and

Conclusion: Finding and accessing a desired resource via the Internet poses challenges that are
substantially different from the challenges faced in navigating to resources in non-digital, non-networked
environments because of differences in content, purpose, description, user community, institutional
framework, context, skill, persistence, and scale.

Fortunately, the Internet has also served as the infrastructure for the development of new means for
responding to these challenges. As the following section and the next chapter will show, a wide range of
navigation aids and services now permit large segments of the Internet to be traversed rapidly and
efficiently in ways previously unimaginable, providing ready access to a vast world of human knowledge
and experience to users across the globe and opening an international audience to purveyors of content
and services no matter where they may be located.


Aids and services to assist Internet navigation have had to evolve steadily to keep pace with the
growth in the scale, scope, and complexity of material on the Internet.29
In the first years of the Internet,30 the 1970s and 1980s, information was accessed through a two-
step process: first, the host on which a desired resource resided was found,31 then the desired file was
found.32 A good example of this "host oriented" navigation is the File Transfer Protocol (FTP).
Retrieving a file using FTP required that the user (using FTP client software) already know the name of
the host (which would be using FTP server software), on which the file was stored,33 have a login name

28 These data are taken from Lyman and Varian (2003).
29 The early period of rapid development of Internet navigation aids was not well documented, nor are there many
good references. Many of the critical developments are included in this brief history, but it is not intended to be
30 In this section, the "Internet" also includes the ARPANET. See Chapter 2.
31 While a few protocols did not explicitly name a host, in most cases that meant that the host name was
implicit in the service being requested (e.g., running a program that queried a central database that resides
only on the Network Information Center (NIC) computer implies the host name of interest).
32 At a technical level, this still occurs, but is less visible to the user. For example, a URL contains a domain (host)
name, and the client (browser in the case of the Web) uses some protocol (usually the Hypertext Transfer Protocol
(HTTP) in the case of the Web) to open a connection to a specified host. The rest of the URL is then transmitted to
that host, which returns data or takes other action.
33 Once the desired FTP server was located, the FTP protocol eventually included capabilities to travel down
directory hierarchies to find the desired files. (At the time FTP was designed, the protocol did not contain any

Prepublication Copy--Subject to Further Editorial Correction
and password for that host (unless anonymous login was permitted), know a filename, and sometimes
have other information such as an account name for computer time billing.34 The File Transfer Protocol
made it possible to retrieve data from any FTP server to which one was connected by a network. File
names and server locations could be obtained in various ways--from friends, from newsgroups and
mailing lists, from files that identified particular resources, or from a few archive servers that contained
very large collections of files with locally created indexing. However, such ad hoc and labor intensive
processes do not scale well. More automated processes were needed as the number of Internet users and
topics of interest increased substantially.

6.2.1 Aiding Navigation via the Internet

The first response to the need for automated processes was Archie,35 a selective index of those
files thought to be "interesting" by those who submitted them to the index and of the FTP servers on
which they were located. When Archie was first released in 1989, many people were skeptical of the idea
of having an index of such large size on one computer, but Archie was a success. It consisted of three
components: the data gatherer, the index merger, and the query protocol. The gathering of data was not
done by brute force. The company that took over operation of the Archie service (Bunyip Information
Systems) relied on licensed Archie server operators all over the world (who knew which FTP archives
were reliable) to run a data gatherer that indexed local files by searching each local site.36 Each local
gatherer passed the partial index upstream and eventually the partial indexes were merged into a full
index having fairly simple functionality, primarily limited to single-level alphabetic sorting. This
"collaborative gathering" process saved time and minimized the bandwidth required. A special protocol
allowed clients to query the full index, which was replicated on servers around the world, to identify the
FTP server that contained the requested file.
Archie still required users to employ FTP to download the files of interest for viewing on their
local computers. The next aid, Gopher, was developed as a protocol for viewing remote files, which were
in a specified format, on a user's local computer. It became the first navigation aid on the Internet that
was easily accessible by non-computer science specialists.37 The Gopher "browser" provided users with
the capability to look at the content of a file on the computer that stored it, and if the file included a
reference to another file, "click" directly on the link to see the contents of the second file. Gopher
included some metadata38 so, for example, when a user clicked on a link, the user's client software would
automatically check the metadata, choose the desired document format, and start the appropriate program
to deal with the data format.

provision for dealing with hierarchical files--that was added somewhat later.) User names and the account
command were, on many systems, the primary mechanism for providing context for the file names. That is, "user"
and "acct" were navigational commands as much as they were authentication and authorization commands.
34 Frequently, some documentation about the content of directories was provided in files with the file name of
"readme.txt" or some variation thereof, such as AAREAD.ME to take advantage of alphabetic sorting when the
directory was retrieved.
35 "Archie" is not an acronym, but is a shortening of the word "archive" to satisfy software constraints. Peter
Deutsch, Alan Emtage, Bill Heelan, and Mike Parker at McGill University in Montréal created Archie.
36 The burden this imposed on the FTP servers lead the operators of many of them to create and keep up-to-date an
Archie-usable listing of the server in its root directory and to keep those listings up-to-date, obviating the need for
Archie to search the server.
37 Gopher, created in 1991 by Marc McCahill and his technical team at the University of Minnesota, is not an
acronym, but is named after the mascot at the University of Minnesota. See "SearchDay," Chris Sherman, February
6, 2002, Number 198, available online at <>.
38 Metadata are data that describe the characteristics or organization of other data. See: Murtha Baca. 1998.
Introduction to Metadata: Pathways to Digital Information. Los Angeles: Getty Information Institute.

Prepublication Copy--Subject to Further Editorial Correction
To overcome Gopher's limitation to files on a single computer, a central search aid for Gopher
files--Veronica--was created in 1992.39 It did for Gopher what Archie did for FTP. Veronica was the
first service that used brute-force software robots (software that automatically searched the Internet) to
collect and index information and, therefore, could be seen as the forerunner of Web search engines. In
1993, Jughead added keywords and Boolean search capabilities to Veronica.40
The next step in the evolution of Internet navigation aids was the wide area information server
(WAIS) that enabled search using word indices of specified files available on the Internet.41 The WAIS
search engine received a query, sought documents relevant to the question in its database by searching the
word indices, and returned a list of documents ordered by estimated relevance to the user. Each document
was scored from 1 to 1000 according to its match to the user's question as determined by how many of
the query words it contained, their importance in the document, etc. WAIS did not index the entire
Internet, but only specific files and servers on it. At its peak, WAIS linked up to 600 databases,
worldwide.42 It was used, for example, by Dow Jones to create a fully indexed online file of its
publications. WAIS was an important step beyond FTP, Gopher, and Archie (and friends) because it built
upon known information retrieval methods43 and standards, including the Z39.50 "search and retrieve"
standard as its key data representation model.44
FTP with Archie, Gopher with Veronica and Jughead, and WAIS demonstrated three ways to
index and find information on the Internet.45 They laid the foundation for the subsequent development of
aids for navigating the World Wide Web.

6.2.2 Aiding Navigation Through the World Wide Web

While those early navigation aids were being developed, a new way of structuring information on
the Internet46--the World Wide Web (WWW)--was under development at the European Organization for

39 Veronica (Very Easy, Rodent-Oriented, Net-Wide Index to Computerized Archives) was created in November
1992 by Fred Barrie and Stephen Foster of the University of Nevada System Computing Services group.
40 Jughead (Jonzy's Universal Gopher Hierarchy Excavation And Display) was created by Rhett "Jonzy" Jones at
the University of Utah computer center. The names of these two protocols were derived from characters in Archie
comics. See: <>.
41 WAIS was developed between 1989 and 1992 at Thinking Machines Corporation under the leadership of
Brewster Kahle. Unlike Archie (and friends), Gopher, and the Web, WAIS was firmly rooted in information
sciences approaches and technology.
42 Richard T. Griffiths. 2002. Chapter Four: Search Engines in History of the Internet. Leiden University,
Netherlands. See <>.
43 Many of the navigation aids and services described in this history drew inspiration and methodology from the
long line of research in information science, as well as computer science. It has not been possible to give full credit
to those antecedents in this brief history.
44 For information about the Z39.50 protocol, see National Information Standards Organization, Z39.50 Resource
Page, Available at <> and Needleman, Mark (2000). "Z39.50­A
Review, Analysis and Some Thoughts on the Future," Library HiTech 18(2), 158-65, <http://www.biblio->, visited June 23, 2004.
45 See Michael F. Schwartz, Alan Emtage, Brewster Kahle, and B. Clifford Neuman. 1992." A Comparison of
Internet Discovery Approaches," Computing Systems 5(4).
46 The Web's structure is hypertext, which was anticipated in the form of "associative indexing" by Vannevar Bush
in a famous article, "As We May Think," published in the Atlantic Monthly in July 1945. The term "hypertext" was
coined by Theodor (Ted) Nelson in the early 1960s. Nelson spent many years publicizing the concept and
attempting early implementations. Douglas C. Engelbart's NLS/Augment project at the Stanford Research Institute
(which became known as SRI International in 1977) first demonstrated a hypertext system in 1968. Apple Computer
introduced HyperCard, a commercial implementation of hypertext for the Macintosh, in 1987. The practical
implementation of a hypertext data structure on the Internet at a time when computer capacity and network speed
were finally sufficient to make it practical was CERN's contribution.

Prepublication Copy--Subject to Further Editorial Correction
Nuclear Research (CERN) in Geneva.47 Many of the concepts underlying the Web existed in Gopher, but
the Web incorporated more advanced interface features, was designed explicitly for linking information
across sites, and employed a common language, called Hypertext Markup Language (HTML), to describe
Web content. With the deployment of the Mosaic browser48 and its widely adopted commercial
successors--Netscape's Navigator and Microsoft's Internet Explorer--it became easy to view HTML
formatted documents and images located on the Web, making it much more appealing to most users than
Gopher's text-oriented system.
Indeed, the use of browsers on the Web eventually replaced the use of WAIS, FTP/Archie and
Gopher/Veronica. At the same time, the Web unintentionally made domain names valuable as identifiers,
thereby raising their profile and importance. While FTP and Gopher sites used domain names, these
names attracted little attention per se. However, the development of the Web and its meteoric growth
linked domain names and the Web very closely through their prominent inclusion in browser addresses as
a key part of Uniform Resource Locators (URLs). (See Box 6.2.) As noted in Chapter 2, it was
commonplace to navigate across the WWW just by typing domain names into the address bar, since
browsers automatically expanded them into URLs. Often, at that time, the domain names were simply
guessed on the assumption that a brand name followed by .com had a high probability of success.
The Web, whose underlying structure is hypertext--resources connected by links--also
introduced a new and convenient means of navigation: "clicking" on hyperlinks on a page displayed in a
browser. Hyperlinks generally have text, called anchor text, which many browsers display by underlining
and color change. When the user moves the pointer to the anchor text of a hyperlink and clicks the mouse
button, the browser finds the associated URL in the code for that page and accesses the corresponding
page.49 Thus, navigation across the Web from an initial site can consist of nothing more than a series of
clicks on the anchor text of hyperlinks.
The World Wide Web has experienced rapid and continual growth50 since the introduction of
browsers. In 1993, when the National Center for Supercomputing Applications' (NCSA) Mosaic browser
was made publicly available, there were just 200 sites on the Web. The growth rate of Web traffic in 1993
in transmitted bytes was 340,000%, as compared to 997 percent for Gopher traffic.51 By March 1994,
Web traffic exceeded Gopher traffic in absolute terms.52 Based on one estimate of the number of Internet
hosts,53 the exponential growth of Web sites continued from 1993 to 1998 with the number of hosts

47 Tim Berners-Lee and his team developed the Web in 1990. It was first released to the physics community in 1991
and spread quickly to universities and research organizations thereafter. See
48 Mosaic, which was released in 1993 by the National Center for Supercomputing Applications (NCSA) at the
University of Illinois, Urbana-Champaign, was the first Web browser with a graphical user interface for the PC and
Apple environments. It had an immediate effect on the widespread adoption of the Web by non-research users.
Marc Andreessen, one of the developers of Mosaic, became a co-founder of Netscape. See
49 Since there is no required association between the hyperlink and its anchor text, there is an opportunity for
malicious or criminal misdirection to deliberately bogus sites, which has been exploited recently as a vehicle for
identity theft and is referred to as "phishing" or "brand spoofing."
50 There is difficulty in measuring the rate of growth of and the volume of information on the Web
because of the lack of consistent and comprehensive data sources. The lack of consistent measurements
over time is also attributable to the very factors that continue to enable the Web's growth--as the
technology and architecture supporting the Web have evolved, so have the methods of characterizing its
51 See Robert H. Zakon. 1997. "Hobbes' Internet Timeline," Request For Comments (RFC) 2235, November.
Available online at <>.
52 See Calcari, Susan. 1994. "A Snapshot of the Internet," Internet World, September, Vol. 5, No. 1: pp. 54-58;
Merit Network, Inc. 1995. Internet statistics. Statistics available at <>; Graphics
and tables of NSFNET backbone statistics are available at <>.
53 Internet Systems Consortium provides host data, which are available at <>.

Prepublication Copy--Subject to Further Editorial Correction
approximately doubling each year. After 1998, the rate of growth in hosts slowed to a doubling every two
years. Web server data54 provides the number of active Web sites located via the DNS as a measure of the
growth in the volume of information on the Web. Although data are discontinuous, from June 1993 to
December 1995, the number of Web sites doubled every three months; from December 1997 to December
2000, the number doubled approximately every 10 months.

