ICANN ICANN Email List Archives


<<< Chronological Index >>>    <<< Thread Index >>>

Comments on "Revisiting Vertical Separation ,of Registries and Registrars"

  • To: crai-report@xxxxxxxxx
  • Subject: Comments on "Revisiting Vertical Separation ,of Registries and Registrars"
  • From: Eric Brunner-Williams <ebw@xxxxxxxxxxxxxxxxxxxx>
  • Date: Fri, 05 Dec 2008 20:58:42 -0500

I wish to make five somewhat critical points concerning the "single-organization" model which I denote as "CRAI-1", I wish to make some not-so critical comments on the CRAI-1 model, which will be the subject of another note. My comments on the other portion of the CRAI note concerning a "hybrid model", which will be the subject of a third note.

1. String space is valuable, the CRAI-1 entities do not require string space.

2. The IANA root servers publish globally accessible public name spaces, the CRAI-1 entities do not require globally accessible public name spaces.

3. Bulk registrars add less than a dollar to the domain/year fee for price capped inventories, the CRAI-1 entities do not provide more value than the bulk registrar margin.

4. RFC 1591 is binding on all zone administrators delegated from the IANA root, ICANN included as the present administrator of that zone, the CRAI-1 entities are incompatible with RFC 1591.

5. From the IAHC period to the present, rational economic actors have attempted to tranform the circa-1999 monopoly market into a competitive market, the CRAI-1 entities operate in the pre-1999 market.

Supporting commentary.

The CRAI-1 model does not need "a name" to communicate between the registry, the registrar, and the registrant, as these are all the same party. Therefore, no necessity for a meaningful name exists. String space is precious, a significant portion of the cost and complexity of the current new gTLD process exists because of the possibility of contention, and the CRAI-1 entities can be assigned strings composed primarily (or entirely, with the necessary restrictions) of digits, with minimal impact to the string space available to contemporary, or subsequent round applicants. We've already spent six figures on the contention algorithm and that's just the tip of the iceberg.

The CRAI-1 model does not need to publish its zone files via the IANA root servers. The CRAI-1 model is that information which "coordinate confidential internal business processes" may exist in some name space, yet not be disclosed to third parties. The IANA root servers do not implement "split DNS" or any "view" of the tree other than the common, shared, public view from the unique DNS root to the arbitrary leaf node. Any entity seeking to not disclose node and leaf existence data of a tree need only publish the existence of the root of the tree to its intended private disclosure set.

This design goal of the CRAI-1 model is trivially achieved via alternate address spaces, e.g., MILNET, DISNET, etc., and by alternate root servers, e.g., Kashpureff's AlterNIC, Karl Denninger's eDNS, Idealab's New.Net, OpenNic, as well as "browser hijack" exploits of the IDN market such as the nearly decade-long attempt to create a proprietary IDN captive market. This 3rd-party non-disclosure design goal, or more generally, exception to the existing gTLD consensus policy regime, is also met by using a second-level domain, rather than a top-level domain.

It is worth noting in passing that using only the concealment of node and leaf data of a tree to provide a restricted access model to data, which may or may not contain references to that node and leaf data, is a remarkable non-contribution to the security of distributed systems. If the SSAC hasn't made that plain to staff, it is still an awkard fact.

The CRAI-1 model is indistinguishable from a high-volume regsitrar. The CRAI-1 model is predicated, in part, on the margin between the price capped base registry price and the retail price being "inefficient". Additionally, the CRAI-1 model is predicated on the ICANN per-domain fee being "inefficient". The only applications of the CRAI-1 model that must fail due to the equal access and/or vertical seperation policies -- these "inefficiencies" -- are those which have a real value less than one dollar per domain year. This efficiency design goal, or more generally, exception to the existing gTLD consensus policy regime, is also met by using a second-level domain, rather than a top-level domain.

It is worth noting in passing that under some fairly common fact patterns, low and even negative margins are subsidized by highly competitive registrars which recover their registration subsidy margin by higher margins on hosting or "advanced" services. The CRAI-1 model cannot offer registrant "hosting" or "advanced" services without negating the argument for sub-dollar "efficiencies".

RFC 1591, Domain Name System Structure and Delegation is quite explicit about the responsibilities of administers of top-level domains. The words "trustee", "duty", "responsibility", "service", "equitable", "non-discriminatory", "no bias", "applied recursively", and "agree" frame the discussion of service, and are binding on the registry operator (ICANN). The CRAI-1 model does not contemplate any of these terms framing the discussion of service.

Since the competitive award of the NIC contract to the precursor of Verisign two deacades ago, to September 13th, 1995, when fees for domain names was first introduced, to 1999 when the shared registration (not registry) system was proposed, to the present, hundreds of well-intentioned people have worked in an attempt to deconstruct the 1999 market, some even employed by the incumbent monopoly, Verisign. The CRAI-1 model places some initially bounded, but incrementally less constrained, set of entities within the pre-1999 policy model. Repeating the pre-1999 market conditions, "in miniature, and en mass" will not assist in achieving the second of the four "critical" principles of the "White Paper", competition.

Finally, a comment concerning an apparent reference to my work in Working Group C, "A Position Paper on some new gTLDs". The quoted material on page 12 of "Revisiting Vertical Separation of Registries and Registrars" reads "For example, a TLD devoted to North American aboriginals, as was proposed to WG-C, may want to ensure that specific tribal names are only assigned to legitimate members of that tribe". This was offered as an indication of support for the CRAI-1 model by some of the Working Group C participants. In fact, the quote is from "Position Paper B on New gTLDs", by Milton Mueller and his gang of idiots, who were ideologically opposed to any policy constraints what so ever, including vertical separation and shared registration system.

As recently as the Los Angeles meeting in 2007 Milton managed to repeat his view that "Cherokee" belongs to the fastest six-dollar buy in the west.

Eric Brunner-Williams,
Disclaimer: I work for CORE Internet Concil of Registrars as CTO, and was CTO of the .cat registry for some time prior to its launch, and had similar responsibilties for .biz prior to its launch, and I expect I'll present an updated application based on "A Position Paper on some new gTLDs", whether Milton likes that or not, so read my comments with those present and past interests in mind.

<<< Chronological Index >>>    <<< Thread Index >>>

Privacy Policy | Terms of Service | Cookies Policy