We considered proposed economic structures of generic Top-Level Domain (gTLD) name registries. These range from the for-profit, market-managed model, to the constrained cost-recovery model. We develop a model for selecting among these, beginning with widely-accepted assumptions in the Internet (e.g., use market mechanisms whenever practicable) and criteria for choosing economic structure (stability, economy, innovation, and flexibility). For-profit registries are found to be not viable due to abuses which follow from the monopoly each holds over their respective domain names. Not-for-profit registries, without additional constraints, are also rejected because they are similarly able to abuse their monopoly position. Recommendation is made for not-for-profit registries, with the additional requirements of cost-recovery, audit and public oversight, open access by any registrar, and no claim to property rights in gTLDs. Such constraints must be applied uniformly to all registries by the new IANA.

0. Introduction

One of the most important issues concerning the central coordinating functions of the global Internet is the economic structure of gTLD names registries. This question is usually characterized as: should registries be for-profit or not-for-profit? Although establishing not-for-profit registries enjoys significant support, including positive comments by the US Government, support by comments to the US Government "Green Paper", and the consensus of the European Consultative Meeting on Internet Governance (http://www.ispo.cec.be/eif/dns/conclusions.html), discussion continues. This paper describes the significant considerations surrounding this question, suggests perspectives from which the new IANA institution should consider this issue, and recommends how the new IANA should proceed with the selection and further development of registry structure.

1. Working Assumptions

In Principle, Private, Competitive, For-Profit Structures are Best

As is commonly accepted in the Internet community, this paper takes as given the general principle that market mechanisms are the best means for managing Internet resources, where practical. Arguments to this end have been made elsewhere and there is no need to repeat them at length at this time. In brief, the advantages of private, for-profit entities include:

- driving down costs,

- rewarding and stimulating innovation, and

- simpler and less costly administration for the new IANA.

Some Structures Cannot Be Managed Strictly by the Market

In spite of the desirability and popularity of open, competitive market mechanisms, it is not possible for all Internet resources to be managed by the market. Some resources must be administered centrally. Just as the broadcast of RF signals on the same frequency in the same location will result in interference, there is an analogous necessity under the current design for the assignment of unique IP addresses and unique domain names to ensure that the Internet can function.

It is essential that the structure of the Domain-Name System (DNS) and the allocation of blocks of IP numbers be coordinated,. The requirement for some central administration is long established in the Internet and in other communication infrastructures. It is also recognized as the principle responsibility of the new IANA, which is endorsed by the US Government in the White Paper.

Each Top-Level Domain Name Will Be Allocated to a Single Registry

Ideally, each gTLD would be allocated across all, or at least multiple, registries. However, this is not presently possible due to a limitation in the state of the art for distributed computer systems. Were it possible for top-level domains to be multiply allocated, market mechanisms would probably be viable and registries could be competitive for-profit businesses. As an aside, there is no significant technical constraint on the number of gTLDs that could be managed by a single registry.

The Internet Continues to Evolve and Some Aspects are Not Yet Fully Understood

The nature of the Internet is still being revealed. The characteristics of Internet property (for example, with respect to domain names and IP numbers), associated rights, hoped-for new technologies (such as directory services), and Internet law, are far from understood. Similarly, the appropriate dividing line between free-market-manageable Internet resources (e.g., assets traditionally managed by ISPs) and common resources which require common administration (e.g., the allocation of IP numbers) are still being developed.

Sale of Second-Level Domain Names Will be Competitive

We take it as given that Registrars of Second-Level (and probably deeper levels as well) Domain Names can and will be competing enterprises which will have the option of being for-profit. Market mechanisms are fully capable of managing this part of the Internet's processes. Further details of a structure and relations among naming-related entities may be found in the "POC Open Letter to the Internet Assigned Numbers Authority" dated 12 June 1998, from the Policy Oversight Committee for Generic Top-Level Domain Names (http://www.gtld-mou.org/docs/ianaopenletter.htm). Under that model, registries will be owned and operated by registrars.

Our investigation leads us to ask where the registries should be located on the range from for-profit to not-for-profit. Not-for-profit structures include the IANA, the number registries, and the DNS root servers. For-profit structures include the registrars and ISPs. Hence, one can view the question which is the subject of this paper as: "to what extent do registries share requirements with the New IANA, and to what extent do they share characteristics and operating requirements with registrars?" Once this question is investigated, it is necessary to ask if either approach is sufficient, or if additional considerations must be addressed.