BOX 6.2
Uniform Resource Identifiers (URIs)55

The Uniform Resource Identifier (URI) is a general name for resource identifiers on the Internet.
Several distinct types of URIs have been defined: Uniform Resource Locators (URLs), Uniform Resource
Names (URNs), and Uniform Resource Characteristics (URCs), each of which identifies a resource
differently. URLs are, by far, the most commonly used of the three.
URLs are URIs that identify resources by their location. Although URLs can be used to locate many
different types of resources on the Internet, the most common URLs are those used to identify locations
on the World Wide Web.56 Web URLs, such as <>, include "http" to designate the
Hyper Text Transport Protocol (HTTP) and "" to designate the Web location of interest. The
example below shows the URL for the committee's charter document in HTML format located in the
directory57 "/cstb/committees/indns", located on the HTTP server.

h t t p : / / w w w . n a s . e d u / c s t b / c o m m i t t e e s / i n d n s / c h a r t e r . h t m l
Directory or other
Document or
other resource

A URL cannot separate the location from the name of a resource. Therefore, a URL is outdated when a
resource is moved to a new location, yet it remains unchanged even when the resource at that location is
The URN type of URI was defined to meet the need for a more persistent identifier. A URN
identifies a resource by assigning it a permanent and globally unique name drawn from a specified
namespace that is not tied to a location.58 However, there needs to be a continually updated table linking
each URN to the current location of the resource it names.
A third class of URI, the URC, was proposed to incorporate metadata about resources and their
corresponding URNs. URCs have not achieved widespread acceptance or use.

54 Matthew Gray of the Massachusetts Institute of Technology compiled data on the growth of Web sites from June
1993 to June 1995, which is available at <>; the
Netcraft Web Server Survey began in August 1995 and is available at <>.
55For a formal definition, see Request For Comments (RFC) 2396, August 1998, Uniform Resource Identifiers:
Generic Syntax.
For further information, see RFC 3305, Joint W3C/IETF URI Planning Interest Group: Uniform
Resource Identifiers (URIs), URLs, and Uniform Resource Names (URNs): Clarifications and Recommendations.
Available at <>.
56 The Internet Assigned Numbers Authority maintains a register of URI scheme names at
57 More generally, this string is passed to a server and then interpreted. Frequently, this string is interpreted as a
(partial) directory path and file name.
58 IANA maintains a register of namespaces at <>.

Prepublication Copy--Subject to Further Editorial Correction
A different approach than URIs for naming resources on the Internet motivates the Handle System.59
A handle consists of two parts separated by a forward slash. The first part is a naming authority--a
unique, arbitrary number assigned within the Handle System, which identifies the administrative unit that
is the creator of the handle, but not necessarily the continuing administrator of the associated handles.
The part after the slash is a local name that identifies the specific object. It must be unique to the given
naming authority. The Handle System allows a handle to map to more than one version or attribute of a
resource and resolve more than one piece of data. The multiple resolution capabilities of handles support
extended services such as reverse lookup, multi-versioning, and digital rights management.

As the volume of material on the Web grew exponentially, simple guessing and hypertext linking
no longer served to find the sites users wanted. Nor was the almost daily un-indexed list of new sites that
NCSA published (as "NCSA What's New") from June 1993 to June 199660 sufficient. New methods for
navigating to Web resources were badly needed and the need unleashed a flurry of innovation, much of it
based in universities and often the product of graduate students and faculty, who recognized the
opportunity and had the freedom to pursue it. The new navigation services took two primary forms:
directories and search engines. Directories organized Web resources by popular categories. Search
engines indexed Web resources by the words found within them.
The sequence of key developments from 1993 through 2004 in both forms of Web navigation
system is shown in detail in Box 6.3. A broad overview of the development follows:
Web navigation service development began in 1993, the year the Mosaic browser became
available, first to universities and research laboratories, and then more generally on the Internet. The first
directory and the first search engines were created in that year. But it was 1994 when the first of the
widely used directories--Yahoo!--and the first full-text search engine--WebCrawler--were launched.
Over the next few years, technological innovation occurred at a rapid pace, with search engines adding
new features and increasing their speed of operation and their coverage of the Web as computing and
communication technology and system design advanced. Lycos, launched in 1994, was the first Web
search engine to achieve widespread adoption as it indexed millions of Web pages. It was followed in
1995 by Excite and Alta Vista. Alta Vista in particular offered speed improvements and innovative
search features. With the launch of Google in beta in 1998 and as a full commercial offering in 1999, the
general nature of the technology of search engines appeared to reach a plateau, although there is continual
innovation in search algorithms and approaches to facilitate ease of use. The commercial evolution of
search services continued rapidly both through additional entries into the market and an increasingly rapid
pace of consolidation of existing entries. The evolution of directory technology has been less visible and
probably less rapid. The focus of that evolution appears, rather, to have been on the means of creating and
maintaining directories and on the addition of offerings including search engines, to the basic directory
By the first years of this century, the two worlds of search engines and directories had merged, at
least commercially. In 2004, Google offered Google directory (supplied by Open Directory) and Yahoo!
offered Yahoo search (provided by their acquisition--Inktomi) with paid ads (provided by their
acquisition--Overture). By 2004 most commercial navigation services offered advertisements associated
with specific responses. These paid ads are the principal source of funding and profit for commercial
navigation services. (The business of Internet navigation is discussed in further detail in Section 7.2.)
Associated with this latter development has been the rapid rise of the business of search engine

59 The Handle System was developed in 1994 by Robert E. Kahn, founder and chief executive officer of the
Corporation for National Research Initiatives (CNRI). For the foundational paper describing the components of this
system, see Robert E. Kahn and Robert Wilensky. "A Framework for Distributed Digital Object Services," May 13,
1995, Corporation for National Research Initiatives, Reston, Virginia, available at <
w.html>. Current status and available software can be found at <>.
60 An archive of those listings is available at <

Prepublication Copy--Subject to Further Editorial Correction
BOX 6.3
Key Events in the Development of Navigation Aids and Services
for the World Wide Web


Work started on the InQuery engine at the University of Massachusetts that eventually led to the Infoseek


First Web robot or spider.
61 World Wide Web Wanderer created by MIT student Matthew Gray. It was
used to count Web servers and create a database--Wandex--of their URLs.
Work on Architext search engine using statistical clustering started by six Stanford University
undergraduates, which was the basis for Excite search engine launched in 1995.
First directory. WWW Virtual Library created by Tim Berners-Lee.
ALIWEB created by Martijn Kosters at Nexor Co., U.K. An Archie-like index of the Web based on
automatic gathering of information provided by webmasters.
First robot-based search engines launched: World Wide Web Worm, JumpStation, and Repository-
Based Software Engineering (RBSE). None indexed full text of Web pages.


World Wide Web Worm indexed 110,000 Web pages and Web accessible documents; received an
average of 1500 queries a day (in March and April).
First searchable directory of the Web. Galaxy, created at the MCC Research Consortium, provided a
directory service to support electronic commerce.
First widely used Web directory. Yahoo! created by two Stanford graduate students, David Filo and
Jerry Yang, as a directory of their favorite Web sites. Usage grew with growth in entries and addition
of categories. Became public company in 1995.
First robot-based search engine to index full text of Web pages. Web Crawler created by Brian
Pinkerton, student at University of Washington.
Lycos created by Michael Mauldin, research scientist at Carnegie Mellon University. Quickly became
largest robot-based search engine. By January 1995 had indexed 1.5 million documents and by
November 1996, over 60 million--more than any other search engine at the time.
Harvest created by the Internet Research Task Force Research Group on Resource Discovery at the
University of Colorado. Featured a scalable, highly customizable architecture and tools for gathering,
indexing, caching, replicating, and accessing Internet information.
Infoseek Guide launched by Infoseek Corporation. Web directory, initially fee based, then free.
OpenText 4 search engine launched by Open Text Corporation based on work on full-text indexing and
string search for the Oxford English Dictionary. (In 1996 launched "Preferred Listings" enabling sites
to pay for listing in top ten results. Resultant controversy may have hastened its demise in 1997.)

Infoseek search engine launched in February. In December made deal with Netscape to be its default
search service, displacing Yahoo!
First meta search service. SearchSavvy, created by Daniel Dreilinger, a graduate student at Colorado
State University, queried multiple search engines and combined their results.

61 A spider is a program that collects Web pages by traversing the Web, following links from site to site in a
systematic way. See Box 7.2.

Prepublication Copy--Subject to Further Editorial Correction
First commercial meta search service. MetaCrawler developed by graduate student Erik Selberg and
faculty member Oren Etzioni at the University of Washington licensed to go2net.
Excite commercial search engine launched. Based on Stanford Architext engine.
Magellan directory launched by the McKinley Group. Complemented by a book, The McKinley Internet
Yellow Pages, that categorized, indexed, and described 15,000 Internet resources and which accepted
Search engine achieved record speed: 3 million pages indexed per day. AltaVista, launched by Digital
Equipment Corporation (DEC) combined computing power with innovative search features--
including Boolean operators, newsgroup search, user addition and removal of their own URLs--to
become most popular search engine.


First search engine to employ parallel computing for indexing: 10 million pages indexed per day.

Inktomi Corporation launched HotBot search engine based on work of faculty member Eric Brewer
and graduate student Paul Gauthier of the University of California at Berkeley. Used clusters of
inexpensive workstations to achieve supercomputer speeds. Adopted OEM search model: providing
search services through others. Licensed to Wired magazine's Web site, HotWired
First paid listings in an online directory. LookSmart launched as a directory of Web site listings.
Contained both paid commercial listings and non-commercial listings submitted by volunteer editors.
Also adopted OEM model.
First Internet archive. launched by Brewster Kahle as an Internet library with goal of
archiving snapshots of the Web's content on a regular basis.
Consolidation begins: Excite acquired WebCrawler and Magellan.


First search engine to incorporate automatic classification and creation of taxonomies of responses

and to use multidimensional relevance ranking. Northern Light search engine launched. Also
indexed proprietary document collections.
AOL launched AOL NetFind, its own branded-version of Excite. The Mining Company directory service
started by Scott Kurnitt and others. Used a netwok of "guides" to prepare directory articles.
First "question-answer" style search engine. Ask Jeeves launched. Company was founded in 1996 by a
software engineer, David Warthen, and venture capitalist, Garrett Gruener. Service emphasized ease
of use, relevance, precision, and ability to learn. launched by Brewster Kahle. Assisted search users by providing additional information: site
ownership, related links, link to Encyclopedia Britannica. Provided copy of all indexed pages to
Alta Vista, largest search engine, indexed 100 million pages total and received 20 million queries per day.
Open Text ceased operation.


First search engine with paid placement ("pay-per-click") in responses.
Idealab! launched GoTo
search engine. Web sites were listed in an order determined by what they paid to be included in
responses to a query term.
First open source Web directory. Open Directory Project launched (initially with name GNUhoo and
then NewHoo) with goal of becoming Web's most comprehensive directory through use of open
source model: contributions by thousands of volunteer editors.
First search engine to use "page rank," based on number of links to a page, in prioritizing results of
Web keyword searches. Stanford graduate students Larry Page and Sergey Brin announced Google,

Prepublication Copy--Subject to Further Editorial Correction
which was designed to support research on Web search technology. Available as "beta version" on
the Web.
Microsoft launched MSN Search using Inktomi search engine.
Direct Hit search engine introduced. Ranked responses by the popularity of sites among previous
searchers using similar keywords.
Yahoo! Web search powered by Inktomi.
Consolidation heats up: GoTo acquired WWW Worm; Lycos acquired Wired/HotBot; Netscape
acquired the Open Directory: Disney acquired large stake in Infoseek.


Google, Inc. (formed in 1998) opened fully operational search service. AOL/Netscape adopted it for
search on its portal sites.
Norwegian company Fast Search & Transfer (FAST) launched AllTheWeb search engine.
The Mining Company was renamed
Northern Light became the first engine to index 200 million pages.
FindWhat pay-for-placement search engine launched to provide paid listings to other search engines.
Consolidation continued: CMGI acquired AltaVista, At Home acquired Excite.


Yahoo! adopted Google as the default search results provider on its portal site.
Google launched advertising program to complement search services; added Netscape Open Directory to
augment its search results; and began 10 non-English language search services.
Consolidation continued: Ask Jeeves acquired Direct Hit; Terra Networks S.A. acquired Lycos.
By end of year, Google had become largest search engine on Web with index of over 1.3 billion pages,
answering 60 million searches per day.


Google acquired Usenet archive dating back to 1995.
Overture, new name for GoTo, became leading pay-for-placement search engine.
Teoma search engine launched in April, bought by Ask Jeeves in September.
Wisenut search engine launched.
Magellan ceased operation.
At end of year, Google indexed over 3 billion Web documents (including Usenet archive dating back to


Consolidation continued:
LookSmart acquired Wisenut.
Gigablast search engine was launched.


Consolidation heated up:

Yahoo acquired Inktomi and Overture, which had acquired AltaVista and AllTheWeb.

FindWhat acquired Espotting.

Google acquired Applied Semantics and Sprinks.
Google indexed over 3 billion Web documents and answered over 200 million searches daily.


