Report on TLD Applications: Application of the August 15 Criteria to Each Category or Group (9 November 2000)

This group of nine applications involves proposals for new TLDs that would provide a type of use or service that is currently not generally available in the DNS. For analysis, they have been grouped as follows:

As might be expected with proposals for new services, the proposals in each of these groups differ significantly from one another, so that the groupings are often more useful for structuring discussion than for comparisons among the applications in each group.

a. Telephony-Related Group

The four applications in the telephony-related group of the new services TLD category are set forth in the following table:

Technical Summary of Proposals

Each of these proposals relates, at least in part, either directly or indirectly to Internet telephony. However, these proposals exhibit a variety of sub-domain naming schemes and stated objectives. To summarize:

Group One Registry, Inc.

This applicant proposes to form SLDs from strings of numbers. Among the objectives are uniquely identifying Internet connected devices, people and perhaps telephony devices. Certain SLD strings would be reserved or restricted. For example, all strings of eight digits or fewer are reserved. Claimed advantages of the numeric-only SLD numbering scheme include uniqueness of identifiers, reduction of "land rush" problem and universality with respect to languages.

The proposal does not purport to be targeted at registration of telephone numbers as SLDs. Nonetheless, its use of number strings and its potential for application to telephony services indicates it is natural to evaluate the application in this category. Additionally, Group One Registry expects that over 80% of its registrations will be digital mobile phones.

As potentially applied to telephone-number registrations, all country-code telephone-number strings (which are included in the eight-digits-or-fewer group) are restricted to use by the telephone authority in the corresponding country. Any SLD string longer than eight digits, on the other hand, would be unrestricted and could correspond to a full telephone number. ICANN, for example, could register its telephone number as 13108239358.one, but not as 3108239358.1.one. The target of the address translation is unrestricted, but could include telephones.

Number.tel, LLC.

Much like the Group One Registry proposal, this proposal would assign SLDs as simple number strings that would correspond to PSTN (Public Switched Telephone Network) telephone numbers. There would be no sub-domains. According to the proposal the point is not to reach a telephone (or voice enabled device), but rather to provide a way to link a phone number with an arbitrary IP address. One stated purpose of this linkage is to allow someone to locate a person's home page using the person's phone number. Other claimed advantages include unique assignment of names (because they relate to a unique number string) and universal, language independent representation (every language can express numbers). The applicant would require that a registrant currently have that PSTN number assigned to them. Thus, ICANN's web site could be accessed through http://13108239358.tel.

Pulver/Peek/Marschel.

This application proposes to identify voice-enabled Internet devices using the DNS as a lookup facility. The proposed naming scheme is similar to a reverse address tree, so that the following domain name would be associated with +1 310 823-9358: <8.5.3.9.3.2.8.0.1.3.1.tel>. The approach offered mimics the <e164.arpa> database, an existing number mapping system intended to record within the DNS information about the world-wide PSTN numbering plan. The translation would lead to a voice-enabled Internet device, but not necessarily a person's actual PSTN telephone (although presumably that might occur in the future). No other TLD sub-domain schemes would be allowed and there would be no reservations of blocks of numbers. The applicant would require registrant to prove that it has "day-to-day" control over the PSTN number used to form the sub-domain string.

Telnic Limited.

Like the Pulver/Peek/Marschel application, this application proposes to identify voice-enabled Internet devices, but does so directly from character strings. Thus, a sub-domain name under this proposal might take the form <john.smith.london.uk.tel>. This would correspond to an Internet voice device of the John Smith who lives in London. Four character or shorter domains would be reserved at the SLD to establish country codes. Similarly, the registry would reserve location names (e.g. cities) to itself. In a sense the Telnic Limited proposal would follow current domain naming schemes (i.e. character- and word-based) except that the target IP address would be a voice device. Since the sub-domain is non-numeric there is no underlying requirement to identify an "owned" PSTN number. Because Telnic Limited proposed to tightly regulate the SLD and 3rdLD sub-domain names "land rush" problems will probably be manageable.