In Time there will Be Many Generic Top-Level Domain Names

Over the next few months there will be an increase in the number of generic Top-Level Domain names (gTLDs). More gTLDs are required to satisfy two requirements: a need for a larger name space to permit many different domain names with the same SLD (e.g., 'United'), and a need to satisfy the very significant demand for SLDs within the gTLD name space (now numbered in the millions of SLDs).

2. Criteria for Choosing For- or Not-For-Profit Structure

Using the above assumptions as a starting point, the decision to choose for-profit or not-for-profit registries should be consistent with and find an appropriate balance among, the following criteria.

Maintain and Enhance Stability

There are two types of stability: the actual technical and operational reliability of the network and its facilities, and the extent to which users believe that they can rely upon them. Both must be preserved and enhanced. Whatever structures are chosen must be well considered in terms of both the near- and long-term. Where possible, consideration should be given to the question of how structures are likely to evolve in the future so that stability can be maintained during this later evolution.

Maintain the Internet's Cost Economy

One of the great values of the Internet is its profoundly economical nature. Users have found that Internet services often permit an order of magnitude reduction in costs for similar non-net-based functions (e.g., E-mail vs postal mail). This differential creates a real and serious threat that the Internet could become a vehicle for the extraction of excessive 'rents' by many niche monopolies rather than a driver of new cost savings. Artificially high prices will not only reduce the economic efficiency of Internet use, they will also distort the way that it is used: users will migrate to lower-cost services which may be technically inferior, and threaten Internet infrastructure.

Encourage Innovation and Evolution

The Internet is in its infancy and there are many new technical developments on the way. The new IANA should ensure that the Internet is not frozen in its present state, and that innovations can emerge and provide new economic and social opportunities.

Preserve Options

The Internet will surely provide many future surprises, and the new IANA must be capable of flexible response to the challenges and opportunities that will present themselves. For this reason, it is imperative that the new IANA avoid structures now that are predestined to limit flexible response in the future. With the benefit of hindsight, it is now clear that enabling for-profit, monopolistic control of registration in the existing gTLDs was a mistake and it should not be repeated. Where there is a choice between structures that are fixed or flexible, there should be a bias toward the flexible. Similarly, the new IANA should implement structures in stages, in order to be able to investigate the various problems as they occur, and thus have the best chance of realizing solutions. In other words, the new IANA should initially act conservatively, so as to minimize later need to reverse decisions. This is particularly important for decisions that are known in advance to be essentially irreversible.

3. Problems Associated with For-Profit Registries

One of the principal questions of the economic structure of registries revolves largely upon whether they are monopolies. That question in turn depends significantly upon the nature of "switching costs" in domain names.

A registry controls name assignments for a particular TLD. There is no competition between registries within a single TLD, because current technology does not permit it. Hence, if a user is sufficiently unhappy with the administration of the TLD's registry, the user has no choice but to change to another TLD. In other words, the user must change its domain name.

Switching costs are those costs associated with a user's or an organization's move from a second-level domain name (SLD) in one TLD to the same or a different SLD in another TLD. If switching costs are negligible, then the registry will have no special hold over the SLD holder. (Note: The concept of "switching costs" comes from economics, and is unrelated to the technology of "packet switching.")

Scenario: Switching Costs at PackCo Incorporated

In order to investigate the problem of switching costs, we construct a scenario regarding a fictitious packaged goods company, "PackCo Inc". PackCo has the SLD "packco.goods", which it has acquired from a registry, RegInc., which is a for-profit company. It acts as both a registrar and the sole registry for the top-level domain name ".goods", as well as for a number of other top-level domains. For an initial one or two year period, PackCo is charged a fair price for its SLD - a price which is reasonably close to RegInc's registration-related costs. RegInc has set the initial price at this level in order to attract many new customers.