Prepublication Copy--Subject to Further Editorial Correction

Competition became more intense:
Yahoo! switched its search from Google to its own Inktomi and Overture services. entered the market with, which added Search Inside the BookTM and other
user features to the results of a Google search.
Google included (in February) 6 billion items: 4.28 billion Web pages, 880 million images, 845 million
Usenet messages, and a test collection of book-related information pages.
Google went public in an initial public offering in August. Ask Jeeves acquired Interactive Search
Holdings, Inc., which owns Excite and iWon.
In November, Google reported that its index included over 8 billion Web pages.62
In December, Google, four university libraries, and the New York Public Library announced an
agreement to scan books from the library collections and make them available for online search.63

SOURCES: Search Engine Optimization Consultants. 2003. "History of Search Engines and
Directories." July. Available at <>; iProspect.
2003. "A Brief History of Search Engine Marketing and Search Engines." Available at
<>; Danny Sullivan. 2001. "Search
Engine Timeline." Available at
<>; and Wes Sonnenreich.
1997. "A History of Search Engines." Available at

marketing, which helps commercial Web sites to decide in which search engines and directories they
should pay for ads; and search engine optimization, which helps to design Web sites so search engines
will easily find and index them. (Organizations that provide these services, generally also provide
assistance with bidding strategies and assistance in advertising design.) Finally, by 2004, many search
services had established themselves as portals--sites offering a variety of often customizable sources of
information--news, weather, stock prices, entertainment listings--and links to services--shopping,
auctions, travel sites, mapping--in addition to their basic search services.
The development of Internet navigation aids and services, especially those focused primarily on
the Web, stands in interesting contrast to the development of the Domain Name System as described in
Chapter 2.

Conclusion: A wide range of reasonably effective, useable, and readily available Internet
navigation aids and services have been developed and have evolved rapidly in the years since the World
Wide Web came into widespread use in 1993.

Large investments in research and development are presently being made in commercial search
and directory services. Still, many of the unexpected innovations in Internet navigation occurred in
academic institutions. These are places with strong traditions of information sharing and open inquiry.
Research and education are "problem-rich" arenas in which students and faculty nurture innovation.

Conclusion: Computer science and information science graduate students and faculty played a
prominent role in the initial development of a great many innovative Internet navigation aids and services,

62 See "Search Swagger," Vise, David A., Washington Post, November 11, 2004, p. E1.
63 John Markoff and Edward Wyatt. 2004. "Google Is Adding Major Libraries to Its Database." NY Times.
December 14. Available at <>.

Prepublication Copy--Subject to Further Editorial Correction
most of which are now run as commercial enterprises. Two of those services have become the industry
leaders and have achieved great commercial success--Yahoo! and Google.

Conclusion: Because of the vast scale, broad scope, and ready accessibility of resources on the
Internet, the development of navigation aids and services opens access to a much wider array of resources
than has heretofore been available to non-specialist searchers. At the same time, the development of
successful Internet navigation aids and services opens access to a much broader potential audience than
has heretofore been available to most resource providers.

One cannot know if the past is prelude, but it is clear that the number and the variety of resources
available on the Internet continue to grow, that uses for it continue to evolve, and that many challenges of
Internet navigation remain. Some of the likely directions of technological development are described in
Section 8.1


Searching on the public Web has no direct analog with searching libraries, which is both an
advantage and a disadvantage. Library models provide a familiar and useful comparison for explaining
the options and difficulties of categorizing Web resources.
First, locating an item by its URL has no direct equivalent in library models. The URL usually
combines the name of a resource with its precise machine location. The URL approach assumes a unique
resource at a unique location. Library models assume that documents exist in multiple locations, and
separate the name of the document (its bibliographic description, usually a catalog record or index entry)
from its physical location within a given library. Classification systems (e.g., Dewey or Library of
Congress) are used for shelf location only in open-stack libraries, which are prevalent in the U.S. Even
then, further coding is needed to achieve unique numbering within individual libraries.64 In closed-stack
libraries where users cannot browse the shelves, such as the Library of Congress, books usually are stored
by size and date of acquisition. The uniqueness of name and location of resources that is assumed in a
URL leads to multiple problems of description and persistence, as explained in this chapter.
Second, resources on the Web may be located by terms in their pages because search engines
attempt to index documents in all the sites they select for indexing, regardless of type of content (e.g.,
text, images, sound; personal, popular, scholarly, technical), type of hosting organization (e.g.,
commercial, personal, community, academic, political), country, or language. By comparison, no single
index of the contents of the world's libraries exists. Describing such a vast array of content in a consistent
manner is an impossible task, and libraries do not attempt to do so. The resource that comes closest to
being a common index is WorldCat,65 which "is a worldwide union catalog created and maintained
collectively by more than 9,000 member institutions" of OCLC and currently contains about 54 million
items.66 The contents of WorldCat consist of bibliographic descriptions (cataloging records) of books,
journals, movies, and other types of documents; the full content of these documents is not indexed.
Despite the scope of this database, it represents only a fraction of the world's libraries (albeit most of the
largest and most prestigious ones), and only a fraction of the collections within these libraries (individual
articles in journals are not indexed, nor are most maps, archival materials, and other types of documents).
The total number of documents in WorldCat (54 million) is small compared to those indexed by Google,
InfoSeek, AltaVista, or other Internet search engines.

64 The shelf location ZA3225.B67 2000 for (Borgman, 2000) at the University of California at Berkeley library
consists of the Library of Congress Call Number (ZA3225) plus a local number to order the author name (B67 for
Borgman) and the date (2000) to create a unique shelf placement in this library.
65 See <>.
66 Visited May 7, 2004.

Prepublication Copy--Subject to Further Editorial Correction
Rather than create a common index to all the world's libraries, consistent and effective access to
documents is achieved by dividing them into manageable collections according to their subject content or
audience. Library catalogs generally only represent the collections of one library, or at most a group of
libraries participating in a consortium (e.g., the campuses of the University of California67). These
catalogs describe books, journals (but not individual journal articles), and other types of documents.
Many materials are described only as collections. For example, just one catalog record describes all the
maps of Los Angeles made by the U.S. Geological Survey from 1924 to 1963. It has been estimated that
individual records in the library catalog represent only about 2% of the separate items in a typical
academic library collection.68 Thus, library catalogs are far less comprehensive than most library users
realize. However, online library catalogs are moving away from the narrower model of card catalogs.
Many online catalogs are merging their catalog files with records from journal article databases. Mixing
resources from different sources creates a more comprehensive database, but introduces the Web
searching problem of inconsistent description.
Nor do libraries attempt the international, multi-lingual indexing that search engines do. The
catalogs of libraries may be organized consistently on a country-by-country basis, at best. The descriptive
aspects of cataloging (e.g., author, title, date, publisher) are fairly consistent internationally, as most
countries use some variation of the Machine Readable Cataloging (MARC) metadata structure. (Metadata
are data about data or, more generally, about resources.) However, many variations of the MARC format
exist, each tied to a national or multi-national set of cataloging rules. The U.S. and U.K. share the Anglo-
American Cataloging Rules, but store their data in USMARC and UKMARC formats, respectively. These
formats are finally being merged, after nearly 40 years of use. In 2004, the national libraries of the U.S.
and the U.K. implemented a common format, MARC 21, and other countries are following suit.69 Other
MARC formats include UNIMARC, HUMARC (Hungary), and FINNMARC (Finland). The OCLC
WorldCat database merges these into an OCLC MARC format. Each catalog describes its holdings in its
local language and may also include descriptions in the language of the document content. For example,
OCLC WorldCat contains records describing resources in about 400 languages; each record has some
descriptive entries in English. Thus, libraries achieve interoperability through highly decentralized
cataloging activities. The cataloging enterprise is economically feasible in the United States because most
published resources are described by the publishers and the Library of Congress and contributed to OCLC
WorldCat and other shared databases. Despite the relative national and international success in
establishing cataloging rules and formats among libraries, incompatibilities continue to exist within these
communities, and the archives and museum communities employ yet other metadata formats. Given the
difficulty of achieving agreement on basic descriptive models among these established institutions run by
information management professionals, the likelihood of getting universal agreement on descriptive
standards for Web documents is low. Decentralized models for data creation, combined with mapping
between similar formats, are the most feasible way to achieve interoperability.
Thirdly, while a rich subject index to the Web would certainly be extremely useful, universal
subject access is almost impossible to achieve. American libraries attempt general subject access via the
Library of Congress Subject Headings (LCSH), but these only apply two or three headings per book. Few
other countries appear to use the LCSH, as many of the concepts are specific to American culture. Unified
access via classification systems such as the Dewey Decimal Classification (DDC) and Library of
Congress Classification (LCC) are more common. These are also country- and culture-specific. DDC and

67 See <>.
68 See "The 98% Solution: The Failure of the Catalog and the Role of Electronic Databases," Tyckoson, David A.,
1989, Technicalities 9(2): 8-12.
69 See The British Library, MARC 21 and UKMARC. 2004. London: British Library. November 24,
available online at <>. MARC 21 Concise
Format for Bibliographic Data (2003). Concise Edition. Washington, D.C.: Library of Congress.
November 24, available online at <>.

Prepublication Copy--Subject to Further Editorial Correction
LCC are little used outside the U.S.; the Universal Decimal Classification (UDC) has broader
international adoption. Many other country-specific classifications exist.
Consistent subject access in any depth is feasible only within topical areas due to a number of
well-understood linguistic problems.70 These include synonymy (multiple terms or phrases may have the
same meaning), polysemy (the same terms and phrases may have multiple meanings), morphological
relationships (structure of words, such as variant endings (e.g., acid and acidic; dog and dogs; mouse and
mice), and semantic relationships (conceptual relationships (e.g., two words may have the same meaning
in one context and different meanings in other contexts). Both automatic and manual methods to provide
consistent retrieval by controlling the meaning of words work best when the subject area is constrained.
Controlled sets of terms (e.g., thesauri, ontologies) can be constrained to their meaning within one field,
such as computer science, economics, arts, or psychology. Libraries construct or purchase indexes
specific to each field within the scope of their collections.
Fourth, Web directories are more analogous to topic-specific library indexes than to library
catalogs. However, Web directories cover only a small portion of the content of the Web and their
descriptions of each item are often less complete and can be less reliable than those created by
professional librarians. Consequently, structured and formal characterizations of material can be
accomplished most effectively in a library catalog or bookstore database, rather than in a general Web
Selecting the proper resource to search remains an important starting point in seeking information
whether online or offline.

70 See Richard K. Belew. 2000. Finding Out About: A Cognitive Perspective on Search Engine Technology and the
Cambridge, U.K.: Cambridge University Press; Peter Brusilovsky and Carlo Tasso. 2004. User Modeling for
Web Information Retrieval: Introduction to Special Issue. User Modeling and User-Adapted Interaction: The
Journal of Personalization Research
14(2-3): in press; and William A. Woods. 2004. Searching vs. Finding. ACM
2(2) available at <>.

Prepublication Copy--Subject to Further Editorial Correction
Internet Navigation: Current State


At this point in the development of Internet navigation, there are at least seven basic ways for a
user to navigate to a desired Web resource, which is generally located on a page within a Web site. Five
of them are direct--users' actions take them immediately to the desired resource. Two are indirect--users
must first employ a navigation service, either a directory or a search engine, to find the address of a
desired resource and then, using that information, go to the desired resource. These basic ways can be and
often are used in combination with one another. Table 7.1 summarizes and characterizes the various
Internet navigation aids and services.
This discussion is concerned with navigation across the Internet and not specifically with
navigation within sites, although the tools deployed in both cases are usually similar. Most Web sites--
except those with only a few pages--now incorporate one or more means of navigation within the site
itself. These include hyperlinks, directories (menus), site maps, and search engines. Because they are
usually limited to the contents of the site, the problems of general-purpose Web navigation aids are
diminished. For example, the context is delimited, the users are relatively homogeneous, the scale is
relatively small, and material that is difficult to automatically index (such as multimedia and images) can
usually be manually indexed.

TABLE 7.1 Principal Internet Navigation Aids and Services
Indexing Process
File Structure
1. Domain Name--
1 or 2
known or guessed
2. Hyperlink
3. Bookmark
flat or hierarchical
flat or hierarchical
5. Metadata
flat or hierarchical
6. Directory
hierarchical or
7. Search engine

7.1.1 Direct Access via a Uniform Resource Locator or Domain Name

1 "KEYWORD" is capitalized to distinguish it from the use of keywords in traditional information retrieval or in
Internet search engines (see Sections 7.1.4 and 7.1.7). In this use, each KEYWORD is part of a controlled
vocabulary for which the match to a specific Internet resource is one to one.

Prepublication Copy--Subject to Further Editorial Correction
One of the major factors in the success of the Web was the development of Uniform Resource
Locators (URLs) for Web sites. Because those identifiers offered a standardized way to identify
resources on the Web, resource providers and resource seekers had a common way to refer to their
locations. But the URLs were intended as codes hidden behind meaningful objects--the anchor text--in
a document, not directly typed by the user. The designers of the Web may have been surprised when
URLs began appearing on the sides of buses and on billboards.
URLs are typically not managed to be permanent and can be difficult to remember, especially
when they require many elements to describe a resource deep within a Web site. (Examples of such URLs
abound in the footnoted references throughout this report.) Despite their flaws, however, they have
thrived as a robust means of navigation.
In some browsers, users can also navigate through the Web by typing only a domain name
because the browsers will automatically expand it into a URL that may identify a Web site. This use of
domain names for navigation is effective to the degree that the searcher knows or is able to guess the
domain name exactly and is satisfied with being taken to the home page of a Web site. If the name
entered is incorrect, then the browser, e-mail server, or other Internet service consumes resources (in the
local computer and in the Internet) trying to find a DNS match. Mistaken guesses can create extra traffic
and burden the DNS, as discussed in Chapter 3. However, most browsers now treat invalid domain names
as search terms and return a list of possible matches.2
Furthermore, as Web sites have grown more complex, discovering the site has often had to be
followed by a second navigation process to find relevant information or pages within the site. But some
users would prefer to go directly to the page that contains the specific information being sought.
Remembering or guessing does not suffice for such navigation because the URLs of inner pages comprise
more than the domain name.
In addition, as network services proliferate and as additional top-level domains are added, users
will have many more sites of interest to which to navigate, but at the probable cost of domain names that
are more difficult to remember or guess. Furthermore, not only information, entertainment, and service
resources, but also many personal electronic devices and home appliances may well be connected to the
Internet. For convenience, users will probably want to assign easy-to-remember domain names to such
devices. But because of competition for the easiest and shortest names, they may have to settle for less-
readily remembered ones. In either event, they can use bookmarks (see Section 7.1.3) to simplify access.
For these reasons, remembering or guessing correct domain names is likely to become less
dependable and, therefore, a less important aid to navigation as the number of locations on the Internet
continues to expand.

