Mr. Chairman and members of the Subcommittee, my name is Michael Froomkin. I would like to thank the Subcommittee for inviting me to appear today at this hearing on ICANN governance. I commend the Subcommittee for its wisdom and foresight in recognizing the importance of this issue.

I believe it is useful to separate this complex issue into three parts: (1) ICANN's mission or, if you will, ICANN's "jurisdiction"; (2) ICANN's internal organization; (3) The extent to which ICANN is subject to oversight by the Commerce Department, the U.S. Congress, or any other outside forces.

These three issues are intertwined. The nature and extent of ICANN's powers over the Internet and over Internet users that determines the type of internal governance structures which are appropriate for it. Similarly, the nature and quality of both ICANN's powers and its internal representativeness, not to mention checks and balances, determines the extent to which it needs to be subjected to searching external oversight. In particular, it is appropriate for this committee to enquire into the nature of the workings of the relationship between the Department of Commerce and ICANN.
 
 

ICANN's go-very-slow policy on new gTLDs had no technical basis. Why then would ICANN adopt such a policy? The reason is that ICANN's policies are a product of an internal deliberative process that under-weighs the interests of the public at large and in so doing tends towards anti-competitive, or competitively weak, outcomes skewed by special interests
 

ICANN routinely claims to be either a technical standards body or a technical coordination body. If this were correct, then it might be proper for the Department of Commerce to defer to ICANN's presumed technical expertise and rely on ICANN's standards or allocation decisions without undertaking independent Administrative Procedure Act (APA) -compliant processes of its own. When, however, ICANN acts as policy-making rather than a standard-making body, then due to ICANN's unrepresentative nature its decisions do not carry any presumption of regularity or correctness and the US Government cannot rubber-stamp its decisions without additional independent fact-finding and deliberation.
 

We would all be better off if ICANN could confine itself to true standards issues, or to true technical coordination. If ICANN cannot, then ICANN needs to be subjected to constant scrutiny.
 

Terminological note: A "registrar" is a firm that contracts with clients ("registrants") to collect their information and payment in order to make a definitive and unique entry into a database containing all domain names registered in a top-level domain (TLD). This database is maintained by a "registry." Top-level domains are sometimes grouped into "generic TLDs" (gTLDs), which are currently three- or four-letter transnational domains, and "country code TLDs" (ccTLDs) which are currently two-letter TLDs. The "root" is the master file containing the authoritative list of which TLDs exist, and where to find the authoritative registries that have the data for those TLDs. Registrants typically register second-level domains (e.g. myname.com), but sometimes are limited to third-level domains (e.g. myname.genericword.com).
 

I. ICANN's Mission
 

ICANN's processes little resemble either standard-making or technical coordination. To date, ICANN's "standard making" has produced no standards. ICANN's "technical coordination" has been neither technical nor has it coordinated anything. Rather, in its initial foray into the creation of new gTLDs, ICANN has acted like a very badly organized administrative agency. Instead of engaging in standards work, ICANN is instead engaged in recapitulating the procedural early errors of federal administrative agencies such as the Federal Communications Commission (FCC)..
 

What real standard-making would look like

A standard-based (or, at least, standardized) approach to gTLD creation would required ICANN to craft a pre-announced, open, neutral, and objective standard of competence rather than to pick and choose among the applicants on the basis of the ICANN Board's vague and inconsistent ideas of aesthetic merit, market appeal, capitalization, or experience. All applicants meeting that standard would be accepted, unless there were so many that the number threatened to destabilize the Internet (as noted below, if there is such a number, it is very large). ICANN might also put in reasonable limits on the number of TLDs per applicant, and on sequencing, in order to keep all of them going online the same day, week, or month.
 

Under a standards-based approach ICANN would have tried to answer these questions in the abstract, before trying to hold comparative hearings in which it attempted to decide to which of specific applicants it should allocate a new gTLD registry:

• What is the minimum standard of competence (technical, financial, whatever) to be found qualified to run a registry for a given type of TLD?
• What open, neutral, and objective means should be used to decide among competing applicants when two or more would-be registries seek the same TLD string?
• What are the technical limits on the number of new TLDs that can reasonably be created in an orderly fashion per year?
• What open, neutral, and objective means should be used to decide among competing applicants, or to sequence applicants, if the number of applicants meeting the qualification threshold exceeds the number of gTLDs being created in a given year?