After a number of years, RegInc decides that PackCo is obtaining exceptional value from its very popular domain name and that RegInc can succeed at extracting a larger share of that value by increasing the price by a factor of 10. PackCo is not happy with this price increase , but is unwilling to make a change: the higher charge levied by RegInc is much smaller than the cost of changing to another domain name. Furthermore, there is another company, a vendor of backpacks, which holds the domain names "packco.sport" and "packco.recre", and yet another company which is an international shipping company that holds "packco.ship". Although PackCo is not really in the same business as any of these other companies, PackCo is concerned that changing gTLDs will result in name confusion by their customers. PackCo executives cannot guarantee that all of its customers will be informed about the name change - it has millions of anonymous customers. PackCo has an additional problem: it has more than one domain name; in fact, it has hundreds, each corresponding to the different packaged goods it sells, each in turn potentially subject to near-collisions with other companies' domain names. If PackCo were to change all of its domain names, it would represent a significant fraction of its annual profits. PackCo is thus forced to continue to use the original domain name even in the face of rising costs.

The problem is exacerbated by the conflict created between IANA - the ultimate authority over gTLDs - and a registry which claims intellectual property rights in its particular gTLD. If IANA is forced to consider conflicting intellectual property rights claims in its administration of the name space, it will face extremely costly, burdensome, and likely long-term litigation risks.

Problem: The Human Element - The Persistence of Memory

The Domain Name system was developed expressly to solve an inherent problem with IP numbers: obscurity. Even sophisticated early Internet users found it awkward and inefficient to maintain lists of IP numbers of the machines they wished to connect to. The domain name system was deliberately designed to be simple and intuitive by being readily identified by, and memorable to, users. It has been remarkably successful at this.

The memories of possibly millions of users are the first source of resistance to switching. Both an SLD and a gTLD together are the basis for the company's reference, and the string that is associated in the mind of the user. Thus, there will be confusion if the SLD is removed from one gTLD and assigned to another. At first blush, this may not seem to be a complex problem, but to a user trying to find a specific name, possibly intersecting with other similar or even competing businesses, it can be time-consuming and difficult. Proprietors could face the threat of significant business losses.

Changing a domain name is akin to changing a major brand name. It is very well-understood in the business community that changing a brand name can be exceptionally difficult, expensive, and result in long-lasting, even permanent damage to a company's market position. Users unable to change to another registry are said to be subject to "lock-in."

Problem: The Technical Element - The Persistence of Embedded Pointers

More significant than the persistence of human memory is the persistence of computer-based pointers. One well-known example is that of the Internet browser bookmark. Users have grown accustomed to the convenience of saving references to Web locations (Universal Resource Locators, or URLs) in their browsers. If users find that their references have begun to fail, they will need to spend time and effort at finding the new URLs for the sites and organizations that interest them. More and more such transactions are strictly machine-machine; there is no "human in the loop." In these cases, all manner of highly efficient transaction processing could be retarded while the new address is determined, if at all.

Pointers embedded in documents will become invalid if their gTLDs change. The reference earlier in this document would break in such a situation. Documents are on servers of various sizes and connectivity all over the world. How are they to be updated? Authors regularly move on to other projects and it is highly unlikely that they, being mortals, would continue to update or correct reference many years after the original publication date.

Embedded pointers in search engines are yet another point of concern. Search engines use automated tools to collect pointers. If a pointer becomes invalid due to a change in a gTLD, it is very difficult for the search engine to know to remove the invalidated reference. Is a reference invalid when it has been unavailable for a specific time? Servers and networks go down, sometimes for significant periods. As more and more pointers become incorrect, users will find search engines less and less useful due to the proliferation of invalid references.

It will be extremely difficult, perhaps impossible, to develop automated, software-based solutions to these sorts of problems, and such proposals are at best speculative. They do not exist today and it is entirely inappropriate to make current operations depend upon their uncertain development.

Problem: Monopolistic over the Long Term

It is clear that initially, there will be competition among domain names . Before a position has been taken in a given SLD/gTLD, a prospective name holder will have significant degrees of freedom in choosing among gTLDs. When additional gTLDs are made available, the preponderance of such customers will drive competitive behavior by registries: prices will be kept low, and services will be maintained at reasonable levels of quality. However, over time, as indicated by our scenario, once a name is allocated, domain name holders will tend to be locked in to their SLDs and their associated gTLDs. Correspondingly, over time, allocated domain names will generally increase in value. . This creates a real and potentially strong incentive for a for-profit registry operator to increase prices- they will have control over their customers.