Threshold Review

Three of the four applications seek delegation of the same TLD, and all four request TLDs for related services, so a threshold review was conducted to determine the relative strength of the applications in terms of their presentation of technical and financial plans for stable and reliable long-term provision of the proposed services.

If viewed simply as conventional registries, the telephony-related TLDs are of intermediate size and difficulty to operate. The applicants' projections for names under management over a four-year period (50% level of confidence) are as follows.

In the initial years these registries would be of intermediate size. The evaluation team concludes that Group One Registry, Number.tel, and Pulver/Peek/Marschel clearly have sufficient technical resources to operate this size of registry. These applications were significantly better designed and provisioned than the Telnic Limited application to provide long-term stable services of the type proposed. The Group One Registry, Number.tel and Pulver/Peek/Marschel applications showed a good understanding of core issues such as availability, performance estimation, sizing, and security.

Telnic Limited's application showed a design, resources, and experience that are somewhat less robust. The technical team concluded that Telnic Limited's application was roughly adequate to the stated task, although somewhat under-specified in some areas, especially with respect to redundancy. Similarly, the business/financial team concluded that Telnic Limited's application was not as strong as Peek/Pulver/Marschel and Group One Registry.

Based on information provided in the applications the four proposals would provide the following levels of service (in this table, "DNS Constellation" refers to the number of geographically distinct locations where the proposal promises to support one or more DNS servers):

All applicants show a good level of availability, although the Telnic Limited proposal did not make use of geographical dispersion for the primary site services. Assessing other aspects of operation for the applicants is problematic. True telephony services may have a different usage profile than current domain name applications, especially if message routing and mobility are factored in. One possibility is that caching of name translations will not work as well when the traffic includes frequent accesses between personal voice enabled devices. Another issue is that mobility may require much more frequent SRS and DNS updates. Very high performance DNS servers with instant updating may be required to support such services. All the applications are vague or silent on how mobility will be handled.

Furthermore, all four applications fail to address the special requirements of telephony-coordinated applications. These requirements make operations under any of these proposals more demanding than conventional TLDs because, to at least some extent, they may require coordination with the existing, highly regulated, PSTN numbering system. The Pulver/Peek/Marschel application specifically envisions connections that might terminate within the PSTN. The other three applications do not explicitly address this possibility, but the flexibility to terminate Internet-originated calls on the PSTN will likely be an essential aspect of all voice over IP (VoIP) activities, such as explicitly mentioned in the Pulver/Peek/Marschel and Telnic applications.

The ITU Letter

In evaluating this group of applications, it is important to note that they may be affected by correspondence dated 1 November 2000 from the International Telecommunications Union (ITU). In response to an inquiry regarding its position on the .tel applications, the ITU (which coordinates a large number of national telephone number assignment bodies and regulatory agencies) stated (in part):

"The issues surrounding the intersection between telephone numbering and the DNS are very complex and currently under discussion between experts in the IETF and ITU as well as with governments responsible for numbering plans. As I am sure you are aware, the E.164 international public telecommunication numbering plan is a politically significant numbering resource with direct implications of national sovereignty. It is subject to a multitude of national approaches, regulatory provisions, and, in some cases, multilateral treaty provisions. Considering this, governments should be given the opportunity to fully reflect upon how their particular numbering resource responsibilities relate to DNS-based telephony resources."

"In this regard, the ITU is working with the IETF to progress a careful exploration of these complicated issues in the context of its joint work concerning the ENUM protocol. As there are still considerable areas of coordination work needed at this time, until there is an opportunity to further explore the issues within the context of joint work underway and with national governments; it is the view of ITU that it would be premature for ICANN to grant any E.164-related TLD application as this may jeopardize these cooperative activities or prejudice future DNS IP Telephony addressing requirements."

Thus, the ITU has expressed concerns that TLDs should not be established that would either impinge on the current numbering plan or foreclose possible future telephone-related assignments.