7.1.2 Direct Access via Hyperlinks

Because the Web is a network of sites through which users can navigate by following links
between documents on the sites, once the first site has been found, one can move across sub-networks of
related and relevant information and services. The address of the linked-to information may be visible or,
more typically, hidden behind anchor text. A human being defines the linkages within and from a site
during site design.
There is no publicly-available Internet-wide file of links; they are maintained locally. However,
linkage information is collected and used by all major search engines as an important part of the ranking
of responses. For example, Google maintains an extensive file of linkages and it is possible to use Google
to find all the pages that link to a given page (within the scope of what is indexed by Google; see "The
Deep, Dark, or Invisible Web" in Section 7.1.7).

2 As discussed in Chapter 4, VeriSign tried to offer a service to users who enter an incorrect .com or .net domain
name that directed them to possible matches, raising technical and policy issues.

Prepublication Copy--Subject to Further Editorial Correction
Navigation by following hyperlinks is an effective tool for moving between related sites once the
first relevant site has been found. However, since Web site operators establish the linkages, they may or
may not lead to the specific sites of interest to the user. Thus, navigation by hyperlinks is both a valuable
and a limited aid. It generally must be supplemented by other means of finding starting sites and of
identifying sites of interest that may not be on the radiating set of paths from the initial point.

7.1.3 Direct Access via Bookmarks

The URLs for sites of continuing interest that have been found after a search and those of
frequently accessed sites can be stored locally--"bookmarked" or placed on a "favorites" list--and
managed in most browsers. By doing so, the user can return directly to a location, perhaps deep within a
site, with a single click. However, these local files can become difficult to manage over time, due both to
scaling problems (as the list of bookmarks grows, it may require its own database) and to the likelihood of
broken links or changed content as URLs age. For these reasons, bookmarks may become less useful with
the scaling and maturing of the Internet, leading users to rely on search engines to find even familiar sites
and Web pages.
The bookmark/favorite mechanism as implemented in current browsers and described above is
fairly weak, providing a simple association between a name (provided by either the user or the Web page)
and a URL. Richer methods are possible. For example, prior experience in both information retrieval and
software engineering suggests that it would be useful to store, in some form, both the query that produced
the reference and information about how long the reference was likely to remain current. With this
information available, it would become easier to repeat the discovery process when a link went bad,
perhaps even automatically. Some work is now underway to recast bookmarks as a type of local cache
with this information included and some reference updating and recovery capabilities. That work also
expects to unify the results of multiple types of navigation, from search engine output, to Uniform
Resource Identifiers (URIs) obtained from colleagues, to links obtained from pages being traversed, into a
single framework. (In information retrieval and library practice since the early 1960s, queries have been
stored and then periodically executed to support Current Awareness or Selective Dissemination of
Information services.3 However, unlike the bookmark case, the queries are run by a service on a regular
schedule and not by users only when they need to update their bookmarks.)

7.1.4 Direct Access via KEYWORDS

The term "keyword" is used in several contexts, with slightly different meanings, in Internet
Its most common current use is to denote the terms entered into the search window of a search
engine for matching against the search engine's index of words appearing on Web pages.4 In this
meaning, a "keyword" can be any phrase and can be treated in a variety of ways by individual search
mechanisms. It is also used in this sense in search engine marketing to refer to the search terms for which
a particular marketer is willing to pay.5

3 See Korfhage, Robert R. 1997. Information Storage and Retrieval. New York: Wiley; and Hensley, C. B.,
Savage, R. R., Sowarby, A. J., and Resnick, A. 1962. "Selective Dissemination of Information ­ A New Approach
to Effective Communication," IRE Transactions of the Professional Group on Engineering Management, EM-9:2.
4 This is similar to the sense in which "keyword" has conventionally been used in information retrieval, where a
"keyword" is "one of a set of individual words chosen to represent the content of a document." See Korfhage
(1997), p. 325.
5 The marketer's site or advertisement will appear as one of the responses to any query that includes those keywords.
In this context, there generally are several keywords entered in the query and many responses in the list produced by
the search engine. This use of "keyword" is treated in detail in Section 7.2.2.

Prepublication Copy--Subject to Further Editorial Correction
However, "keyword" has also been used to denote terms in a controlled vocabulary linked to
specific Internet locations by a specific Internet (generally, Web) service. To distinguish this meaning, it
is written here in capitals. Typically, KEYWORDS are associated with a particular organization or
service and that organization or service has paid to have them linked uniquely to its location. Usually, just
a single KEYWORD (or phrase) is entered and only one site appears in the response. They apply,
however, only within a specific Web service and are not generally interpretable in the same way by
others. One of the best known uses of KEYWORDS is that of America Online (AOL) in which
KEYWORDS can be typed into the AOL address bar.6 AOL KEYWORDS link uniquely to a network
resource--"NYTIMES" links to, or to an AOL feature or service--"STOCK
MARKET" links to the AOL Market News Center. (The latest versions of AOL now offer a choice
between: "Go to AOL keyword: `NY Times'" or "Search the Web for `NY Times'".) Typing the AOL
KEYWORDS into MSN or into Internet Explorer will not necessarily lead to the same location. Indeed,
both "NYTIMES" and "STOCK MARKET" when typed into Internet Explorer and Netscape Navigator7
are treated as search terms (keywords in the more general sense) and the response is a ranked list of
possibly matching sites.
Several years ago, there were several attempts to offer more widely applicable KEYWORDS on
the public Internet. A service offered by RealNames, Inc. was available for several years. It was adopted,
for example, by MSN, which, however, terminated its use in June 2002.8 RealNames closed shortly
thereafter. KEYWORDS have been replaced in most cases, except for services catering to non-English
language users9 and AOL--by search engines, which provide a wider ranging response to keyword terms,
and by the sale of search engine keywords to multiple bidders.
KEYWORDS have many of the same strengths and weaknesses as domain names for navigation.
If known, they lead exactly to the location to which the purchaser of the KEYWORD wishes to lead the
searcher (which may not be the same as the searcher's intent). If guessed, they either succeed, lead to the
wrong site, or fail. However, since many browsers and services now treat non-URL entries in their
address lines as search terms, "failure" now generally produces a ranked list of possible matches. Thus,
KEYWORD systems--including AOL's--now default to search systems, just as domain name guesses
generally do.10
Unlike the DNS, a variety of KEYWORD systems applicable to specific topic areas and with or
without hierarchical structure are conceptually possible. Implementation of a KEYWORD system on the
Web requires an application or a service, such as a browser or Netpia, that recognizes the KEYWORD
terms when entered into its address line or when they reach the service's name server. And, whereas in
the early days of the Web, such an innovation might have been relatively easy, the general
implementation of standardized browser software in various versions makes the widespread introduction

6 See Danny Sullivan. 1999. "AOL Search Big Improvement for Members,", available
online at <>. See also Dominic Gates. 2000. "Web
Navigation for Sale," The Industry Standard, May 15. Available online at
7 Test carried out in March 2005.
8 See Danny Sullivan. 2002. "RealNames to Close After Losing Microsoft,", June 3.
Available online at <>. The committee heard
testimony from Keith Teare, then chief executive officer of RealNames, at its July 2001 meeting.
9 Two prominent native language KEYWORD systems are: (1) Netpia, a Korean Internet service that offers Native
Language Internet Address (NLIA) for 95 countries. NLIA enables substitution of a native language word or phrase
(a KEYWORD) for a unique URL. See <>. (2) Beijing 3721 Technology Co., Ltd., which has
offered Chinese language keywords since 1999. See <>.
10 In July 2004, Google added a "Browse by Name" feature to its search, enabling a user to enter a single name in
the tool bar and returning a single site if the term is specific or well-known; if not, it defaults to a traditional search.
It is not clear how the single response names are selected and, whether or not they are paid for. See Scarlett Pruitt.
2004. "Google Goes Browsing by Name," PC World. July 15. Available at:

Prepublication Copy--Subject to Further Editorial Correction
of a new feature much more difficult (although specific services, such as AOL or Netpia, can still
implement them for their users.)
Moreover, within any specific database, digital library, or community repository (such as the
large databases of primary scientific data being assembled around the world), terms can take on local
meanings. Generally speaking, meanings are constrained by the use of a controlled vocabulary, which
defines each term as applied in this system. Well-known examples of controlled vocabularies include the
Library of Congress Subject Headings, the Medical Subject Headings (MeSH), Subject Headings for
Engineering (SHE), and the Association for Computing Machinery (ACM) Classification System.
KEYWORD systems also face the problems that arise from scale. The larger the number of
locations to which they seek to assign simple and unique names, the greater the pressure to become
complex and structured. The system must either remain manageably small or develop an institutional
framework that allows decentralization, while centrally determining who can use which names to
designate which locations. AOL and Netpia both centrally determine the assignment of names. However,
Netpia implements KEYWORDS in 95 languages through decentralized name servers located at
collaborating ISPs, while AOL implements its smaller list of KEYWORDS through its own service

7.1.5 Direct Access via Metadata

Since the early days of the Web, there has been a desire--especially, but not only, by those in the
library and information science community--to establish a more consistent and more controlled way to
categorize and describe Web resources based on the use of "data about data" or metadata,.11
However, differences between the Web12 and conventional libraries and data collections
complicate fulfillment of that desire. First, the number, scope, and diversity of resources on the public
Web exceed that in any library. Second, the quality of the, often self-provided, metadata is highly
variable. And third, there is no organization or group of organizations that is able and willing to assume
responsibility for assigning metadata tags to a significant portion of the resources accessible on the Web,
as the Library of Congress does for books.
Efforts to adapt metadata for the description and categorization of sufficiently valuable Web
resources began in the mid-90s, when standard ways to express metadata--metadata schemes--were
proposed as the answer to interoperability and scaling of the expanding Web.13 But a re-examination of
the mid-90s' recommendations in 2002 forced their proponents to consider why metadata had not been
successfully used.14 The error was in their assumptions: they had expected to find high quality--clean and
honest--information, not the large amount of misrepresented and deliberately incorrect metadata that was
provided for resources on the Web.

11 For an overview, see Tony Gill. 2000. Introduction to Metadata: Metadata and the World Wide Web. July. Getty
Research Institute. Available online at
12 Hypertext Markup Language (HTML)--the programming language of Web site construction--specifies the
expression of metadata in the form of "meta" tags that are visible to search engines (as they collect data from the
Web--Box 7.2), but are not typically displayed to humans by browsers. To see meta tags, if they are present on a
Web page, go to View/Source in Internet Explorer or View/Page Source in Netscape Navigator.
13 See Lynch, Clifford A. and Garcia-Molina, Hector. 1995. Interoperability, Scaling, and the Digital Libraries
Research Agenda
. Available online at
<>. Visited July 9, 2004.
14 Christine L. Borgman. 2002. "Challenges in Building Digital Libraries for the 21st Century," in Ee-Peng Lim;
Schubert Foo; and Christopher S.G. Khoo (eds.). 2002. Digital Libraries: People, Knowledge & Technology:
Proceedings of the 5th International Conference on Asian Digital Libraries (ICADL 2002),
Singapore. December
12-14, 2002 and Lecture Notes in Computer Science. Heidelberg, Germany: Springer-Verlag, pp. 1-13,

Prepublication Copy--Subject to Further Editorial Correction
It seems that any feasible attempt to develop metadata schemes and apply them broadly to the
Web would have to be decentralized and based on the efforts of a large number of autonomous
organizations with specific knowledge of the content and quality of the resources they describe. Yet
decentralization raises the question of coordination among the many potentially inconsistent and non-
interoperable metadata schemes that the autonomous organizations might otherwise develop. Through
coordination, their separate efforts could cover a significant portion of the Web and open access to their
resources to a wider audience beyond the organizations themselves. Two approaches have been taken to
the coordination of metadata schemes produced by autonomous organizations.
The first approach to coordination is for organizations to collaborate in defining common
metadata elements that will be used by all of them as a core for their metadata schemes. The best known
and best developed of these is the Dublin Metadata Core Element Set, known as the Dublin Core,15 so
named because it originated at a meeting in Dublin, Ohio that was sponsored by the Online Computer
Library Center (OCLC). It comprises fifteen metadata elements, which were thought to be the minimum
number required to enable discovery of document-like resources on the Internet. Thus, Dublin Core can
be used to discover an item, to determine where fuller descriptions can be found, and to identify its
detailed format (e.g., MARC for books or the Environmental Markup Language for biocomplexity data).
Dublin Core works best when a professional cataloger creates descriptions. To achieve wide adoption,
some believe that it needs to be made more suitable to machine-generated descriptions.16
The second approach to coordination is to provide a higher-level structure that can incorporate
multiple metadata schemes, enabling them to be deployed in combination to describe a resource with the
assurance that the resultant description will be correctly interpreted by any computer program that is
compatible with the higher-level structure. The best known and best developed of these higher-level
structures is the Resource Description Framework (RDF) developed by the World Wide Web Consortium
(W3C).17 It extends another W3C standard, Extensible Markup Language (XML),18 which is used to
describe data where interchange and interoperability are important, to describe resources. Any Web
resource--that is, any object with a URI--can be described in a machine understandable way in the RDF,
although for some objects the description might contain little information. The resource--Web object--is
described by a collection of properties--its RDF description. These properties can come from any
metadata scheme since the RDF description incorporates reference information about the metadata
scheme and the definition for each property. The advantage of the RDF is that it provides a widely
applicable framework within which specialized metadata sets can be combined. For example, to describe
geographic resources on the Web, an RDF description might incorporate the Dublin Core to describe the
bibliographic provenance and a geographic metadata scheme to describe the geographic coverage of each
resource. The developers of the RDF believe that its existence will encourage the development of a large
number of metadata schemes for different resource domains and that where there is overlap in their
coverage, they will, in effect, compete for adoption by those who describe resources.

