Comments on CRAI "Revisiting Vertical Separation of Registries and Registrars" report

[Patrick Mevzek <contact@xxxxxxxxxxxx>]

Please find below some of my comments on the CRAI study about the
current registry/registrar separation in domain names market.
I'm writing these comments as an individual generic Internet user, owner of some
domain names for personal and business needs, and a founder of a
company working with registries and registrars.



The registry-registrar split was decided at a time where 2 conditions
were true: there was a clear picture of the market, with on one side
suppliers of domain names and on the other side buyers of domain
name, and the number of buyers was very large and on the rise.

As long as these two conditions hold, I see no reason to change the
current split, or at least no reason to make it more lax, and on the
contrary every reason to enforce it even more.

As an example, some registries decided to rise their wholesale
(registrar) prices thank to their new ICANN contract. What happened
on the market is that some registrars decided to change their prices
to take that into account, while some others decided to lower their
own margins so that the end customers price would not change.
This shows how this registry/registrar split was successful, as it
enabled end customers and users of domain names to be insulated in
part from some decisions by registries. It keeps the market
competitive and open to new actors.

Of course we can find some drawbacks of this split. For example, some
registry services (like .NAME second level email redirections handled
by the registry itself) did not sell very well, in part because few
registrars decided to offer them to end clients where the registry
had no other choices. For some services, it may seem that registries
and registrars are in direct competition.

Now, the problem may be with new gTLDs coming in a near or far
future.
If we agree on the previous 2 conditions outlined above, we may see a
problem for a split in the new gTLDs because 
- either there is not anymore a clear separation on who operates/sells
domain names and who buys/uses them,
- or the number of buyers/users is very low.




It is not necessarily true that the opening of new gTLDs will bring
many new actors (as registries) since the entry barrier is high, and
not even taking into account prices. It may happen that many new
applicants will be current registries, registrars, groups of them, or
entities using them as technical operators. This can be seen as a
failed attempt to bring more competition (besides bringing more TLD
strings), and allowing any form of smaller split between registries
and registrars as we have today would bring even less competition
(in a market where one gTLD accounts for
more than 70% of all gTLD domain names, and one registrar accounts
for 27% of all gTLD domain names - where the report says the market
appears "unconcentrated" it can be said that the HHI value found
(952) is close enough to the level (1000) when we start to speak
about "moderate concentration" even more since registrars were taken
one by one and not as groups of related entities as they exist today).




The report says:
"For registries not operating under a binding price cap, the arguments
in favor of vertical separation and equal access requirements are 
less clear cut. We would recommend that ICANN take steps towards 
relaxing one or both of these requirements.
Any such liberalization of the vertical separation and equal access
requirements should be taken
gradually, as these sorts of reforms are difficult to reverse."


Let me just start with the last sentence: after providing no hard
facts in favor of what is recommended, we are warned that any change
may be dangerous because "difficult to reverse".
So the recommandation is to take a one way road going somewhere we do
not know yet, which I found at least scary.

Now for the main part about relaxing the two requirements.
I would like again to put that in perspective with the current state
of market (its concentration as outlined above) which is probably the
state it will remain in for many years, even with the introduction of
new gTLDs (some may argue even that it will become even more
concentrated because of the introduction of new gTLD but this is
another debate).
If someone believes that some requirements should be relaxed for
registries without binding price cap (that is all sponsored
registries), this means that a problem is perceived in these
registries, and we first should know what the problem is. Is it the
low number of registrations ? Something else ? 
Before offering solutions, there is a need to clearly state the
problem, how the solution should work to solve this problem, and what
would be the consequences and side effects. To do that, monitoring
TLDs is needed (as discussed later in the report and in my comments
below).

As for the two models offered (on page 3),  I do not agree at all
with the "single-organization TLD" case, as in my view this case
should not exist at all. Introducing new gTLDs being captured by
some organization for their own use provide no benefit to the
Internet population at large and even more harm it because it creates
even more contention, now on the top level of domain names, without
providing any new competition, or innovation.
Therefore no exemptions to the current split system should be granted
so that these cases do not happen too much, or even not at all.
By mandating that the split exist, this may refrain some applicants
to grab such "vanity" gTLDs which in the end is good for the Internet
at whole.
This is correlated to one of my two conditions explained above: has
the market a clear distinction between sellers and buyers of domain
names? If not, I believe there is a problem, for the specific gTLD
considered.
Let us take section 5.2 of the report:
"For example, a corporation striving to use its brand name TLD as a
sales channel would
likely prefer to have complete control over what domain names are
registered in the TLD and
may also want to keep some registrations confidential, for example if
the company is planning
to launch a new product that has yet to be announced."

If a entity is in this case, it has a very simple solution with the
benefit of not costing $185000 or more but only something around $10
: they should not apply for their own gTLD but just register one TLD
in any currently available gTLD. With their own domain name, and the
power of unlimited subdomains, they would be able to completely
control how their multiple subdomains are used, by whom, for free and
it can be completely confidential as needed (since there is no whois
requirements for subdomains).
On the opposite, if the entity do not wish to follow this simple and
logical advise but prefer to go the long way of grabbing their own
vanity gTLD for their own use and noone else, they should expect to
hit the problems outlined and this is normal and no requirements
should be lowered to simplify their case.