The effect of the ITU's concerns might be considered as follows. Telnic Limited does not make use of phone numbers in the sub-domain name, but instead uses names to designate the intended destination of VoIP calls. Although it is likely that Telnic's system will be required to support termination of Internet-originated calls on the PSTN, and should do so in a manner that is consistent with PSTN number assignments, the Telnic application appears to have the least impact on PSTN numbering. (It should be noted that Telnic's application does not explicitly renounce the future use of numbers as SLD strings.) The Number.tel application makes allowance for use of SLD names for telephone routing purposes, although the implication is that they are simply numbers and are not competing with the international number plan. This implication, however, runs strongly counter to the assumption, likely to be broadly held by the technical community and the public alike, that calls placed to a phone number from either the Internet or the PSTN should terminate at the same place. Group One Registry explicitly reserves part of the SLD space for possible regulation by telephony authorities. Of the four applications for .tel, the Pulver/Peek/Marschel is the only one to suggest directly that the assigned sub-domain string might represent a physical PSTN phone. Section E27 of the Pulver/Peek application explicitly addresses the potential for competition between the applicant and the regulatory agencies represented by the ITU. Basically, the applicant believes that "legacy" telephone numbers and functions will eventually migrate to the Internet domain name scheme.

Comparative Analysis

After the threshold review of the telephony-related applications, the evaluation team examined how each of the August 15 Criteria should be applied to the remaining application in this group in light of the diversity in purpose and targeted markets reflected in the categorization:

Maintaining Internet stability, meeting unmet needs, proof of concept and enhancement of the utility of the DNS are particularly relevant to analysis of the telephony-related applications.

Stability

The applications in this group can be assessed against the stability factor in a manner much like applications in other categories and groups. In addition, however, to one degree or another the new services present additional technical requirements that must be met if they are to achieve a stable, authoritative, consistent, and reliable mapping between the domains registered within the .tel TLDs and the PSTN dialing plan.

The evaluation team concluded the Group One Registry, Number.tel and Peek/Pulver/Marschel applications provided a level of general stability (i.e. not considering the extra coordination requirements inherent in the new services implied by these applications) that is somewhat superior to the Telnic application.

When reviewing the stability implications of the coordination requirements, however, additional distinctions emerge. In the case of the Group One Registry, Number.tel, and Peek/Pulver/Marschel applications, numbers (which can be, and often will be, telephone numbers) are explicitly used and, if a consistent telephone numbering system is to be achieved between the Internet and the PSTN, technical mechanisms should be established to ensure that the DNS registrations are consistent with the PSTN databases, both at the time of registration and on an ongoing basis. Group One Registry recognizes the possibility of conflict with the PSTN number plan and imposes some restrictions to address this issue.

The Telnic proposal, because it uses words (or character strings) rather than numbers for SLD names, does not have the same type of coordination requirement. To the extent that the Telnic service would be used to direct calls to a domain name (such as <john.smith.london.uk.tel>) to the PSTN there is a coordination requirement, but providing that coordination is not inherent in the Telnic domain-naming system and can be viewed as the responsibility of other (unspecified) systems.

Another unusual technical requirement for these proposed services is the ability to manage dynamic mapping requirements in the DNS. Telephony applications are among the first substantially mobile Internet devices. It is not clear from the proposals how even simple issues such as call forwarding would be handled. If this were to be handled by the registry then SRS and DNS capacity requirements might be much higher than applicants anticipate. If it were to be handled at the registrar level then registrar availability and security requirements might be much higher than current models require.

Meeting Unmet Needs

The four proposals in this group address a variety perceived needs. The needs may be analyzed as follows:

Group One Registry

The applicant assumes that there is a large pool of Internet connected devices that need to be addressed uniquely. This unique addressing might also extend to individuals or companies through the use of their phone numbers as a kind of serial number.

Number.tel.

The stated focus of this application is to allow DNS addressing of web sites based on telephone numbers. Thus, if one can remember a person's or company's telephone number, one can go to the corresponding web site simply by typing the number and ".tel" into a browser. Although there is some slight increase in convenience achieved by this procedure, it seems unlikely that telephone numbers would be broadly viewed as easy-to-remember identifiers and, to the extent they are, the same benefit could be more easily achieved by implementation of a search engine or other directory service including telephone numbers and conventional URLs. (The applicant's market assessment does not explicitly evaluate the potential demand for number strings used in the manner proposed.)