Today, reasonable people could no doubt disagree on the fine details of some of these questions, and perhaps on almost every aspect of others. Resolving these issues in the abstract would not necessarily be easy. It would, however, be valuable and appropriate work for an Internet standards body, and would greatly enhance competition in all the affected markets.

Once armed with a set of standards and definitions, ICANN or any other allocation body, would be on strong ground to reject technically incompetent or otherwise abusive applications for new gTLDs, such as those seeking an unreasonably large number of TLDs. A thoughtful answer would inevitably resolve a number of difficult questions, not least the terms on which a marriage might be made between the Department of Commerce's "legacy" root and the so-called "alternate" roots.
 

What technical coordination would look like
 

An alternate approach to gTLD creation, one that would most certainly enhance competition, would take its inspiration from the fundamental design of the Internet itself-and from major league sports. The Internet was designed to continue to function even if large parts of the network sustained damage. Internet network design avoids, whenever possible, the creation of single points of failure. When it comes to policy, however, ICANN is currently a single point of failure for the network. A solution to this problem would be to share out part of ICANN's current functions to a variety of institutions.
 

In this scenario, ICANN would become a true technical coordination body, coordinating the activities of a large number of gTLD policy partners. ICANN's functions would be: (1) to keep a master list of TLDs, (2) to ensure that there were no 'name collisions' - two registries attempting to mange the same TLD string; (3) to fix an annual quota of new gTLDs; (4) to run an annual gTLD draft; (5) to coordinate the gTLD creation process so that new gTLDs came on stream in an orderly fashion instead of all at once.
 

Each of ICANN's policy partners would be assigned one or more draft choices, and then ICANN would randomly (or, perhaps, otherwise) assign each one their draft picks. As each policy partner's turn came up, it would be entitled to select a registry - imposing whatever conditions it wished - to manage any gTLD that had not yet been claimed on ICANN's master list. In keeping with the transnational and public/private nature of the Internet, ICANN's policy partners could be a highly diverse mix of international, national, and private "civil society" bodies.
 

While I think this alternate solution would best achieve the ends of internationalization, competition, and diversity, it might well require legislation since it is unclear if the Department of Commerce has the will (or the authority) to implement such a plan, and we have seen no sign that ICANN is about to divest itself of any policy authority unless forced to do so.
 

What ICANN actually did: select an arbitrarily small number of gTLDs based on arbitrary appraisals of aesthetic merit, market appeal, capitalization, and experience.
 

Rather than adopt either a standards or a technical coordination approach, ICANN instead adopted an arbitrary approach. First it set an arbitrarily low ceiling on the number of TLDs, then it allocated most but not all of that quota based on its arbitrary appraisals of the applicants aesthetic merit, market appeal, capitalization, and experience.
 

ICANN's decision to impose an arbitrary limit on the number of new gTLDs
 

The closest thing to technical standards work that ICANN has done to date was to adopt an artificially low limit on the number of gTLDs it would recommend the Commerce Department create - under the guise of a so-called "proof of concept". The grounds on which ICANN based this arbitrarily low limit on the number of new gTLDs demonstrate as clearly as anything else that ICANN is not making technical decisions but instead making policy choices on the basis a wholly inadequate an unrepresentative structure.
 

ICANN has never claimed that the technical stability of the DNS would in any way be threatened by the introduction of a very large number of new gTLDs. Indeed, it could not easily make this claim, since all the technical evidence is to the contrary. Rather, the dangers that ICANN seems concerned about are social - potential consumer confusion, and a potential 'land rush' mentality due to the enormous pent-up demand. (In my opinion, however, ICANN has selected a policy that maximizes the risk of a 'land rush'. Panic buying happens when consumers fear a shortage. Here, ICANN is proposing the creation of a very small number of gTLDs, with no assurances as to when if ever the next batch will be created. This gets it exactly backwards: the way to avoid a land rush would be to have a very predictable path for new gTLDs so that everyone understands that there's no need to panic since plenty of names will always be available.)
 