15 The Dublin Core Web site is at <>. Official reference definitions of the metadata elements
can be found there.
16 See Carl Lagoze. 2001. Keeping Dublin Core Simple: Cross-Domain Discovery or Resource Description. D-Lib
Magazine, 7(1), available at <>. Lagoze provides a useful
discussion of the tradeoffs in simple and complex metadata descriptions and the relationship between Dublin Core,
RDF, and other schema.
17 See Resource Description Framework (RDF)/W3C Semantic Web Activity available at <>
and RDF Primer Primer (correct as written) available online at <>.
18 See Extensible Markup Language (XML) available at <>.

Prepublication Copy--Subject to Further Editorial Correction

BOX 7.1
The Semantic Web

Despite the problems of characterizing most resources on the public Web with RDF metadata,
there are islands of application and, over the long term, they may extend to cover ever more terrain. With
that prospect in mind, Tim Berners-Lee and his colleagues at the W3C have proposed a way of linking
these islands into a formalized network of knowledge that they call the "Semantic Web."19 They do so by
introducing "ontologies" that consist of relational statements (propositions) and inference rules for
specific domains of knowledge and are used to define terms used on the Web. In their vision, the
Semantic Web would enable Web agents to draw upon that network of machine-accessible knowledge to
carry out complex functions with less explicit direction than is currently required. While its area of
application is far broader than navigation, its developers foresee, for example, that software agents will
"use this information to search, filter, and prepare information in new and exciting ways to assist the Web
user."20 Like metadata and RDF, the applicability and feasibility of the Semantic Web remains the subject
of dispute between its advocates and the skeptics.21
The practical implementation and use of the Semantic Web is highly dependent upon the broad
adoption of RDF and the ontologies it requires. That work has proceeded slowly thus far.

While the RDF may provide a useful framework within which various metadata schemes may be
developed and combined, it does not resolve the more difficult problem of actually using these metadata
schemes to describe resources on the Web. That problem has three components: determining who will
assign the metadata to each resource; finding incentives for metadata use; and determining how the
metadata will be used.
The resolution of that three-component problem is easiest within communities, whether organized
by topic, geographic region, or some other shared subject area.22 Individual communities in several
academic disciplines are creating their own repositories with their own metadata frameworks. Among the
repositories that have been established are IRIS for seismology, KNB for biocomplexity, and NCAR--
among others--for environmental data.23 Other communities have established portals to gather resources
and links to other resources on their topic of interest. Communities build these metadata-based
repositories and portals out of self-interest--with accurate metadata they can provide better access to
community resources. As both the creators and the users of the metadata, the self-interest of cohesive

19 See Tim Berners-Lee, James Hendler, Ora Lassila. 2001. "The Semantic Web," Scientific American. May.
Available at <
20 See James Hendler, Tim Berners-Lee, and Eric Miller. 2002. "Integrating Applications on the Semantic Web,"
Journal of the Institute of Electrical Engineers of Japan. October, vol. 122(10) 676-680. Available online at
<>. For an imaginative exploration of the possibilities, see: Paul Ford. 2002.
"August 2009: How Google beat Amazon and Ebay to the Semantic Web. July 26. Available online at
21 See, for example, Clay Shirky. 2003. "The Semantic Web, Syllogism, and Worldview." November 7. Available
online at <>. Also see Paul Ford. 2003. "A Response to
Clay Shirky's `The Semantic Web, Syllogism, and Worldview'," November 13. Available online at
22 See Chris Sherman. 2002. "Search Day ­ Metadata or Metagarbage," March 4,
Available online at <>.
23 IRIS (Incorporated Research Institutions for Seismology) is at <>; KNB (The Knowledge
Network for Biocomplexity) is at <>; and NCAR (National Center for
Atmospheric Research) is at <>.

Prepublication Copy--Subject to Further Editorial Correction
communities leads them to want trustworthy metadata and to provide the resources needed to create and
keep them current and accurate.24
Solving that three-component problem is more difficult for the general Web user "community."
Metadata would either have to be supplied by independent editors (as it is now for use in directory
services) or applied by the resource providers and collected automatically by search engines. Although
search engines look at the meta tags--a type of information about a web page that can be placed in the
code describing the page but not made visible to users on Web sites, it is not always clear whether and
how they make use of the metadata they find there. And the fundamental difficulty of unreliable self-
assigned metadata is difficult to overcome through automatic means.
However, one important current use of metadata is to characterize images and audio and video
files on the general Web so that they can be indexed and found by search engines. The metadata tags are
generally either extracted from text accompanying the images or supplied manually by editors or the
resource provider and appear, generally, to be reliable. (See Section 8.1.3.)
Thus, it is highly unlikely that general metadata schemes, even if they were designed, could be
reasonably implemented for the Web generally. However, metadata schemes may be practical and useful
for specific sets of resources with interested user communities, such as professional organizations,
museums, archives, libraries, businesses, and government agencies and for non-textual resources, such as
images, audio, and video files. Moreover, even in specialized resources, establishing the framework and
assigning the metadata terms to a large number of items are very different matters, since the latter is far
more labor intensive. Thus, the widespread use of metadata would become easier with the improvement
of automatic indexing (automatic metadata tagging), a topic that has long been pursued in information

7.1.6 Navigation via Directory Systems

In general, the term "directory" refers to a structured collection of objects organized by subject,
much like a library card catalog or yellow pages telephone listing. Structuring a directory--usually using
a taxonomy of some form--and placing objects under specific subject headings are done by humans.
They may assign the same subject heading to more than one object and the same object may be assigned
to more than one subject--that is, it may appear under more than one heading. In the case of Internet
directories, the objects are Internet locations--Web sites or Web pages, typically.
Only the Web sites that have been submitted to or found by an Internet directory service and
whose content has been classified and described, either by the editors of the directory or by the creators of
the Web site, will be available in the directory. As a result of the heavy requirements for skilled labor,
Internet directories can include only a small selection of all the sites connected by the Web. However, in
contrast to search engines, they have the advantage of being able to incorporate listings of many Web
sites in the "Dark Web" (see "The Deep, Dark, or Invisible Web" in Section 7.1.7) because the sites
themselves solicit a listing or because they become known to the directory editors through other means.
The listings, categorized under subject hierarchies, can be browsed by narrowing down subject
categories or can be searched by matching search terms against summary descriptions of the Web site
content.26 For example, the Yahoo! directory of the Internet classifies Web sites by 14 topics. Under the
"Computers and Internet" topic, there are two commercial (sponsored) categories and 48 additional
categories. The latter grouping includes topics such as "communications and networking" with 1108

24 See, for example, work done by the Education Network Australia. EdNA Online. 2003. The EdNA Metadata
, available online at <>, and the listing of activities at
25 See, for example, Gerard Salton, editor. 1971. The SMART Retrieval System: Experiments in Automatic Document
. Englewood Cliffs, N.J.: Prentice Hall.
26 Three of the largest directories include the Open Directory, Yahoo, and LookSmart.

Prepublication Copy--Subject to Further Editorial Correction
entries grouped into 40 subtopics; "supercomputing and parallel computing" with 9 subtopics and 11
sites; and "software" with 50 subtopics.27
Users search a typical, hierarchically-organized directory by starting at the top of the tree and
moving deeper along branches labeled (by the directory editors) with terms that seem to match their
interests.28 Thus, a user interested in administration tools for the DNS could trace the following path in

Directory>Computers and Internet>Communications and
Networking>Protocols>DNS>Administration Tools29

The success of the user's navigation depends on the user's skill in identifying appropriate paths
(but backing up and trying another path is relatively easy), on the editors' skill in describing and placing
the site in the directory, and on the site's accuracy in describing itself to the editors (unless they visit and
characterize each site themselves).
Directories present the user with a taxonomic view of some of the Web sites on the Internet,
which can enable users to reach their goals through successive refinement. However, if the taxonomy is
poorly implemented or does not roughly match the view of the user, it can be very difficult to use. There
are hundreds of directories, general and specific, available on the Web and listed in various directories of
Work is now underway by several groups (a majority with at least loose links to each other)31 to
reexamine analogies to the time-tested "yellow pages" model for the Internet. Whereas search engines
are modeled on a "search through the visible Web and see where appropriate material can be found"
model (augmented, as discussed below, by paid placements), a yellow pages system is one in which all of
the entries are there because their owners want them there, there is considerable content-owner control
over presentation, and while a classification system is used to organize the information, the categories in
which a listing is placed are generally chosen by the listing (content) owner, not the site operator or an
automated process. In its paper form, that type of system has been thoroughly demonstrated over the
years and is useful precisely because the information content is high and the amount of extraneous
material low. Since the combination of several of the factors discussed in this section have resulted in
most searches for products leading to merchants who sell those products, and to a good deal of extraneous
information as well, a directory that is organized the way the merchants want (and pay for) may be more
efficient technically and economically. Of course, these models can apply to many types of listings, goals,
and content other than commercial ones.
One of the major differences between a traditional yellow pages and some of these systems is that
the yellow pages (and even their online versions which support broader geographical scope32) are
organized around a single hierarchical category system. That is necessitated by the organization of the
material for publication on paper. But a computer-based system can take advantage of a multi-hierarchical
environment in which, for example, the location of an entity (at various degrees of precision) is specified
in a different hierarchy from content descriptors (such as category of store), rather than having to be a
superior category in the hierarchy.
Because directories are typically built and maintained by humans, they can cover only a small
portion of the Web, have difficulty keeping up with changes in locations, and are labor intensive.
(However, as noted above, specialized directories that are supported by those who want to be found, such

27 Directory accessed on March 5, 2005.
28 Many directories have added the ability for users to jump directly to a subtopic within a directory.
29 Directory accessed on March 5, 2005.
30 See, for example, the directory of directories at Galileo (Georgia's Virtual Library)
31 For example, the work by Beijing 3721 Technology Co., Ltd. (also known as `3721'), which was recently
acquired by Yahoo!.
32 See <>.

Prepublication Copy--Subject to Further Editorial Correction
as commercial yellow pages listings, can overcome the latter two problems.) One way to address all of
these problems is to decentralize responsibility for maintaining the directory, just as the DNS name files
are decentralized. For example, 475 "guides" who are selected and trained to cover a specific subject area
maintain the directory.33 Its 475 "guides" are responsible for "more than 50,000 topics with
over 1 million pieces of original content, which are grouped into 23 channels."34 Another way, which is
becoming common, is to combine directories with search engines, which index a much larger portion of
the Web using automated processes. Netscape began an ambitious directory project using volunteers
called Open Directory, which continues to this day and is incorporated into Google.

7.1.7 Navigation via Search Engines

Search engines rely on indices generated from Web pages collected by software robots, called
crawlers or spiders.35 These programs traverse the Web in a systematic way, depositing all collected
information in a central repository where it is automatically indexed.36 The selection and ranking of Web
pages to include in the response to a query is done by programs that search through the indices, return
results, and sort them according to a generally proprietary ranking algorithm. Consequently, basic search
by search engines is often referred to as "algorithmic search." (See Box 7.2 for an expanded description
of how search engines work.) When Web site operators pay to have their sites listed in response to
queries, the search is referred to as "monetized search." The desire of Web sites to obtain advantageous
positions in the listing of responses to queries relevant to their offerings has lead to the development of
specialized strategies and tactics that are called "search engine marketing and optimization." Despite
their best efforts, search engines cannot, for a variety of reasons, reach all sites on the World Wide Web.
The untouched part of the Web is called, variously, the "Deep," the "Dark," or the "Invisible Web."
Finally, the searcher can obtain more comprehensive results to a query by looking at the results of
searches conducted by several search engines combined by a "meta-search" engine. The following
sections examine these search engine topics in more detail.