Pulver/Peek/Marschel

This application proposes to provide a reverse-lookup service to allow routing of VoIP and similar services based on the recipient's telephone number. While this is an important service, it is not clear why establishment of a TLD for this purpose enhances the service in any meaningful way. Reverse-lookup schemes are not intended for direct human entry, but rather to provide DNS entries that can be located by equipment designed for that purpose. For example, a caller might dial ICANN at +1 310 823-9358. The caller's terminal or other device would, under the Pulver/Peek/Marschel proposal, convert this key sequence to <8.5.3.9.3.2.8.0.1.3.1.tel> and do a DNS lookup of that name to determine how to handle establishment of communication with that destination. It is not apparent how this system is benefited by establishment of the .tel TLD, as opposed to use of a second- or lower-level domain. Since the DNS query is generated by the terminal or other device, there is no particular technical advantage to employing a TLD for this purpose. Indeed, two other telephony-related reverse-lookup services have already been initiated using the second-level domains <enumworld.com> and <e164.arpa>. Thus, it seems that the Pulver/Peek/Marschel proposal would not address unmet needs in any significant way. Instead, it may be viewed as an effort to have ICANN give an imprimatur to the proposal as the "official" reverse-lookup directory for telephone numbers.

Telnic Limited

The Telnic proposal, like the Pulver/Peek/Marschel proposal, is directed toward routing in the Internet-telephony field. It provides an additional feature, however, because it provides name-based addressing rather than requiring specification of the recipient's PSTN telephone number when the call is being established. This not only allows callers to avoid having to look up PSTN telephone numbers (since the underlying service would be configured to automatically route calls to <john.smith.london.uk.tel> to the appropriate location), but also allows VoIP destinations on the Internet to be established without the need for assignment of a PSTN telephone number. Thus, of the four proposals in this group, the evaluation team concludes that the Telnic proposal is strongest in meeting unmet needs.

Proof of Concept

Group One Registry

The applicant identifies proofs of concept related to its registry/registrar business model. It is not clear how proof of this concept would be relevant to anything beyond the Group One proposal itself. No technical or usage proofs of concept are suggested.

Number.tel

This applicant would determine if domain name based only on numbers could become a unique, internationally accepted means of identifying an IP address. They also believe that this scheme would reduce name conflicts. Telephone numbers are considered useful for these purposes because they are uniquely assigned and identifiable.

Pulver/Peek/Marschel

Proofs of concept identified by the applicant include explicit migration of the current telephone numbering scheme in to the domain name space. Also expected is that telephone numbers will be useful and unique identifiers for locating other objects associated with an entity, such as their LDAP directory.

Telnic Limited

This applicant believes that the character string based approach they propose will retain the user-friendly aspects of current domain addressing when used with voice devices. Additionally, the structured assignment approach may reduce name disputes and "land grabs".

Enhancement of Utility of the DNS

Enhancement of the utility of the DNS is another of the August 15 Criteria. Three applications request delegation of the .tel string; one requests delegation of .one. Although these strings are semantically far from existing TLD strings, so they are not likely to be confused with them, there are other aspects of these applications that may confuse the DNS hierarchy. The Number.tel application proposes to use the .tel strings for web sites, not necessarily for telephony-related uses, which is likely to be confusing to many users of the Internet and would not appear to enhance the utility of the DNS. To the extent the applications are being used to designate service types by their TLD (this is particularly relevant to Pulver/Peek/Marschel and Number.tel), they are likely to create confusion by designating particular services with a ".tel" at the end of the URL, when the established convention [RFC 2806] is to designate it at the beginning of the URL, in the form: <tel://icann.org>. Indeed, this is likely to result in confusion that will hamper consumer acceptance of VoIP designations using the existing, technically standardized, conventions.

Protection of the Rights of Others

The Group One, Number.tel, and Pulver/Peek/Marschel proposals use naming schemes that would appear to inherently reduce name disputes. Because each of these three proposals each define the available second-level names in terms of PSTN numbers, any dispute could be resolved based on a determination of the right to the PSTN number. These proposals, if accepted, however, would need to provide a mechanism for making the PSTN-number-entitlement determination.