I am not an expert on Internet engineering. However, my understanding is that while experts do not agree on precisely how many gTLDs could be created without adverse consequences to DNS response time, there appears to be a technical consensus that we are nowhere near even the lowest possible limit. ICANN At-Large Director Karl Auerbach, himself a technical expert, has suggested that the smallest technically-mandated upper level for the number of gTLDs might be as high as a million. ⁽¹⁾ Persons with long experience in DNS matters, including BIND author Paul Vixie, apparently agree. ⁽²⁾ Others have performed tests loading the entire .com file as if it were a root file, and found that it works. In principle, this is not surprising, as there is no technical difference between the root file containing the information about TLDs and a second-level domain file. Given that there are currently about sixteen million registrations in .com, if this argument is right, then the maximum number of TLDs may be very high. ⁽³⁾ Some experts worry, however, that a very large number of new TLDs, such as a million, might affect DNS response time. ⁽⁴⁾ If so, that still means that with fewer than 300 TLDs in operation today (gTLDs + ccTLDs), we can afford to create tens of thousands, and probably hundreds of thousands, more.
 

It is an article of faith among Internet entrepreneurs that possession of a good domain name is a necessity for an Internet startup. Many traditional firms also consider the acquisition of a memorable or short domain name to be of strategic importance. Recently, for Internet startups, possession of a "good" name was seen as a major asset - reputedly enough in some cases to secure venture financing.
 

For some time now, however, it has also been an article of faith in the Internet community that "all the good names are taken" Recently it has seemed as if simply all the names that were a single word were taken. This apparent shortage, especially in .com, has driven firms seeking catchy names into the aftermarket. There does appear to be a reasonably large stock of names in the existing gTLDs being held by domain name brokers for resale in the aftermarket. Prices are very variable. Although few firms paid millions of dollars like the purchasers of business.com, and loans.com, it appears that at least until the .com bubble burst, the shortage of attractive names in .com , and the resulting need to purchase them at high markups in the aftermarket created what amounted to a substantial "startup tax" on new businesses.
 

ICANN justifies its very tentative initial foray into gTLD creation as a "proof of concept" but it has not disclosed the concept that is believes it is trying to prove, nor described how one tells if the test is successful, nor even when one might expect ICANN to do the evaluation. The "concept" cannot be gTLD creation itself: There is no rocket science to the mechanics of creating a new gTLD. From a technical perspective, creating a new gTLD is exactly like creating a new ccTLD, and creating new ccTLDs is quite routine. Indeed, .ps, a TLD for Palestine, was created less than a year ago with no noticeable effect on the Internet at all. ⁽⁵⁾
 

In fairness, ICANN is not originally responsible for the gridlock in gTLD creation policy, which in fact long predates it. Indeed the Department of Commerce - which currently has the power to create new gTLDs - called ICANN into being because it wanted to find a politically feasible way to create new TLDs in the face of difficult political obstacles, not least a belief in the intellectual property rights holders community that new TLDs might add to the risk of customer confusion and trademark dilution.
 

This political fear, more than any mythical technical consideration requiring a "test" or "proof of concept", explains why ICANN imposed a needlessly low limit on the number of new gTLDs it would recommend the Department of Commerce create in this first round, and why ICANN has as yet not been able to consider when if ever it will contemplate future rounds of gTLD recommendations. It does not explain, however, why ICANN why ICANN persists in falsely claiming consensus for its artificially low number of TLDs, nor why went about selecting its seven finalists in the manner it did. Indeed, as described below, ICANN's gTLD selection procedures were characterized by substantial failures.
 

Nevertheless, it might seem that despite any procedural irregularities, ICANN's recommendation that the Department of Commerce create a small number of new gTLDs can only be good for competition as it will increase supply and thus drive down prices. And indeed, supply will increase. Unfortunately, of the new gTLDs, only .biz and maybe .info are likely to be of attractive to the majority of startups and other Internet newcomers. Because there are only two such domains, and because there is no easily foreseeable date at which additional gTLDs might become available, there is a substantial risk of a speculative frenzy in which domain name brokers, cybersquatters, and amateur arbitragers all seek to register the catchy names that have not already been snapped up by trademark holders who took advantage of their pre-registration period. I am concerned that the faction which controls ICANN will use this very predictable speculative frenzy as 'evidence' that new gTLDs are a bad idea, or that the number must be kept down in the future.
 