It may be that for some time there will be relatively larger demand for registration of new domain names compared to the number already registered. In this case, there would be some pressure to keep long-term prices at a reasonable level. However, the number of new names to be registered will eventually diminish relative to those already installed, and the motivation to keep prices low will diminish significantly or disappear entirely.

Each registry thus holds a monopoly over the top-level domain names it administers.

Problem: Rising Prices Delinked from Costs

The appearance of monopoly registries will cause prices to rise if the registry organizations are for-profit. There may be periods during which prices will reasonably track costs, but this will not continue reliably. It is unreasonable to expect for-profit organizations not to migrate to value-based pricing (as seen in our scenario) and thus begin to charge based on switching cost, rather than based on the cost of provision. . A for-profit company must seek to maximize profit. The more control it has over its customers, the more profit margin it can extract. Value-based pricing will allow substantially higher profit margins than will cost-based. This is a clear violation of the criterion: "Maintain the Internet's Cost Economy."

High switching costs are not the only problem. There are additional problems associated with for-profit registries.

Problem: Inappropriate Property Claims

The above scenario demonstrates that a registry might be able to establish property rights over a domain name. To preclude this, the new IANA must constrain registry operators so that they will not be able to make property claims over TLDs, which must be kept as public resources. It is clear that allowing intellectual property rights in a gTLD held by a registry is an invitation to monopolistic abuse of economic power. In addition, a registry should not be permitted to use as its company name, the name of the gTLD that it operates.

Problem: Additional Monopolistic Practices

It is in the nature of monopolies that they are able to impose constraints upon users far beyond simple overcharging. They could, for instance, require that name holders or registrars sign up for a special level of "gold service", thus providing "improved and guaranteed" performance. Users and registrars could be motivated to step up to such new services as they discover that the old services and performance levels are simply not sufficient to keep up with the changing loads and requirements of the Internet. Registries could let contracts expire, and thereafter require name holders to enter into new, onerous agreements. Such techniques cannot be easily predicted, especially given the changing nature of the Internet (e.g., new protocols). Hence, initial guaranteed-price contracts, and detailed contract terms, will not reliably keep name holders from being subject to creative techniques used by monopoly registries to separate them from their money. Imposing a corresponding regulatory burden on the new IANA is equally problematic; the specific techniques are far from understood at this time, and even if some of them were, management of new and changing techniques would likely impose a very expensive, slow, and time-consuming cat-and-mouse game on the IANA. Enforcement is equally problematic, and especially so due to many possible jurisdictions and legal systems. Legal and other costs incurred by monopoly registries in this area would surely be passed on to their customers. Registries would surely argue that these are a regular part of their operating expenses.

Problem: A Proliferation of Claimants "Rights" in gTLDs

Even before the creation of additional gTLDs, there are numerous claimants to "rights" in particular gTLDs, some of whom are operating so-called DNS "root servers" that are recognized by a tiny minority of the servers comprising the Internet. There are claimants to the entire alphabet, there are claimants to every word in the English dictionary, there are claimants to sexually oriented TLDs - the list goes on and on. If the rights of any single one of these claimants is recognized, what is to prevent other claimants with plausible intellectual property rights in the same or similar names from filing law suits to press their claims? What is to prevent an enormous number of claimants from making claims that they are each entitled to operate one or more for-profit registries? It would of course be possible to establish a global lottery offering a pre-determined (by IANA) number of gTLDs to the universe of would-be registries who meet some minimum standards This is surely the worst possible way to provide for a registry system that is stable, efficient and operating in the best interests of the Internet as a whole.

Problem: Irreversibility

Should the new IANA choose to permit for-profit registries, it will be taking a substantial risk. If for-profit registries are permitted, and are shown to have a negative impact on the Internet or cost economics, it will be quite difficult to convert them into not-for-profit registries. On the other hand, it is far easier to convert not-for-profit registries into for-profit. Our criterion, "Preserve Options", is best supported by choosing not-for-profit registries.

Past Experience

Past experience in the Internet with Networks Solutions Incorporated (NSI) demonstrates many of the above problems, including registration fees far beyond cost-recovery (as evidenced by their unusually high profitability) and inappropriate property claims over ".com". This proof by demonstration shows that similar problems could readily occur in other TLDs, if similarly structured.