The Group One Registry also reserves the part of the name space that conflicts with the E.164 number plan for the relevant country based telephone numbering authority. As noted above, however, this reservation would not preclude registration, independent of the telephone numbering authority, of digit strings representing telephone numbers within the authority's jurisdiction.

Telnic, on the other hand, uses a naming system that is somewhat more prone to naming disputes (though involving less risk of conflict with the PSTN numbering system). The proposal describes a scheme whereby certain SLD strings would be reserved. This includes all four or fewer character strings, the full names of regions, countries and some cities. Some offensive words would also be reserved. Under this proposal, disputes should be somewhat rarer than in the current TLD since all names will be localized (as in <john.smith.london.uk.tel>). Telnic proposes that disputes that do occur would be resolved through ICANN-approved dispute-resolution procedures. Trademark issues would be alleviated with warnings to registrants to carefully consider potential conflicts when applying for a character string. There are no automatic mechanisms for detecting conflicts.

Recommendations

Based on application of the August 15 Criteria, the evaluation team believes that none of the four proposals in the telephony-related group should be selected at this time. Each of the four proposals appears not to have adequately addressed requirements for stable, authoritative coordination with the PSTN numbering system, particularly when dynamic-routing considerations are taken into account. (Of the four, Group One, Number.tel, and Pulver/Peek/Marschel are of particular concern in this area.) In addition, the Group One Registry, Number.tel and Pulver/Peek/Marschel proposals would do little to address unmet needs. Moreover, if a TLD were established in which the service available at URLs was defined by the TLD rather than the prefix, this would likely increase confusion regarding URL naming conventions. Finally, the concerns raised and caution urged by the ITU counsel against establishing a telephony-related TLD until further study and consensus-building within the Internet and telephony technical communities.

b. Message Routing Group

The two applications assigned to the message routing group of the new services TLD category are set forth in the following table.

Technical Summary of Proposals

Dada SpA proposes a name space that would represent individuals, usually by their actual name. DNS translation would point back to the registrar where the final target IP address would be generated. The purpose is to base the final target on the individual's profile so that a message, for example, would be routed differently depending on the time of day, where the person was, or the type of Internet device they were currently using. The mechanism for supporting this activity is left to the registrars and is not detailed in the application.

The Nokia application applies to mobile Internet devices. SLD would be assigned to vendors of such devices (i.e. in a sense they are the registrars). Names at lower levels would be arbitrary. The Nokia application does not provide any discussion or examples of how this capability would be useful or interesting.

Threshold Review

The evaluation team performed an extensive threshold examination of these two proposals for technical and business/financial feasibility. As a result of this review, the evaluation team concluded that neither proposal merited more extensive review in connection with the current proof-of-concept program.

The technical team concluded that DADA SpA's application did not demonstrate realistic technical and operational plans. The applicant has not demonstrated that it has the relevant technical expertise and the proposed technical plan did not support the proposed business plan in the several areas, including under-specification of fault tolerance and security.

With respect to Nokia's application, the evaluation team concluded that the application was not complete and did not demonstrate soundness and feasibility from technical and business-process perspectives. Although Nokia and its registry operator, Sonera, have impressive capabilities in the communications field, the application by itself does not adequately describe the architecture of the proposed system and did not address all the topics required by the ICANN call for proposals. Instead, it sketched out general requirements for hardware and software systems without providing a detailed solution to the requirements. The business/financial aspects of Nokia's proposal were also sketchy: the application included a weak marketing plan and a weak assessment of estimated demand and the resources to meet demand.

c. Other Group

The three applications assigned to the "Other" group of the new services TLD category are set forth in the following table. Although grouped together, these applications provide for a variety of different services.

Applicant	Strings Requested	New Service
Monsoon Assets Limited (BVI) (dba dotYP, Inc.)	.find, .yp, ypa, .ypi	Yellow Pages
Novell, Inc.	.dir	LDAP Directory
SRI International	.geo	Georeferenced Information

Technical Summary of Proposals

The Monsoon Assets Limited proposal requests a new TLD that will be used to support a lookup service for businesses in a manner similar to a yellow pages phone book. By issuing a name under this TLD, the search service would be invoked and would return an IP address to the desired service. For example, <plumbing.Hollywood.yp> would return plumbers listed in business directories covering Hollywood.