The surest way to drive down and keep down the price of domain names, thus eliminating the "startup tax" and enhancing the ability of new firms to enter new markets and incidentally greatly reducing, perhaps even almost eliminating, cybersquatting, is to create healthy expectations. As soon as participants in the market understand that a steady supply of new domain names in attractive gTLDs will continue to become available on a predictable schedule, the bottom will fall out of the after-market, and the incentive (albeit not the opportunities) for cybersquatting will be greatly reduced, thus helping e-commerce by making attractive names available on reasonable terms to a much greater number, and wider variety, of persons and firms.
 

Selection of gTLDs
 

In ICANN's recent gTLD process, ICANN acted not as a standards or coordination body, but as if it were allocating scarce broadcast spectrum is some kind of comparative hearing process. ICANN created no standard. It 'coordinated' no projects with running code being deployed by outside parties. Rather, ICANN acted like a foundation grant committee, trying to pick 'winners.' In practice, ICANN's exercise of its gatekeeper committee role contributes to the artificial shortage of gTLDs. Worse, the selection processes ICANN employed were amateurish and arbitrary.
 

Although all applicants were charged the same non-refundable $50,000 fee, a sum that immediately skewed the process towards commercial uses and away from non-profit or experimental uses, it appears not all applicants received equal treatment. During the Los Angeles ICANN Board Meeting, it transpired that the staff had not subjected all the proposals to the same level of analysis. Thus, when Board members sought more detailed information about proposals that interested them, but which the staff had relegated to the second tier, that information sometimes did not exist, although it existed for the staff's preferred picks.
 

ICANN then attempted to hold a one-day comparative hearing between more than 40 applicants, each of whom had complex applications that referenced multiple possible gTLDs. During this process, each applicant was given three minutes to speak.
 

Both before and during the one-day Board meeting, both the staff and the Board seemed excessively concerned with avoiding risk. Although true competition in a fully competitive market requires that participants be allowed to fail if they deserve to do so, there are reasonable arguments as to why it makes sense to have a body like ICANN require potential registry operators to meet some minimum standard of technical competence. One can even make a case for requiring a showing of some financial resources, and for requiring the advance preparation of basic registry policy documents spelling out who will be allowed to register names and under what terms. Perhaps there are other neutral criteria that should also be required and assessed. This is a far cry from ICANN's apparent tendency to tend to prefer established institutions and big corporations, and to downplay the value of experience in running code. If in 1985 the Internet itself had been a proposal placed before a committee that behaved as ICANN did in 2000, the Internet would have been rejected as too risky. Risk aversion of this type is antithetical to entrepreneurship and competition.
 

Worst of all, ICANN applied its criteria arbitrarily, even making them up as it went along. The striking arbitrariness of the ICANN decision-making process is illustrated by the rejection of the ".union" proposal based on unfounded last-minute speculation by an ICANN board member that the international labor organizations proposing the gTLD were somehow undemocratic. (That this same Board member was at the time recused from the process only adds to the strangeness.) The procedures ICANN designed gave the applicants no opportunity to reply to unfounded accusations. ICANN then rejected ".iii" because someone on the Board was concerned that the name was difficult to pronounce, even though the ability to pronounce a proposed gTLD had never before been mentioned as a decision criterion. I am not in a position to vouch for the accuracy of each of the claims of error made by the firms that filed reconsideration requests after the Los Angeles meeting (available at http://www.icann.org/committees/reconsideration/index.html) but as a group these make for very sobering reading.
 