The above analysis demonstrates that domain name registries are inherently monopolistic. For-profit market mechanisms are thus inappropriate; they will not work. Past discussion that has referred to one of the choices of registry structure as being 'competitive' is inaccurate. They are in fact anti-competitive. Competing domain name registries are not possible until the state of the art in distributed systems changes.

If for-profit structure will not work, will not-for-profit structure? We consider that question in the next section.

4. Not-For-Profit gTLD Registries

Not-for-profit registries are superior to for-profit registries simply because they are not necessarily driven to make profits. However, being relieved from this requirement does not mean that they are in and of themselves an appropriate structure. There are five potential problems with not-for-profit registries. We analyze each and consider if additional mechanisms can make them appropriate for gTLD domain name registries.

Registries Must Be Cost-Recovering with Public Review

A cost-recovery model is not on its own sufficient to preclude abuse. Not-for-profit organizations may provide unusually high staff salaries or perks, or accumulate assets unrelated to the operation of a registry. Other abuses could occur by maintaining a large bank account in common with an owning registrar - which in turn could act as a hidden source of profits for the registrar. It will be necessary to ensure that no such "hidden profits" are permitted by requiring that registries operate in a manner consistent with other, similar not-for-profit organizations, and that they consist of a single financial "envelope". This can best be accomplished by having public review of registry financial practices and management, with the assistance of an internationally recognized public accounting firm.

Registries Must Allow Registration or Reregistration by Any Registrar

We assume at the beginning of this paper that registrars will be competitive. In order promote effective competition among registrars, each will require open access to any registry both for initial registration and for re-registration. Similarly, to avoid the problem of lock-in between users and registrars, registrants must be able to freely move their domain names to any other registrar (domain/registrar portability).

Domain Names Must Be Public Assets

As demonstrated above, for-profit registries could claim inappropriate property rights in gTLDs; not-for-profit registries could possibly make similar claims. In order to preclude this, there must be no property rights in any top-level domain. This must include a requirement that a registry not be permitted to use any gTLD as part of a brand or company name (e.g., no "ORG.COM Registry Incorporated").

Registries Must Be Motivated to Innovate

Under our model the registries will be financially distinct entities without a profit motive. Even though the existence of multiple registries will encourage a basic tone of competition among them, they could conceivably lack motivation to foster innovation and could have difficulty raising capital. However, these problems are precluded if the owners of the registries are the registrars (as cited in the "Open Letter"), which will generally be businesses operating for-profit. This will enable them to raise capital and will provide the motivation to drive innovation within the registries. Whether registries may be owned by organizations other than registries is a item for consideration by the IANA and depends in part upon satisfying this criterion.

Constraints on Registries Must Be Applied Uniformly

The new IANA, as a corporation subject to the laws of the jurisdiction in which it is located, must uniformly apply the above constraints on registries to avoid violation of anti-trust laws.

This concludes our development of an appropriate economic structure for registries: not-for-profit organizations with the additional constraints of cost recovery with public review, open registration, public-asset domain names, motivation for innovations, and uniform application of constraints by the new IANA.

5. Criteria Reviewed and Summary

The implementation of a not-for-profit, cost recovery registry system will meet the four criteria set forth above:

Maintain Internet Cost Economy

Not-for-profit registry prices will track actual costs. Overhead costs, which will be transferred ultimately to registrants, will be kept at industry-standard levels through the auditing and public review process.

Maintain Stability

Not-for-profit registries will be long-standing stable structures. They should not subject their users to concerns about possible monopoly pricing and registry long-term business stability.

Encourage Innovation and Evolution

Use of not-for profit registries will be motivated to innovate under at least one model, as described above. Additional models are open to specification by the new IANA.

Flexible Response by the New IANA

The new IANA will be able to observe the behavior of registries, registrars, users, and associated technical developments. If it is desired to make changes to the system, it will be easier to go from not-for-profit to for-profit than the other way around.

We have observed that for-profit-Registries cannot appropriately limit the negative consequences of the monopoly control over of gTLDs. We have demonstrated that not-for-profit registries, with additional control mechanisms (public oversight, etc.), are an effective way do so. Economy, stability, and innovation will be well-supported by not-for-profit registries. Finally, as experience is gained, and new developments occur, the new IANA will have the option of shedding part of its regulatory burden by moving part or all of the registry process to open, competitive market mechanisms.