The second recommandation (to let registries own registrars or the
opposite as long as the registrar does not register domain name in
the registry it owns or that owns it) makes more sense, and I could
generally agree with it. Of course proper safeguards need to be put
in place, and specific attention to the fact that "registries" are
sometimes and probably more and more often a two side coin with on
one side the sponsoring organization dealing with the policies and
such, and on the other side a technical operator chosen by the
sponsoring organization to conduct day to day technical operations.
The separation outlined in this recommandation must be made on the
two level, the sponsoring organization and the technical operator.
It should also be noted that other kind of structure could be
created, such as having an holding company controlling at the same
time a company acting as a registry and another company acting as a
registrar. In cases like that, when the registrar do register domain
names with the registry, questions could arise on proper separation
or not. So the created contractual limits should take into account
that there is always a risk, maybe small, of people building specific
structures and models to circumvent those limits.

This also means two other things: it should be clearly *publicly*
known all data about registries and registrars, that is enough data
to be clearly able to see who owns who through which companies and so
on. It is not a big problem already for registries, due to their
current low numbers, but it is already a huge problem currently for
registrars, as some other studies have shown even basic data such as
true postal address and phone numbers are not really available for
all current registrars. So first requirements should be put in place
so that data is published and available to anyone to make sure that
owning requirements are followed by everyone. And the other thing is
that there should be a process inside ICANN that first periodically
reviews and verifies that all registries and registrars do in fact
follow this requirement, and at the same time review if this
requirement still make sense and does not seem to be circumvented.


As a side note, one recommandation ask ICANN to "monitor the
performance of these new TLDs", however no explanation of what the
performance of a TLD is: is it the absolute number of domain names ?
the rate of increase ? the number of registrars ? their diversity 
(and what kind of diversity ?) ?
the number of end users ? their diversity ? the percentage of domain
name really used (in opposition of simple warehousing) ? other
factors ? a combination ?
In all cases, I agree with the recommandation to monitor TLDs, new or
actual, but performance criteria should be established *before* any
new TLD is introduced, so that the metric is not changed along the
road to suit some needs.


Another point worth mentioning (top of page 10), if the absence of
the concept of resellers in each and all ICANN contracts, policies
and procedures, which may create, as the report says, some loophole.
ICANN should take time to study the presence and influence of
resellers in domain name markets and take that into account in all
future documents, starting with the registrar agreement and of course
all new gTLD applications.

There are currently around 1000 ICANN accredited registrars. It can
be argued with such a huge number, there always should be at least
some registrars interested by any new gTLD which implies that the
requirement for new applicants to deliver their domain names only
through ICANN accredited registrar is not an hard requirement to
achieve.
Or, this may mean that the current 1000 ICANN accredited registrars
are not all "true" registrars with end users activity. But this would
be another debate.

New gTLD applicants should take into account in their business plan
how much their new offering will be accepted by registrars, and what
it can or should do to make sure that many registrars will sell their
gTLD to the end user. It is a critical part of their business plan,
and should be taken into account while reviewing applicants
proposals.

It would be hard on one side to cite the eligibility requirements to
get a domain name under their new gTLD as a major advantage of their new
offering, showing how it is tailored to a specific population, hence
bringing value and exposure for them to the Internet, and at the same
time to say that these eligibility requirements make it more
difficult to have registrars willing to deliver their products.
They should balance both sides and find a compromise they can live
with. But removing the requirement to go through ICANN accredited
registrars just to simplify things in this case is not a good move.

In that way, there is no need to relax the current requirements of
split and new registries should be required to use current and future
ICANN accredited registrars only to deliver their TLD, but to foster
innovation, some temporal or space exceptions could be accepted,
such as some delay after starting during which the registries can
find other channels (as it has been written in the 2000 new contracts and
reviewed by the CRAI report on page 14) as long as the total volume
of domain names do not cross some limit. These plans should be
disclosed in advance, and be part of new applicant proposal, with
details on all limits (delay and domain names volumes). These should
not be changed after the gTLD has been awarded.

At the same time, ICANN registrars accreditation should be made
faster, simpler and less costly. Also ICANN could offer incentives to
register in some new gTLDs by monmentarily lower their fees which
happen on multiple level : the yearly fixed amount paid by each
registrar (only depending on the total number of registrars), 
the yearly variable amount paid by each registrar (depending on the
percentage of total domain names owned by the registrar), and the
ICANN fee to be paid by each domain name transaction.
Even the other indirect contractual requirements, such as insurance
and payment bonds or equivalent.
These fees should be made progressive, so that two effects can be
achieved: easy to create new registrars for new gTLDs, and easy for
current registrars to add a new gTLD to their portfolio.
Otherwise, as some registrars will not enter some gTLD before
believing a "sensible" amount of registrations could be made, there
is indeed a risk not to have a lot of competition in some new gTLD.
But again, this is something applicants should think about before
offering their proposal, and not regretting it after and asking for
requirements to be relaxed to better suit their cases.

This makes everything simpler as there is no need to try defining
exactly what a "single owner TLD" is nor a "money-making venture"
(that the report wish to exclude).

In short, as a summary, I see no need to relax the current
requirements (for a registry to deliver products only through ICANN
accredited registrars) as a whole and generally, while some
exceptions could be put in place (lowering prices for registrars in
new gTLD, allowing the registry to try other channels during some
limited periods and below some volumes, etc.) to help new gTLDs find their
market at the beginning.
Any such exception should be known in written before the gTLD starts
and its application be accepted, and specific "performance" criteria
should be defined also before starting, to be able to judge the
results of each gTLD after some specific delay has passed
(the criteria may not necessarily be the same for all new gTLDs as
they should of course take into account the corresponding community
served by the gTLD on one side, and the global Internet community on
the other).
Other than that no vertical integration should be allowed.

-- 
Patrick Mevzek
Dot and Co <http://www.dotandco.com/>