If ICANN were to limit itself to either standard making or technical coordination it would have approached its mission very differently from the arbitrary and amateurish procedures it used. It is critical to note that the relevant standards of comparison for ICANN's decision making are not the private sector. As a non-profit standards body contracting with the US government, ICANN should either be held to standards of openness, professionalism, and neutrality appropriate for standard-making or, if making political and social choices, be treated as a state actor and expected to act in conformity with fundamental norms of due process. Suggestions heard from some victorious gTLD applicants that ICANN's processes compare favorably with those used for procurement in the private sector are both erroneous and irrelevant. ICANN is not engaged in procurement. It is not "buying" anything. And ICANN paid almost no attention to the prices proposed by would-be registries.
 
 
 

II. Internal Organization
 

ICANN's go-very-slow policy on new gTLDs had no technical basis. Why then would ICANN adopt such a policy? The reason is that it is a product of an internal deliberative process that under-weighs the interests of the public at large and in so doing tends towards anti-competitive, or competitively weak, outcomes skewed by special interests.
 

The source of this predisposition is the distribution of decision-making authority on the ICANN Board, and in ICANN's subsidiary institutions, which have been manipulated to neuter the public voice, and the role of individuals, non-profits, and civil society groups. Originally, half of ICANN's governing Board would have been elected by at-large members of ICANN. Instead, ICANN has worked at every turn to prevent this.
 

In July, 1999, ICANN Chair Esther Dyson told the House Commerce Committee's Subcommittee on Oversight and Investigation that ICANN's "highest priority" was to elect nine at-large Board members, ⁽⁶⁾ exactly as ICANN had committed to do as an original condition of being approved by the Department of Commerce. Instead, ICANN reneged on its commitment to the United States government, and to the public, that half its Board would be elected by an at-large membership. Thus, today:
 

• Instead of half (nine) of the Board members being elected at large, as promised to NTIA and to Congress, ICANN amended its by-laws to allow only five members to be elected at large;
• Instead of all the self-appointed nine original directors leaving office as they promised Congress and the public they would do, four remain in office;
• Instead of allowing the five elected at large members to participate in the selection of the new gTLDs, ICANN amended its by-laws to seat them at the close of a meeting, instead of at the start (the process used for all previous new directors). Then ICANN rushed its processes so that it could make the final decisions minutes before the new directors took office.
• In a move that risks further neutering the five elected at-large members, ICANN announced that their jobs would all be abolished at the end of their two-year terms, unless a majority of the full Board voted (after a "clean sheet study") to re-establish elected at-large Board seats. [Note that under the current by-laws, the un-elected directors apparently get to keep their jobs indefinitely.]
• The internal institutions that ICANN created to take the lead in domain name policy - the seven constituencies in the "Domain Name Supporting Organization" (DNSO) - were designed from the start to exclude individuals from membership. The very engineers who built the Internet are not represented in their personal capacities - only if their employers choose to send them.
• All non-commercial groups, including all universities, all consumer groups, all political groups throughout the world are shoehorned a single DNSO constituency. They are, in the main, ineligible for full voting membership of any of the other six constituencies.; Meanwhile, many businesses such as Internet first-movers and others who have an interest in reducing on-line competition for established firms are eligible to be in two, three, or even four of the seven constituencies, thus allowing them multiple votes-and a certain majority.

 

ICANN is a highly complex organization (see attached charts, prepared by Tony Rutkowski). It is simply impossible for anyone to keep track of what is happening in all the different pieces,

except an organization capable of deploying a fleet of lawyers. Similarly, because ICANN sees its mission as global, it meets four times a year on four different continents. Next month's meeting, for example, is in Australia. The result of this laudable attempt at internationalization is that only interests wealthy enough to attend all these meetings - with several representatives - can achieve the continuity of participation required to influence ICANN's decisions in any sort of a consistent manner. The result tends to be a 'consensus' of those with the necessary expense accounts.
 
 
 
 
 

III. External Checks on ICANN
 

I do not deny that one can identify potentially serious social issues that might be caused as side effects of the creation of new gTLDs. I do submit that ICANN has no competence to deal with them, and that its actions have to date in creating special domain name registration rights for trademark holders, well in excess of the rights granted to them by Congress, have been anti-competitive, unfair, and counterproductive.
 

ICANN's mandate and its competence is, at most, for technical matters. Social policy issues such as the intellectual property consequences of new gTLDs, the number of days a person should have to respond to an arbitration over a domain name, or issues of content management, should not be decided by engineers or by the people who happen to have seized control of ICANN. Rather, they should be decided via the means we traditionally use for making social policy choices - markets and representative democracy.
 