Novell's application states that while the Internet has revolutionized the world, it has also created the problem of multiple e-mail addresses, phone numbers, web sites, and mailing addresses, making it difficult to find anyone or anything. Also, many different and incompatible methods exist to obtain basic information about people and things. Novell's solution to these problems is the creation of a .dir TLD. The TLD would, in Novell's view, become a rendezvous point for all existing and new directories, and would provide a standard mechanism for looking up information about entities on the Internet. According to Novell's application, with the .dir TLD, identities that already exist can be used and the information associated with those identities is always available and is always secure. Novell proposes that existing domain names that have associated directory support can obtain a new domain name from the .dir TLD that points to Novell's directory service. For example, <abc.xyz.com> would register its directory server under the name <abc.xyz.com.dir>.

SRI International's application for .geo asserts that the .geo TLD will provide a complete, virtually free-of-charge, and open infrastructure for registering and discovering Georeferenced information on the Internet. Georeferenced information is information that represents a place, object, or process within a geographic location.

Threshold Review

The evaluation team did a threshold feasibility review of these applications to determine whether they were generally consistent with preservation of Internet stability and whether they demonstrated well-thought-out plans, with provisions for adequate resources, to accomplish their stated objectives. As a result, the evaluation team concluded that the Novell and SRI International proposals merited further review.

After considerable threshold review, the technical team concluded that the Monsoon Assets Limited proposal was not a strong candidate, as presented, for inclusion in the initial, proof-of-concept phase. Their new service would require substantial search capabilities, including the ability to handle misspellings and near matches, which would require technical approaches and provisioning not specified in the proposal. Further, the proposal did not answer many of the technical questions that were posed in the application. Also, the proposal's business plan was not as thorough as other plans in many areas. For example, a key underlying assumption of the proposal is that other third party businesses (yellow page directory companies) would fundamentally change their businesses in the manner proposed by Monsoon. While the proposal did provide some support, it did not provide a reasonably clear strategy and rationale for convincing those other businesses to make this fundamental change. Overall, it did not demonstrate a thorough analysis of what would likely be involved in implementing the proposal, a careful explanation of all relevant issues, or well-articulated assessments of the business, financial, technical, operational, and marketing requirements for implementing the proposal.

Application of Criteria

The evaluation team considered how each of the August 15 Criteria should be applied to the remaining two applications in this group (Novell and SRI International). The applicability of those criteria may be summarized as follows:

The applicable factors identified above may be analyzed as follows:

Stability

The threshold review concluded that both Novell and SRI International presented proposals that provided for stable operation of the proposed TLD. Since these proposals entail new services that might become "orphaned" due to failure of the service to catch on, however, it is important to consider the proposed plans for long-term sustainability of the service in the absence of broad market acceptance, and the consequences for stability due to discontinuation of the service.

Novell

Since the registration process for Novell's proposed TLD requires verification that the registrant owns the base domain name <example.com> from which the .dir name is derived <example.com.dir>, this new TLD will put some new load on existing TLD services to check registration of those base names. In the long run, this traffic is unlikely to be heavy enough to cause significant problems, but it could produce excess traffic during any startup surge. Generally, Novell's proposal seemed to underestimate the initial demand for new registrations, which could cause various stability problems when it comes on line. The Whois service is replicated only in a single central site, leading to possible loss of this service in case of catastrophe. Since the Whois service is not necessarily crucial to operations, this aspect of stability is of secondary importance.

SRI International

The .geo proposal requires a default GeoRegistry to handle information on how to access Geodata for all cells on earth. For many cells that do not cover "popular" geographic locations, no agency might be willing to pay the expenses of maintaining such a lookup service. SRI hopes that organizations will contract to serve as the default registries for various portions of the Earth, but some areas might have no takers for this role. Therefore, SRI proposes to provide default services for some cells. The result is that if SRI decides to stop supporting the .geo service for whatever reason, moving the basic registry to another provider will be insufficient. The default GeoRegistry for the unpopular cells would also need to find a home.

