Re: [gnso-vi-feb10] Question on Proposals

[Eric Brunner-Williams <ebw@xxxxxxxxxxxxxxxxxxxx>]

On 4/9/10 11:45 PM, Richard Tindal wrote:
> 
> Eric B-W and Jeff N,
> 
> Your proposals strictly limit a registry operator from controlling a 
> registrar in its TLD.

Correct, at least w.r.t. CORE's.

> What is the purpose of this limitation  (i.e.  what harms do you see coming 
> from a registry controlling a registrar in its TLD)?

A minority interest cap sufficient to allow equitable access to
registries operating a shared registrar is adequate to solve the
problem that at registry start-up, where the registry is sufficiently
distinct from the current market, it will lack registrars.

A consequence of economically executing a shared registrar is that the
EPP and business model divergence among the participating registries
is self-limiting, and any registrar implementing the same EPP and
business model supporting registration profiles has access to all of
the participating registries, reducing the cost of entry for second
and subsequent competitive registrars.

The competition policy benefit is that applications are not invisibly
constrained to the existing dominant model defined by ASCII, dollars
(or euros or ...) and the OEDC economies, creating diversity of models.

The public interest policy benefit is ending structural economic
discrimination reducing the abilities of unserved and underserved user
populations to use the domain name system.

The ICANN policy benefit is that exception to the general requirement
for registrars is not necessary to achieve these fundamental policy goals.

But you've read all that twice and are really asking why not some cap
significantly in excess of the 11.2% to 15% in the current contracts.

First, because the policy goal of ending structural economic
discrimination, in the presence of significantly larger than the 2000
to 2006 experience cohorts of registries entering operations per
quarter, is removed by a lower cap.

Second, because there does not appear to be even a linear relationship
in control and benefit, to support a 2x claim for the nominal 30%, or
a 3x claim for the nominal 45%, ... so the reduced diversity of
control, the reduced size of the pool of registrar sharing registries,
has no corresponding benefit, and at the figure you asked privately,
75% control, is indistinguishable from 100% control, which is an
exception from Recommendation 19.

Third, because the least change to effect an identified goal is the
best change.

Fourth, because the number of registries which must offer an EPP
server set of extensions which a registrar is economically rational in
implementing in its client is significantly greater than 1, the
variation in EPP extensions, a problem since the 2000 round, is 1/Nth,
where N is the number of registries necessary to form a registrar,
nominally 7 at a 15% cap, the variation possible, and unfortunately
likely, if the registry associated registrar had no economic
motivation for commonality.

Restated, where registries are incented to deviate minimally from the
core EPP standard, registrars are incented to implement their
deviations from the core EPP standard, to mutual benefit.

Re-restated, a minority ownership or control cap provides nearly an
order of magnitude reduction in the overall technical and contractual
complexity of registry and registrar relationships. Complexity is
cost, and as the number of registries approaches the number of
registrars, the growth of complexity cost, to registrars, and to
ICANN, and others with multi-registry interests, does not scale. This
affects no registry directly until a complexity cost limit, at which
point new registries have no default registrar interest.

Reminder, at Paris and subsequent I've asked the RC, now RSG, "who is
going to implement the EPP warts of the new gTLDs?" I've never gotten
a single registrar to indicate interest in the new gTLDs -- as a
registrar.

Thanks for your question about the registry cap on ownership or
control of registrars. Don't forget the other direction.

Eric