Algorithmic Search

Because they are automated, search engines are the only currently available navigation aids
capable of finding and identifying even a moderate fraction of the billions of Web pages on the public
Internet. To index and retrieve that much information from the Web, search engine developers must
overcome unique challenges in each of the three main parts that make up a search engine: the crawler, the
indexer, and the query engine. (See Box 7.2.)
The principal challenges facing the crawler (robot or spider) are determining which Web sites to
visit and how frequently to do so. Search engines vary substantially in the number of Web sites visited,
the depth of their search and the intervals at which they return. In general, increasing the size of the
computer facility can increase the number, the depth, and frequency of visits and, consequently, it is a
business judgment by the search engine operators that determines these parameters.
The principal challenge facing the indexer is in determining when two terms are equivalent,
bearing in mind singular vs. plural versions, misspellings, differing translations or transliterations,
synonyms, and so on. Moreover, these challenges are language-specific. These problems can be
addressed to some extent during the creation of an index, but are often left to the search engine. A search

33 See <>.
34 Web site. 2005. March 5. For more information see <>.
35 See <> for a wealth of information about search engines and directories.
36 Many search algorithms do not depend on the script or language used, so almost all the Internet's visible Web
pages can be included--regardless of their language.

Prepublication Copy--Subject to Further Editorial Correction
engine typically allows for partial matching to the query and returns many results. Searching for "cook
rutabaga" may return pointers to results with those words, but also to results containing "cooked
rutabaga" or "cooking rutabaga."
The major challenges facing the query engine are matching the query words appropriately and
calculating the relevance37 of each response. Determining the relevance of a response to a query is a
complex problem, since relevance depends on the user's specific needs, which may not be clear from the
words chosen for the query. (See the example of a search for "Paris" in Section 6.1.6.) Different search
engines use different criteria to calculate relevance, with examples being the location and frequency of
words matching the query terms on a Web page and the patterns and quality of links among Web pages.

BOX 7.2
How Search Engines Work

Search engines are technical systems comprising computer programs that perform three
interrelated functions: searching the Web to collect Web pages, indexing the Web pages found, and using
the index to respond to queries.
The Web search programs, called crawlers or spiders, collect Web pages by traversing the Web,
following links from site to site in a systematic way. By beginning with Web sites that are densely
populated with links, a crawler is able to spread across the most frequently accessed parts of the Web, and
then increase speed as it travels to other less frequently visited sites. Search engines use different
algorithms to determine the coverage and depth of the pages that are visited. Some may focus more on
breadth, covering only pages linked from the home page of a Web site, rather than depth, visiting pages
deeper within the Web site's hierarchical structure. Because of the constant changes of Web site content
and the addition of new Web sites, Web crawling by search engines is never completed. The frequency
with which Web pages are re-crawled directly affects the freshness of the results returned from a search
engine query.
Once the Web pages are retrieved, indexing programs create a word index of the Web by
extracting the words encountered on each Web page and recording the URL (and possibly additional
information) of each one. Indexes are then created that list the URLs of all pages on the Web for a given
word. They may also record additional information such as whether the word appeared in a title, anchor
text, heading, or plain text, the relative size of its type, and its distance from other words.
In response to queries, query engines will search the index for the query words to find the pages
where they appear. For example, if the query is for "CSTB committees," the search engine will retrieve a
list of all the pages on which both the strings "CSTB" and "committees" appear. A search may return tens
or thousands or even millions of responses, which users will typically not examine in full. Therefore, the
critical question is the order in which the retrieved list of pages is returned. The goal is to return the pages
in decreasing order of "relevance." The determination of relevance is both critical to the quality of
response and very difficult to assess, because of the large number of indexed pages and the short queries
that search engines typically receive. Each distinct search engine has developed its own proprietary
algorithms for determining relevance.
Google, currently the general-purpose search engine with the largest data base, is noted for its
algorithm, PageRankTM, which uses the link structure of the Web as a key part of the relevance
calculation. It assumes that a link from page A to page B is a vote by page A for the quality of content on
page B. A's vote is given higher weight if the Web site of page A is also highly linked. The relevance
ranking assigned to a page by PageRankTM is based on the intrinsic value of the page plus the

37 Research on relevance in traditional information retrieval has identified dozens of factors, such as degree of
topical match, authority of the source, how current the material is, familiarity of the terminology, ease and cost of
acquiring the content, and so on. See, for example, Linda Schamber. 1994. "Relevance and Information Behavior,"
in Martha E. Williams (ed.), Annual Review of Information Science and Technology, Vol. 29, Ch. 1, pp 3-48.

Prepublication Copy--Subject to Further Editorial Correction
endorsements from the other pages linked to it. The qualitative performance of the PageRankTM algorithm
is better than keyword algorithms alone, since it makes use of more information than just the content on
the pages.38
The retrieved Web pages used to create the index may also be stored, or cached, beyond the time
needed to create the index. The cache allows search engines to increase the speed with which search
results are returned, but they also add to the enormous storage capacity needed to create indexes required
for large-scale search engines. Estimates of the computing resources used by one search engine to index
the Web and handle 200 million queries per day--approximately one-third of the total daily Web
searches--is 54,000 servers, over 100,000 processors, and 261,000 disks.39
Currently operating search engines include AlltheWeb, Alta Vista, Google, Inktomi, MSN
Search, and Teoma.

Relevant results, once retrieved, can be sequenced by other factors that the system or the user
considers appropriate. Few people will view dozens, much less thousands, of matches to a
query; typically, only the first one or a few pages of results are viewed. Among the 1400 or so
respondents to a survey of search engine users in the spring of 2002, only 23 percent went beyond the
second page and only about 9 percent read more than three pages.40 Thus, the matching and ranking
criteria of search engines strongly influence the material that people actually view and use.
No single set of matching and ranking criteria is likely to suit all users' purposes. Consequently, it
is in the general users' interest that multiple search engines employing different relevance criteria be
available. The more options, the more likely it is that users will find a search engine that consistently
ranks highly the content or services they seek. Very skilled and experienced users might even want to
know the criteria by which a search engine ranks its results, enabling them to choose the search engine
whose criteria best meets their needs. However, commercial search services treat the details of their
ranking algorithms as proprietary since they are a primary means of differentiating themselves from their
competitors and of minimizing the capacity of Web site operators to "game" the system to achieve higher
Search engines are able to return a high proportion of all the relevant Web pages that are
available for indexing--though often in such a large number that they exceed the searcher's ability to
review most of them--and to give a reasonable probability of retrieving less well-known Web pages
related to particular topics--though not necessarily in the first few pages of results. Furthermore, since
information on the Web is dynamically changing, heterogeneous, and redundant, no manual system can
list and remain current with more than a small fraction of all sites. Only search engines have the capacity
to keep their indices relatively current by continually revisiting accessible sites, although (as noted above)
the frequency with which sites are visited varies among search engines and, probably also, depends upon
specific site characteristics. As the Web expands, so may the number of responses that search engines
return for a query, although the relationship is not generally proportional. Because of the likelihood of
receiving large numbers of responses to a query, searchers naturally favor search engines whose ranking
of responses reliably provides the pages that best meet their needs close to the top of the listing.

38 See Arvind Arasu, Junghoo Cho, Hector Garcia-Molina, Andreas Paepcke, and Sriram Raghavan. 2001.
"Searching the Web," ACM Transactions on Internet Technology, August, Vol. 1, No. 1.
39 See John Markoff and G. Pascal Zachary. 2003. "In Search of the Web, Google Finds Riches," New York Times,
April 13, Sec. 3, p. 1; and John Markoff, "The Coming Search Wars," February 1, 2004, New York Times.
40 See "iProspect Search Engine Branding Survey," reported in iProspect Press Release. 2002. "iProspect Survey
Confirms Internet Users Ignore Web Sites Without Top Search Engine Rankings," November 14.
Available online at <>.

Prepublication Copy--Subject to Further Editorial Correction
Monetized Search

The growing use of search engines offers the opportunity for information and service providers to
affect the presentation of search results to users through a variety of direct and indirect means. When
viewed from the traditional information retrieval perspective, this appears to contradict a user's "right" or
expectation of neutrality in his or her information sources (except when seeking information from sources
with an obvious viewpoint, such as political or commercial sources.) However, when seen from the
marketing perspective, it is an especially efficient way for providers to reach prospective users just at the
time when the users have expressed interest in what they have to offer.
The search engine companies have responded to the providers' interest through a variety of
advertising, "pay for placement" and "pay for inclusion" opportunities. These practices as a group are
often called "monetized search." Payment can result in having an advertisement placed ahead or
alongside of the search results for specified search terms, having a site visited more frequently or more
deeply by the crawler, having a page assured a place in the results, or having a link placed at the top of the
list of results. (However, concerns about the latter two practices paid inclusion and paid ranking have
led some search engines that offer them either to phase them out or to consider doing so.41)
The consequence of these opportunities is that the first page of results of a search engine query
for, say, "Florida holidays," generally now includes not only the "neutral" results ranked according to the
relevance algorithm used by the engine, but also sponsored listings along the top and/or the sides from
advertisers who paid to have their listings presented whenever the keywords "Florida" or "holiday" or
"Florida holidays" appeared in the query.
Providers' payments have become the major source of revenue for most search engines, which
searchers use for free. (See Section 7.2.2 for a further discussion of advertising and the search engine

Search Engine Marketing and Optimization

The opportunity to pay for listings in response to certain key words on specified search engines
has led to the development of "search engine marketing." Its practitioners help operators of Web sites to
decide which key words to pay for on which search engines to attract the greatest number of prospective
customers to their sites. This is very similar to the role that advertising agencies play in helping
advertisers to decide what ads to run in which media. However, advertising on search engines has the
distinct advantage that the message is presented only to prospective customers who have expressed an
interest in a topic that may be related to the advertisers' wares at the time of their interest. In many cases,
the advertiser pays only if prospective customers actually click on the link that takes them to the
advertiser's site.
Web site operators have also responded to their perceived business need to be ranked higher in
the non-paid results of searches (for example, florists in the search for "flowers") by adopting means to
improve their rankings.42 This has lead to the development of "search engine optimization" in which the
site design is optimized to include simple, common terms likely to turn up in searches and to include
metatags (metadata) that are invisible to users but are picked up by some search engines. Specialized
firms43 have grown up to help companies both in marketing and optimizing their Web sites.

41 See Stephanie Olsen. 2004. "Search Engines Rethink Paid Inclusion," June 23. c/net Available online
at <>.
42 See Mylene Mangalindan. 2003. "Playing the Search-Engine Game," Wall Street Journal, June 16, p. R1.
43 For example,, <>, claims on their Web site that "SEM (search engine
marketing), also known as SEP or SEO, is the art and science of increasing a Web site's visibility across the major
search properties on a strategically defined list of relevant keywords and phrases." Accessed on August 3, 2003.

Prepublication Copy--Subject to Further Editorial Correction
Other approaches are less savory. For example, "pagejackers" falsify information in the meta
tags on their site to emulate the appearance of another Web site that would rank higher; while
"spamdexers" seek placement under search terms that are unrelated to the content of their pages by
placing many repetitions of the unrelated terms on their site in invisible form (e.g. white text on white
background, which can nevertheless be read by the search engine). In response to providers' tactics,
search engines eliminate from their databases those companies that they believe are using unscrupulous
methods to improve rankings.44 Not surprisingly, this has led to a continuing battle between Web site
operators trying new ways to improve their ranking and search engines introducing countermeasures as
each new tactic is discovered.

The Deep, Dark, or Invisible Web

Although they are far more comprehensive than directories, general search engines still index and
retrieve only a portion of the content available on the Internet. First, they do not reach every page that is
visible to them because of limits on how often they will crawl the Web, on the capabilities of their
crawlers, and on how much of each visited site they will crawl. Second, a large majority of the
information potentially reachable on the Web is not visible to them. The parts that they cannot see are
called the "deep," the "dark" or the "invisible" Web.45 In 2000, the latest date for which estimates are
available, the invisible Web was estimated to be 400 to 450 times larger than the visible or public World
Wide Web.46 Web pages can be invisible to search engines for a variety of reasons.47
A primary reason is the increasing use of databases to deliver content dynamically. If material is
available only in response to queries and actually does not exist until a question is asked, there is no
practical way for a general search engine's crawler to "see" it since those systems cannot synthesize the
queries that will generate the relevant material. Thus, engines48 cannot crawl inside searchable databases
such as library catalogs, the Thomas register of manufacturing information, or indexes of journal
literature. A search engine query on "Shakespeare" may retrieve sites that specialize in Shakespearean
memorabilia (as described in their Web pages), sites of theaters that are currently performing
Shakespearean plays, and Shakespeare fan clubs, but usually will not retrieve catalog records for books in
libraries or for records in archives. There are Web sites that serve as directories to many "invisible"
resources, such as library catalogs and databases, on the Internet.49 Moreover, specialized search engines,
such as those used by shopping comparison50 or travel reservation sites,51 are designed to synthesize the
appropriate queries and submit them to multiple databases in order to obtain the comparison information
requested by the user. Furthermore, as noted earlier, directories can be useful in this situation since they
can manually identify or seek submission of database sites, enabling the searcher to find a relevant
database and then submit a specific query to it.