In addition, a new services TLD applicant must realistically assess the business, financial, technical, operational, and marketing requirements for implementing the proposal and procure firm commitments for necessary resources. Some of the significant factors in evaluating whether the proposed general purpose TLDs in the new services category will be attractive to a significant sub-market in which it can compete effectively and the adequacy of marketing and promotion plans are summarized in the following table:

The table seems to indicate that SRI International appreciates, and is willing to make, the resource commitment of equity, marketing and employees required to be attractive to a significant sub-market in which it can compete effectively. Novell, or its registry operator, Tucows, Inc., apparently expects no marketing expenses. Novell does not adequately discuss its marketing plan or expenses.

Proof of Concept

The Novell proposal suggests that their service will replace many existing mechanisms for maintaining and finding information about entities on the Internet. To a large extent, these mechanisms are not related to the existing DNS, so this element of proof of concept is not directly relevant to any DNS concepts. One unusual aspect of the Novell proposal, however, is that names in the .dir TLD would only be granted to owners of existing names in other TLDs. The choice of the Novell proposal might thus provide some insight into whether this concept is feasible, given issues like the expiration of the base names and transfer of their ownership. Also, Novell suggests that introducing this TLD will largely avoid new intellectual property issues because those will all be resolved in the context of the base names that the new .dir names are derived from. Selection of the Novell proposal is likely to prove or disprove that contention. This value of the Novell proposal as a proof of concept for DNS systems is highly related to whether their naming scheme will be used by other TLDs in the future, however.

Selection of the SRI proposal is likely to lead to the proof of concept of a few new ideas in DNS service. SRI chooses one possible method of mapping names to geographical locations. The fate of their service may shed some light on whether this method is a suitable one. Their proposal essentially assigns overlapping responsibilities for providing rich services to portions of the name space to different GeoRegistries. Whether this is a feasible way to handle geodata and other services in the Internet is likely to be demonstrated by the success or failure of this proposal. Generally, .geo is an entirely different way of using names in the Internet. Its success or failure might demonstrate the likelihood of success of other revolutionary naming schemes.

Enhancement of Competition and Meeting Unmet Needs

In the analysis of these proposals, the enhancement-of-competition and meeting-unmet-needs factors are largely parallel, since both of them rely on establishing a new service (for a LDAP directory and for geodata) that must "catch on" with Internet users before it can become competitively significant. The proposed registry-registrar models are also significant to the assessment of competitive effects.

Novell

Technically the Novell proposal is straightforward. The translation process would point to a directory, in particular an LDAP-conforming directory object. However, there are other types of directory objects, so the allocation of a TLD to one particular type may be limiting. In general the directory would be accessed mechanically, not directly by human action.

In terms of meeting unmet needs, the service proposed for .dir could be provided in many other ways that do not require a new TLD. The .dir namespace is essentially a parallel partial duplication of the namespace of other TLDs. Arguably a more technically sound way to achieve the functionality of matching a DNS name to a directory server that keeps information on the name is to attach data to the existing name in some manner, rather than to create a parallel namespace. A significant amount of work on directories for the Internet is being done in other ways, and the proposal does not adequately address how it relates to that work and whether it would be compatible with it.

Novell proposes to use a registry-registrar model and does not restrict the ability of accredited registrars to offer registration services within the TLD, and in that sense enhances competition at the registrar level. Novell plans to use standard methods to choose registrars to work with their services.

SRI International

Evaluation of whether the .geo proposal enhances competition for registration services encompasses several inquiries, including whether the proposed TLD and registry is likely to be attractive to a significant sub-market in which it can compete effectively, whether distinctive services are being proposed that will meet the needs of those not being served adequately by existing services, the adequacy of marketing and promotion plans, the competitiveness of proposed pricing and service levels with other TLDs and operators having significant market share and restrictions on accredited registrars. Many of these issues cannot be easily resolved, but, to the extent possible to determine, the .geo proposal answers these inquiries sufficiently.