Since ICANN's decisions as to its gTLD recommendations were not based on purely technical criteria, as a formal matter ICANN is making social policy choices, not just acting as a standards body. It is therefore right that ICANN's decisions are subject to external checks. Indeed, as I argue in my article Wrong Turn in Cyberspace: Using ICANN to Route Around the APA and the Constitution, 50 Duke L.J. 17 (2000), available onlinehttp://www.law.miami.edu/~froomkin/articles/icann.pdf , as a matter of law ICANN as currently constituted amounts to a state actor, and thus is subject to the same Due Process constraints as apply to any federal agency. Accordingly, its arbitrary and capricious decisions violate both the APA and the Due Process Clause of the Constitution.
 

ICANN and the U.S. Department of Commerce dispute this characterization. They prefer to rely on form over reality, and insist that ICANN is legally private despite the fact that ICANN derives all of its authority and revenue from Commerce's loan to ICANN of authority over the root. It follows, however, that if this characterization of ICANN as a purely private body is correct, then there are strict limits on the extent to which the Department of Commerce can implement ICANN's recommendations without violating the Administrative Procedures Act, or the Constitution's Due Process clause.
 

Once ICANN makes its formal recommendations, the Department of Commerce will have to decide how to proceed. Rubber-stamping of ICANN's decisions by the Department of Commerce would amount to adopting ICANN's arbitrary and capricious choices, since the U.S. government would essentially endorse both ICANN's practices and its conclusions.
 

The Department of Commerce has maintained that its relations with ICANN are not subject to the APA, or indeed to any legal constraint other than those relating to relations with a government contractor and/or a participant in a cooperative research agreement. But whatever the legal arguments, when contemplating decisions which will shape the very nature of the Internet naming system, Commerce should proceed with deliberation, and act only on the basis of reliable information. The need for reliable information, proper public participation, and transparent and accountable decision-making is even stronger when Commerce contemplates making the sort of social policy choices - as opposed to mere technical standard-setting - embodied in creating new gTLDs and imposing conditions on their use. Basic requirements of fairness, due process, and the need to make reasonable decisions counsel in favor of notice, public access, the making of an official record, and deliberation.
 

There is no question but that if a federal agency had acted as the ICANN Board did, its decisions would not satisfy even cursory judicial review. In the circumstances, therefore, it would be unreasonable and a denial of due process for Commerce to rely on the outcome of such a flawed process without conducting its own review.
 

ICANN faces a choice: On one path it becomes a true standards body, or a true technical coordination body, and leaves the social policy choices to those - like Congress - who have the legitimacy to make them. On the other path, the one it currently seems to be following, it is a state actor. In that case, its actions to date have been far too arbitrary to survive judicial review.

1. Posting of Karl Auerbach, karl@CaveBear.com, http://www.dnso.org/wgroups/wg-c/Arc01/msg00195.html .

2. E-mail from Paul Vixie, BIND 8 Primary Author, to Eric Brunner (Dec. 15, 1999) ("A million names under '.' isn't fundamentally harder to write code or operate computers for than are a million names under 'COM.'"), http://www.dnso.org/wgroups/wg-c/Arc01/msg00203.html .

3. See Quickstats, at http://www.dotcom.com/facts/quickstats.html (reporting twenty million registrations, of which 80% are in .com).

4. See, e.g., E-mail from Paul V. Mockapetris, BIND Author, to Paul Vixie, BIND 8 Primary Author, & Eric Brunner (Dec. 15, 1999) (querying whether one million new TLDs would impose performance costs on DNS), http://www.dnso.org/wgroups/wg-c/Arc01/msg00202.html .

5. See IANA Report on Request for Delegation of the .ps Top-Level Domain, at http://www.icann.org/general/ps-report-22mar00.htm (Mar. 22, 2000).

6. Testimony of Esther Dyson, Chair, ICANN, before the House Commerce Committee, Subcommittee on Oversight and Investigations, July 22, 1999, http://www.icann.org/dyson-testimony-22july99.htm .