44 See, for example, Google's guidelines at <>.
45 See Chris Sherman and Gary Price. 2001. The Invisible Web. Medford, N.J.: Information Today, Inc.
46 See Michael K. Bergman. 2001. "The Deep Web: Surfacing Hidden Value," August, Journal of Electronic
, Vol. 7, Issue 1. Available online at <>.
47 See Tyburski, Genie and Gayle O'Connor. April 3, 2003. The Invisible Web; Hidden Online Search Tools.
Presentation to ABA Techshow, 2003, Chicago. Available at <>.
48 See Clifford A. Lynch. 2001. "Metadata Harvesting and the Open Archives Initiative," ARL Bimonthly Report
217, 1-9.
49 One such directory is "Those Dark Hiding Places: The Invisible Web Revealed," which can be found at
50 Such as, for example, epinions ( and bizrate (
51 Such as, for example, Travelocity (, expedia (, or metasites,
such as Sidestep ( and Mobissimo (

Prepublication Copy--Subject to Further Editorial Correction
Content virtualization52 (a form of dynamic data) produces an increasing variety of information
unavailable to Web crawlers. Google has made an attempt to overcome some of the challenges of rapidly
changing content and the increasing use of "virtual" content, particularly among large news Web sites, by
creating special arrangements with news organizations to continually update, retrieve, and index content
from a pre-selected list of news Web sites.53
The dark Web also encompasses the vast intranets of many corporations, governments, and other
organizations. Resources are not indexed if they are behind firewalls, require payment, or are otherwise
coded "off limits" to search engines.
Progress on the dark Web problem is being made via efforts such as the Open Archives Initiative
(OAI) that enables information providers to offer their metadata for harvesting.54 Additionally, new kinds
of Web pages for which indexing is probably infeasible are emerging. Personalized content such as the
"My Yahoo personal news page" is an example of this type of content since it is created only when the
user requests it and is not available to others on the Web. However, it comprises a selection of materials
from public Web pages, so indexing it would add little other than information about an individual's
selections--and that would probably be constrained by considerations of privacy. Some ephemeral
content, such as "instant messaging" and "chat," is not indexed because it is dynamic and not readily
accessible to search engines unless it is archived. Google indexes the complete Usenet archives and the
archives of many important Internet mailing lists.
Another source of new and rapidly changing information on the Web is the profusion of journals
in the form of Web logs, called blogs. Software that has made it very easy for individuals to construct and
modify Web sites and the availability of services to house them have made Web publishing common.
The ease of creating and changing blogs has lowered the barriers to entry for publishing to large
audiences. This has led to their increasing number, types, and ranges of quality. They often incorporate a
large number of links to other blogs and Web sites, making them a distinctive medium that contains not
only the authors' contributions, but also the authors' identification of "communities of interest" whose
ideas are germane to theirs.
These new forms of content pose further requirements for search engines to retrieve information
more frequently and also to return the wide variety of information posted to Web logs in a useful way.
Some of these challenges are being met by specialty search engines, which go beyond the features
presented by Google. One of these is Daypop,55 which uses its own kind of link analysis to identify Web
logs that are pointed to other Web log sites from their front pages, rather than from archived or back
pages, and allows popular commentary, or responses to original posts on a Web log, to be comparatively
searched across a number of Web logs.
Finally, multimedia materials are increasingly populating the Internet. They range from still
photographs to full-length videos and films. These can readily be indexed by their descriptions, but
unless a human indexer provides descriptive terms--metadata a photo or song cannot yet be
automatically searched and indexed by its content at a commercially-viable scale.
The inability to search everything accessible via the Internet is not necessarily a problem, since
much of it may not be useful and there is necessarily a cost associated with sorting through ever larger
amounts of material. However, while some of the unsearchable material may be of little value, it is likely

52 "Content virtualization--or content integration as some know it--leaves data in its originating system and pulls it
together in real time when requested by the user." Quoted from Lowell Rapaport. 2003. "Manage Content
Virtually," Transform Magazine, April. Available at
53 For more information about news aggregation services by Google, see <>. For other
approaches to news aggregation services see Blogdex at <>.
54 Scholarly Publishing and Academic Resources Coalition. The Case for Institutional Repositories: A SPARC
Position Paper. Available online at <>.
55 See Greg Notess. 2002. "The Blog Realm: News Source, Searching with Daypop, and Content Management,"
Online. Vol.26, No.5. For more information about Daypop see <>; see also Technorati for
another search engine approach to Weblogging, at <>.

Prepublication Copy--Subject to Further Editorial Correction
that some of it (say, major library catalogs or government databases) would be of great value, if it could
be readily searched.

Metasearch Engines

"Metasearch" engines take the keywords entered by the user and submit them to a number of
independent search engines. They present the combined results to the user. This would appear to offer
the advantage of time saving and the possibility of getting the best results from the sum of the engines
searched. However, whether those advantages are realized depends upon how the metasearch engine
combines and orders the results from several search engines, each using different relevance and ranking
criteria. Among the metasearch engines available in 200456 were Dogpile, Vivisimo, Kartoo, and

7.1.8 Use of Navigation Aids

How much use is made of these aids to navigation? Complete data are unavailable, but use of
search engines (and their associated directory services) is being measured. According to the Pew Internet
& American Life Project:57

In total, Americans conducted 3.9 billion searches in June 2004.
The average search engine user conducted 33 searches in June 2004.
Eighty-four percent of Americans who use the Internet have used search engines--more than
107 million people.
On an average day, about 38 million of the 64 million Americans who are online use a search
engine. Two-thirds of Americans who are online say they use search engines at least twice a week.
Using search engines is second only to using e-mail as the most popular Internet activity,
except when major news stories are breaking when getting the news online surpasses using search
"There is a substantial payoff as search engines improve and people become more adept at
using them. Some 87% of search engine users say they find the information they want most of the time
when they use search engines."
"More than two-thirds of search engine users say they consider search engines a fair and
unbiased source of information."
"... 92% of searchers express confidence in their skills as searchers--over half of them say
they are `very confident' they can accomplish what they want when they perform an online search."
" ... 44% of searchers say that all or most of the searches they conduct are for information
they absolutely need to find."
"A third of searchers say they couldn't live without Internet search engines." However, about
a half say that, while they like using search engines, they could go back to other ways of finding

56 For a listing of metasearch engines see Chris Sherman. 2004. Metacrawlers and Metasearch Engines. March 15. Available at: <>.
57 See Deborah Fallows, Lee Rainie, and Graham Mudd. 2004. "The Popularity and Importance of Search Engines."
August. Data Memo. Pew Internet & American Life Project. Available at:
<>. The results came both from a telephone
survey of 1399 Internet users and tracking of Internet use by comScore Media Metrix.

Prepublication Copy--Subject to Further Editorial Correction
How do Internet users deploy the available aids to navigation? Do they generally go to the Web
sites they seek by entry of a domain name or keyword or through bookmarks? Do they follow hyperlinks
from Web page to Web page? Or, do they commonly make use of search engines and directories? There
is little publicly available research that addresses these specific questions. One analysis, based on survey
data from March 2003 and one year earlier, provides some insights.58 According to that analysis, search
engines produced 13.4 percent of site referrals on the day measured, which was an increase from 7.1
percent one year earlier. Navigation through entry of a known or guessed URL or use of a bookmark also
increased from 50.1 percent to 65.5 percent over the year. The decline occurred in the flow along
hyperlinks, which decreased from 42.6 percent to 21 percent. This survey indicates that Internet users
tend to use certain sites and services consistently, visiting them repeatedly, using their bookmarks or
remembered URLs. This suggests that much of their Web use is routine: checking e-mail, visiting a few
standard sites, and exchanging instant messages with some Internet buddies. The need for search engines
or directories arises primarily when a user needs a specific piece of information or wants a new or a
replacement source of information, entertainment, or other material.
When a search is required, a survey of 1403 Internet users in spring 200259 showed a strong user
allegiance to one or a small number of navigation services. More than half (52 percent) generally relied
on the same search engine or directory and close to 35 percent used several interchangeably. Only 13
percent used different services for different kinds of searches. With respect to the usefulness of search
engines, at that time almost half (45.9 percent) believed their searches were successful almost always.
When they were not successful, 27 percent of the users switched to another search engine, rather than
refining the search with more terms--only 7.5 percent did that. One third of the users felt that their
searches were successful three-quarters of the time and 13 percent reported successful searches only half
the time. (Though presented differently, these figures are not inconsistent with the results of the Pew
study reported above.)
Users of diverse skills and interests, located across the world have a range of information and
services literally "at their fingertips," whether at work, at home, or on the road, that far exceeds that
available even to information specialists before 1993. This is true especially for those seeking
commercial products and services. Comparable navigation tools for scholarly and public interest materials
are generally less well developed.

Conclusion: The further development of Internet navigation services, such as subject-specific
directories, that enable discovery of specialized databases and similar resources not readily indexed by
search engines, is desirable. They can be of particular value to non-commercial groups, whose
information resources may not be able to support active marketing.

Conclusion: As the material accessible through the Internet continues its rapid increase and its
societal importance grows, Internet navigation aids and services will need to improve further so as to
deliver more precise responses, in more convenient forms, to more diverse questions, from more users
with widely varying skills.

Prospective improvements in Internet navigation technology and processes are discussed in
Section 8.1.

58 The data were collected on March 6, 2003 by WebSideStory's StatMarket from about 12 million visitors to
125,000 sites using its proprietary analytical platform and compared with figures from the previous year. Reported
in: Brian Morrissey. 2003. "Search Guiding More Web Activity," CyberAtlas. March 13. Available online at
59 See "iProspect Search Engine Branding Survey," reported in iProspect Press Release, 2002. "iProspect Survey
Confirms Internet Users Ignore Web Sites Without Top Search Engine Rankings," November 14.
Available online at <>.

Prepublication Copy--Subject to Further Editorial Correction


In contrast to the provision of domain name services, Internet navigation is not the function of a
single integrated technical system. While there is just one Domain Name System, there are many ways of
navigating the Internet, only three of which involve distinct technical systems dedicated to navigation--
KEYWORDS,60 search engines, and directories. Moreover, the institutional framework of the technical
systems supporting Internet navigation is an open market, with many independent and competing
providers offering their services. While some providers are non-profit or governmental institutions, such
as national libraries or professional societies, the most frequently used navigation systems are provided by
commercial organizations. This section will concentrate on the commercial market for directory and
search engine services.

7.2.1 The Commercial Providers of Navigation Services

As noted in Section 6.2.2, the early distinctions between providers of directories and providers of
search engines--when each Web search site featured either algorithmic search engine results or human-
powered directory listings61--have increasingly become blurred. Technology has helped to automate
some of the classification processes for the Yahoo! directory,62 and most general-purpose Web search
sites now feature search results from both human-based directories and crawler-based search engines,
with one type providing the majority of search results. See Table 7.2 for a listing of navigation services
and the sources of the results they provide.
The navigation services market is dynamic. The relationships shown in Table 7.2, which applied
in July 2004, are continually changing. For instance, the early search engine Lycos, which began in 1994,
ceased providing its own search listings in April 1999, and has since used AllTheWeb to power its Web
search site. Google, which generates it own Web search results, also provides algorithmic search services
to others such as AOL and, until March 2004, Yahoo!, which paid Google $7.2 million in 2002 for the
search queries it handled.63
Over the past 4 years from 2000 to 2004 in the United States, Google rose from eleventh position
among navigation sites with a 5.8 percent market share in December 2000, as measured by "audience
reach,"64 to first position with an estimated share of 34.7 percent in February 2004, as measured by "share

60 In June 2004, the commercial market for KEYWORDS comprises primarily AOL, Netpia, and Beijing 3721.
Yahoo! purchased Beijing 3721 in 2004.
61 See Danny Sullivan. 2002. How Search Engines Work. October 14. Available at
62 In "A History of Search Engines," Wes Sonnenreich explains that "as the number of links grew and their pages
began to receive thousands of hits a day, the team created ways to better organize the data. In order to aid in data
retrieval, Yahoo! became a searchable directory. The search feature was a simple database search engine. Because
Yahoo! entries were entered and categorized manually, Yahoo! was not really classified as a search engine. Instead,
it was generally considered to be a searchable directory. Yahoo! has since automated some aspects of the gathering
and classification process, blurring the distinction between engine and directory." Available online at
63 See Yahoo proxy statement filed March 2002, p. 30. Available online at
64 Nielsen NetRatings reported in Danny Sullivan. 2003. Nielsen NetRatings Search Engine Ratings. February.
Available online at <>. comScore Media
Metrix press release, April 29, 2004. "Audience reach" is the percentage of U.S. home and work Internet users
estimated to have searched on each site at least once during the month through a Web browser or some other
"online" means.

Prepublication Copy--Subject to Further Editorial Correction
of search."65 (See Figure 7.1) The previous leading Web navigation site, Yahoo!, fell from a 48 percent
share to second position with 30 percent during that time. Two of the other high-ranking Web navigation
sites were MSN with 15.4 percent and AOL with 15 percent of searches in February 2004. However, note
that during that period Inktomi provided search services for MSN, while Google provided search services
for AOL. For international Internet users (English using populations), Google had an even larger lead in
February 2004, capturing more than 43 percent of searches to Yahoo's 31 percent, MSN's 14 percent, and
AOL's 7 pecent. (Since Google still provided search results to both Yahoo! and AOL in February 2004,
its actual share of searches was closer to 80 percent, both internationally and in the U.S. After March
2004, without Yahoo!, its share dropped to 50 percent.)

TABLE 7.2 Navigation Services and the Providers of their Results66
Process Used to
Provider Of
Provider of Paid
Provider of
Main Results
Directory and/or
Main Results
Backup Results
Search AllTheWeb

Yahoo! acquired)
Alta Vista
Search Alta

Yahoo! acquired)
AOL Search
Ask Jeeves
Google Open

Google Search Google Google Open
Choice of: Inktomi


LookSmart Directory LookSmart/Zeal LookSmart Backup
Lycos Search AllTheWeb

MSN Search

Netscape Search Google Google Open

65 The new metric generated monthly by comScore Media Metrix, beginning in January 2003, provides a better
measure of market share by focusing on the number of searches that a search engine handles per month rather than
the number of searchers that perform at least one query on the Web search site. The Web search site queries are
based on a panel of 1.5 million Web users located within the U.S. and in non-U.S. locations. The February 2004
results are from a comScore Media Metrix press release on April 29, 2004. Available online at
66 Based on, 2003. at
<> and updated in March 2005.