In terms of meeting unmet needs, the functionality provided by .geo could arguably be provided by creating a different form of geographical database and specifying a different mechanism of accessing that database. However, it is possible that the use of DNS is as good a way to achieve this functionality as any other, and this concept may be worthy of testing. There is no existing agreed-upon mechanism for achieving this functionality.

Further, a general purpose, new services TLD applicants should offer attractive pricing and service levels, particularly because the new services TLDs typically must achieve broad market acceptance in order to succeed. The following table summarizes the applicants' proposed pricing and service levels:

The Novell price is in line with other general purpose TLD prices. The SRI price is harder to judge, since they are charging for very different things than other TLDs. The service levels provided by both proposals are in line with levels provided by other general purpose TLD proposals. The technical capabilities listed in the table above reflect implementation of a high-quality registry by the registry operator: JVTeam for the SRI application; Tucows, Inc. for the Novell application.

Another competitive issue considered in evaluating the proposal is whether the proposal restricts the ability of accredited registrars to offer registration services within the TLD. SRI does not anticipate the need for domain name registrars during the initial test period, because GeoRegistries will communicate directly with the registry to register cell service domain names. Traditional domain name registrars can be accredited as GeoRegistrars to provide competition.

Enhancement of Utility of the DNS

Enhancement of the utility of the DNS is another of the August 15 Criteria. SRI International's application appears to sensibly add to existing DNS hierarchy, does not appear to create or add confusion to the existing DNS hierarchy, and is semantically far from existing TLDs to avoid confusion.

Novell's application is semantically far enough from existing TLDs to avoid confusion. Novell's plan to create names by appending .dir to existing DNS names might create some human user confusion about the difference between a name and the name with its .dir extension. In most cases, however, the .dir name will be used by software, not by humans, so the possible confusion seems likely to be minimal.

Enhancement of Diversity

Enhancement of diversity is the other of the August 15 Criteria particularly relevant to analysis of proposals for TLDs in the new services category. Evaluation of whether SRI International's proposal enhances diversity encompasses several inquiries, including diversity in business models and of geographic locations. SRI International in based in the United States. Its registry operator, JVTeam, is owned by NeuStar and Melbourne IT, an Australian company. Novell in based in the United States. Its registry operator, Tucows, Inc., is a Canadian company.

SRI's proposal assumes success in enlisting many sites to participate as GeoRegistries. Given the geographical orientation of their system, they anticipate that these GeoRegistries will be spread throughout the world, typically serving geographical areas close to their locations. If SRI is successful with the model they expect, companies throughout the world will serve as GeoRegistries as part of the overall system. The SRI approach does not limit the number of GeoRegistries that can serve a particular area, so for areas where more than one group may wish to provide GeoRegistry services, all could be accommodated.

Novell expects geographically diverse groups to enlist as registrars for its service.

SRI International's revenue model is based upon a fixed license fee and variable subscription fees. Novell's revenue model is a subscription-based model with a one-time registrar set-up fee.

Protection of the Rights of Others

The format of basic .geo names makes trademark and other intellectual property issues unlikely to arise for the basic names in SRI's proposal. However, GeoRegistries are permitted to register different forms of names under .geo, which could lead to these kinds of problems. SRI proposes a serious accreditation process for each GeoRegistry to handle these issues. Because establishment of a GeoRegistry is a heavyweight operation with significant fees, there will be both time and resources to perform such accreditation.

Novell expects no new problems related to property rights because they will only register extensions of existing names, and only to the holders of those names. If the ownership of a DNS name changes, they do not specify how the system will ensure that the ownership of the associated .dir name changes.

Recommendations

The SRI .geo proposal is an interesting and innovative expansion of the functionality of DNS. The technical, business, and legal details provided suggest that the service can be successful, and might revolutionize use of the Internet.

The Novell .dir proposal is technically feasible and has an average business plan. Many alternative methods not involving new TLDs exist to attach directory services to existing names, raising the unanswered question of why a TLD-based approach is superior. Novell's complete concentration on their own LDAP directory system, with no serious consideration of other directory systems, is a negative element in their proposal. In the absence of consensus on how to best handle distributed directories in the Internet, establishing a TLD that is tied to one of the approaches is perhaps unwise at this time.