Prepublication Copy--Subject to Further Editorial Correction

Paid Overture

Directory Open n/a n/a
Search Teoma
Google n/a
(Ask Jeeves-

Yahoo! Search/Directory Inktomi
(Yahoo! owned)




FIGURE 7.1 Company share of U.S. Web searches by home and work users, February 2004. SOURCE:
comScore Media Metrix qSearch press release, April 28, 2004, available at

7.2.2 The Business of Internet Navigation

The primary source of income for commercial Internet navigation services, which provide access
to material on the public Internet, has become selling advertising and placement on their sites.67
Consequently, as in many broadcast media, it is the content and service providers who are subsidizing
users' access to navigation services in order to present advertisements to them at the time of their
expressed interest in a topic. This contrasts sharply with traditional commercial information search
services, such as Lexis, Westlaw, and Dialog, which have obtained their income directly from their users
who pay to access the services' proprietary (but free of marketing influence) information. Typically,

67 Commercial search engine companies are exploring possibilities beyond their own search sites. For example,
publishers such as the Washington Post have turned to Google or Overture to sell advertisements associated with the
content that a visitor selects. See "If You Liked the Web Page, You'll Love the Ad," Bob Tedeschi, New York
, August 4, 2003, available online at <>. In
addition, Google and others license their search engine technology for use by other Web sites and by corporate

Prepublication Copy--Subject to Further Editorial Correction
those pay-for-access companies also provide other services, such as training, documentation, and
extensive customer support, to their users.
The advertising that supports search services can take several forms: banner advertisements,
popup advertisements, or search-linked ads. Banners are typically displayed at the top or side of a Web
page and are generally priced on a per-impression (view) basis, which means that the advertiser pays
based on how many people see its advertisement with prices quoted in CPMs (cost per thousand
impressions), as is traditional in the advertising industry. A typical rate for a generic banner advertisement
is 2 cents per impression, or $20 CPM. Banner sizes are standardized so that sellers and buyers of
advertising space can find it easy to negotiate pricing and other contract terms.68 So-called "skyscrapers"
are vertically oriented banner ads. Popup advertisements are similar to banners, except that they pop up as
separate window. Their shape is also standardized. (The intrusive nature of popup advertisements has led
to a variety of software products--separate programs or browser features--that automatically prevent
them from appearing.69)
Search-linked advertisements appear as the result of a search. For example, the searcher
mentioned above who enters the keyword search term "Florida vacation" might see advertisements for
Florida hotels, condo rentals, theme parks, towns and cities and the like. These may be displayed as
banners, popups, sidebars, or--as noted earlier--presented with the search results themselves.
Sophisticated algorithms are used by the search services to select which advertisements will appear.
These algorithms take into account, among other things, the amount the advertiser is willing to pay if the
user clicks on the advertisement, the relevance of the advertisement, and the historic success of the
advertisement in generating clicks. All services place limits on the number of advertisements they will
Not surprisingly, search-linked advertisements are much more valuable than generic banners.
They are priced both by impression and by click-through. Practices differ among search services, but
Google displays up to 2 advertisements at the top of the page (which it calls "Premium Sponsorships" and
up to 8 advertisements on the right side of the page (which it calls "Adwords Select"). Typically, these
ads can appear on every page of results.
At one time Google priced the top ads on a CPM basis and the side ads on a CPC (cost per click)
basis. In the latter case, the advertiser pays only when the user actually clicks on an advertisement.
However, Google has eliminated the CPM pricing and now all its ads are priced on a CPC basis. The ads
placed at the top of the page are chosen from the side ads on the basis of price and performance. Prices
for a click-through are over 10 times as much as the price of a generic impression.70 However, some
search engines will drop a click-through advertiser who does not produce a sufficient number of hits.
The two leading providers of search-linked advertisements (or monetized search) are Overture
(now owned by Yahoo!) and Google, which also distribute search-linked advertisements to other search
sites. The paid listings provided by Overture to its affiliated network of Web search sites, including
Yahoo!, MSN, Infospace, and Alta Vista, have been estimated to have handled 46.8 percent of all U.S.
based paid searches; and the paid listings provided by Google, appearing on the search results pages of
Google, AOL, Infospace and Ask Jeeves, accounted for 46.6 percent of all U.S. paid searches in January

68 Further information regarding these standards may be found at <>.
69 Additionally, software known as "adware" or "spyware" has been developed that is installed on a user's computer
and covertly gathers information about a user during navigation of the Internet and transmits such information to an
individual or company. In turn, the individual or company transmits information back to the user's computer to
display specific banner advertisements based on the user's navigation of the Internet. Such activity has resulted in
state legislatures considering or enacting spyware regulation laws; see, for example, Utah Spyware Control Act
(H.B. 323 2004); California (A.B. 2787 April 13, 2004; S.B. 1436 March 23, 2004); Iowa (S.F. 2200 March 1,
2004), and litigation to undo such laws; see, for example,, Inc. v. Utah, No. 040907578 (Utah Dist. Ct.
3d Dist. April 23, 2004).
70 From the Annual Report, Overture, January 2003.

Prepublication Copy--Subject to Further Editorial Correction
2003.71 Google provides search services to several hundred other partners, in the U.S. and abroad,
although AOL and Ask Jeeves are the biggest U.S. customers.
Search-linked advertisements have been very successful. According to its initial public offering
(IPO) prospectus, Google had revenues of $961.9 million in 2003 and profits of $105.6 million, but
without some unusual provisions, its operating profit margin is 62 percent. Before its acquisition by
Yahoo!, Overture reported revenues of $103 million in 2000, $288 million in 2001, and $668 million in
2002. In 2003 it claimed over 95,000 advertisers, who received over 646 million clicks in the second
quarter of 2003 for which they paid an average of 40 cents per click.72 These figures dramatically
illustrate that Internet navigation services--unlike many other Internet services that were tested in the
Nineties--have apparently found a financial model that is capable of supporting them and enabling their
continued development. At the same time, the struggle to capture advertising dollars has been one of the
forces driving the continuing consolidation of the industry, as some of the most successful search services
have acquired their competitors in order to increase their share of the market.
Spending on paid listings (guaranteed separate listings on the search engine results pages) and
paid inclusion (guaranteed inclusion in the regular search engine results, but ranking not assured)73--two
of the three forms of search related marketing--grew by 40 times in 4 years from $80 million in 2000,
$290 million in 2001, to $900 million in 2002, to $2.5 billion in 2003, and was expected to reach $3.2
billion in 2004.74 Globally such spending is expected to grow fivefold to $7 billion a year by 2007, from
$1.4 billion in 2002. Outside the United States, tenfold growth to $2 billion in 2007 from about $200
million in 2003 is expected. 75
The revenues from monetized search are often shared between the site and the search service that
provides the advertisements. For example, Google currently provides algorithmic search and monetized
search for AOL and they split the revenues from the monetization. Overture provided monetized search
for Yahoo! on shared revenue basis until its acquisition, while Google provided algorithmic search
service to Yahoo! until March 2004 for a flat fee. Google and Overture both use an auction model to price
their search-linked advertisements: users can specify a price that they are willing to pay for various
positions and the highest bidder gets the highest position in response to a specific query. For example, a
rental car agency could bid to be listed first in any search for "rental cars." Minimum bids vary, but
generally the range of bidding starts at 5 cents a click and goes up to $100 for some mortgage-related
items, although Google caps bids at $50. The model is sufficiently popular that, as noted earlier, a
secondary market of "search engine marketers/optimizers" has arisen to advise Web sites on how to
optimize their bidding for queries.76 The details of the auction systems differ, but the advantage of
auctions is that hundreds of thousands of prices can be set by actual demand rather than having to be
preset and posted.
Since these auctions are subject to gaming, navigation services actively watch for potential fraud
by advertisers and monitor the content of advertisers with editorial spot-checking. If they suspect
cheating, the advertiser will be removed from bidding. To become qualified bidders, advertisers provide

71 See <>.
72 Data collected on August 16, 2002 and December 4, 2003 from
73 Ask Jeeves announced in June 2004 that it was phasing out its paid inclusion program because its algorithmic
search had become sophisticated enough to find all necessary Web sites and refresh them as required, making
paying for inclusion unnecessary. See Stephanie Olsen. 2004. "Search Engines Rethink Paid Inclusion," June 23
c/net Available at <>.
74 Source: Wall Street Journal On-line, accessed May 5, 2004. Data from InterActive Advertising Bureau,
PriceWaterhouseCoopers LLP, eMarketer.
75According to U.S. Bancorp Piper Jaffray as reported by Mylene Mangalindan. 2003. "Playing the Search-Engine
Game," Wall Street Journal. June 16, Available online at <
76 See sites such as Wordtracker, <>; <> for
a description of their Keyword Bid Optimizer (KBO); and Traffick at <>.

Prepublication Copy--Subject to Further Editorial Correction
information about themselves, their business, their interests, the keywords on which they wish to bid, and
how much they wish to bid.
These advertiser-driven business models for navigation services contrast with the non-
commercial model of neutral information searching and navigation of public and academic libraries,
although they more closely resemble the business models for newspapers and other media where
advertising and editorial matter are expected to be rigorously separated. Nevertheless, users need to be
cautious about how they treat the results of Internet searches, especially those about subjects with
commercial significance. As noted above, the major search services currently identify the sponsored
results ("sponsored links" or "sponsored search listings") and set them off from the direct results of the
algorithmic search, following the newspaper model. As long as the distinction is clear and users are aware
of it, sponsored search should present few problems while providing the great benefit of "free" search
services to the user. However, the potential for abuse exists. It would be possible, for example, for a
search service to accept payment for assured placement in the "top ten" of what would appear to be a
neutral listing. (None have been accused of doing so, but some will accept payment to assure inclusion,
but not ranking, in the otherwise neutral listing.) Or the distinct placement and typography of the
sponsored listing could be weakened to the point that a casual user would not be aware of its difference
from the algorithmic search results. Thus far, competition among sites and third party evaluations have
served as important countervailing forces. Should abuses grow, however, search services could find
themselves under increased public pressure for government scrutiny or facing more disputes and criticism
concerning such activities from other commercial entities. (See the discussion in Section 8.2.1.)

7.2.3 The Navigation Services Market

As seen above, a large number of navigation services have entered the market, attempting to
achieve profitability by selling advertising. Although it can be very profitable, this has turned out to be a
difficult and expensive venture. Furthermore, competition among search engines has forced them to
invest in improved software and extensive computer and storage facilities with substantial
communications capacity to increase the breadth, depth, and frequency of their coverage of the Web.


Over the past four years, there has been considerable consolidation in the search services
market.77 Several large search engine service providers have left the market, and others have been
combined into a single firm. At the same time, there has been increased vertical integration as operators
of Web sites have acquired operators of search engines. In 2003, Overture, with a primary focus on
providing paid search listings, acquired Alta Vista for $140 million and AlltheWeb, the Fast Search and
Transfer (FAST) Web search unit, for $70 million. Overture aimed to strengthen its core business of paid
search listings by eventually integrating it with their algorithmic search and paid inclusion services.78
Also in 2003, however, the directory-based Yahoo! purchased the search engine Inktomi for $235 million,
and then in July 2003, acquired Overture for $1.62 billion.79 Google, while not an aggressive acquirer,
went public with an IPO in 2004 that raised $1.67 billion,80 which provided it with a war chest that can be

77 See, for example, <>.
78 See Brian Morrissey. 2003. "Overture to Buy FAST," Australia February 26, available online at
79 See, "Yahoo! to Acquire Overture" 2003 press release. <http://www.corporate->; and Mylene Mangalindan, Nick
Wingfield, and Robert Guth. 2003. "Rising Clout of Google Prompts Rush by Internet Rivals to Adapt," The Wall
Street Journal
, July 16.
80 See Dawn Kawamoto and Stephanie Olsen. 2004. "Google Gets to Wall Street ­ and Lives." August 19. c/net Available at: <

Prepublication Copy--Subject to Further Editorial Correction
used for acquisitions. The one new player that has entered the search services market is Microsoft, which
built the staff and technology to launch its own search service, which it is very likely to integrate into its
next-generation operating system. In February 2005, Microsoft unveiled a revised MSN search that bore
strong visual similarities to Google's search interface.81
Yahoo! has apparently decided to vertically integrate by buying both a paid listing provider and a
search engine. It is now able to produce by itself the paid listings previously supplied by an independent
Overture and the algorithmic search services previously provided by Google. The net result of this latest
phase of consolidation is that there are only a few major independent navigation services left--Google
and Yahoo! are the largest.
In 2004, Google, which then provided search services to both Yahoo and AOL, actually had 80%
share of searches. This dropped when Yahoo! replaced Google with its own algorithmic search engine.
However, whether the acquisitions result in sustained shifts in search shares will depend upon whether,
for example, users of the Yahoo! search site continue to search there or instead switch to another site that
uses Google. These changes in the search services industry are likely to influence other Web search sites,
primarily AOL, which currently outsource both Web search results and paid listings, to consider creating
or acquiring their own in-house services. Now that Microsoft has entered the search engine competition, it
is possible that AOL will feel compelled to do the same.


In the past, as described in Section 6.2, there has been a cycle of innovation, adoption, and
displacement of navigation services. It began when some new search engine or directory emerged with
new technology, or a better user interface, or both than the incumbent favored service. The new service
attracted attention and gained market share. Then as it and the