Re: [wg-c] WG questions

[Kent Crispin <kent@songbird.com>]

Sorry for the length...

On Thu, Jul 15, 1999 at 03:46:40PM -0400, Jonathan Weinberg wrote:
> 	David urged, a couple of days ago, that we need to think about a ICANN
> mechanism for determining which individuals or companies get to run which
> gTLDs.  Along somewhat the same lines, Kent's suggested a process for
> getting gTLD/registries approved:

[description deleted]

The process I proposed is very general.  However, I was specifically
addressing the approval of gTLD names, and registries in the
abstract sense.  I do not think that the process is appropriate for
selecting registry operators, or registries in the concrete.  You
seem to be mixing the two notions. 
> 
> I can imagine a variety of different approaches we could take here, and it
> seems to me that they fall on a continuum between two polar extremes.

I don't see them as poles.  They are just two different proposals, 
in a space with much higher dimension than one.

> On
> one extreme, we might have a decision-making process taking into account
> the particular facts of every proposed gTLD,
[...]

In my proposal, while I don't discount the possibility of a WG
specifying some registry operator for a particular TLD, in general I
think it should be actively discouraged, and in practice I think the
WG process itself would make it almost impossible to do.  I think the
WG process is mainly useful for selecting of names and properties of
TLDs, not specifying who should run them.  

More importantly, any actual selection of a registrar would be a
business affair involving contracts and lawyers, and the DNSO does
not seem like a good platform from which to specify such
relationships.  In fact, I argue that the selection of registry
operators is completely outside the scope of the DNSO. 

To be concrete, I think the DNSO could decide that a .web TLD would
be a good idea.  It could define general characteristics for the .web
registry.  But it simply cannot decide that IODesign (or Emergent, or
NSI) will be the entity that should run the registry.  The name of
the TLD, the policies of the registry, general policies about
registries and registry operators -- these are all things that fit
into the definition of the purpose of the DNSO as a policy making
body.  But selection of a particular commercial entity to run a TLD
is completely out of scope.

This is no different in principle from the DNSO selecting a 
particular business to be a registrar.  (The DNSO should have input 
on the *policies* instantiated in the registrar accreditation 
guidelines, of course.)

I can imagine special cases where a particular entity might be named
in a TLD policy -- some international medical society might be named
in some connection with a .med TLD, for example (though probably not
as a registry).  But those cases are far in our future, and they 
would have to be justifiable as policy.

> At the other extreme, we might have an approach under which the
> decision-maker would do its evaluation by looking to a set of simple,
> hard-edged, black-letter rules, so that the results turned mechanically on
> the answers to specific factual questions.  For example, ICANN might
> establish a set of objective, published rules (1) setting qualifications
> designed to weed out the incompetent, undercapitalized, irresponsible,
> fly-by-night and/or criminal; (2) designating strings that may not be used
> to identify a TLD; (3) limiting any registry operator to __ TLDs; and (4)
> requring that a registry {must, must not} be non-profit and/or shared.
> Having done that, it could select a predetermined number of applicants
> meeting these criteria by FCFS or lottery.[*]

These are almost entirely criteria for selecting registry operators,
and there clearly should be some objective criteria for that purpose. 
However, if you separate out approval of a TLD, then the rules are
requirements for entities that might bid on running a registry. 

*Policies* for the selection of registry operators, and *general
requirements* for registry operators are of course within scope for
the DNSO, and for this WG, and they are more amenable to "hard-edged
rules" than selection of TLDs. 

> An advantage of the first approach is that no relevant factor would be
> off-limits to the decision-makers; they could do their best to choose who
> they thought the *really best* entity to run a given TLD would be.  That
> question, though, in this sort of context, is commonly insufficiently
> determined, and effectively unanswerable.

Indeed, and I don't see that decision being made by the DNSO in any 
case. 

> An advantage of the second
> approach is that it would simplify the decisionmaking process, making it
> more difficult for the decisionmaker to engage in biased or arbitrary
> decision-making, and making results more predictable.

I agree that the decision making would be simpler in most cases. 
That is not an obvious advantage in complex environments, however. 

> I'd like to suggest that we lean towards the second approach.  Requiring
> every new TLD and its affiliated registry to go through a WG, the Names
> Council and the ICANN Board, I suspect, will be a slow process that, in the
> highly politicized and contentious gTLD context, will maximize
> opportunities for lobbying, political considerations, obstruction, and
> hidden bias and arbitrariness.
>
> If we want it to be easy to accredit new
> gTLD/registries, it seems to me, ICANN should set out the governing rules
> at the outset so that its actual approval process can be largely mechanical
> rather than discretionary.  Am I wrong about this?

To paraphrase Peter Venkmen (Bill Murray) in Ghostbusters: "I'm fuzzy
about this right/wrong thing." :-)

Some further points:

0) In no case should we be specifying registry operators, for the 
reasons mentioned above.

1) You speak of a set of hard edged rules, but leave unanswered the
question of how they will be developed, and how they will adapt to
changing circumstances.  The criteria here are rather more
complicated than picking an age for majority -- in fact I suspect
they are a distillation of the criteria that would have to be argued
for approving a single TLD, and I suspect that "hard-edged rules"
will be at least as controversial as selecting TLDs directly, because
everyone will immediately interpret proposed rules in terms of their
pet TLDs.

2) Every WG that accredits a TLD sets a precedent -- if you accredit 
.firm as a tld on the grounds that we need competition for .com, 
then many of the same arguments apply to .biz, and, precisely to the 
extent that there is a demand for another competitor to .com, there 
will be motivation to accredit .biz.  The accreditation process 
becomes easier as the precedents are set.  Another way to look at 
this is that instead of a set of hard edged rules, the wg model will 
evolve a set of softer edged precedents.

3) The nature of the rule model seems implicitly permissive -- you
set the rules, and then let the thousand flowers bloom.  Everything
is permitted unless there is an explicit rule that forbids it.  

That freedom sounds wonderful, but it negates the possibility of
design.  For example, I mentioned the possibility of a set of TLD
names for TM categories.  This would become much more difficult if
the "natural" names for those categories were immediately snapped up. 
While I personally don't think a TM category name space, in
particular, is a good idea, I am seriously opposed to a model that 
makes it very difficult to have designed name spaces in the TLD
space. 

4) [This point deals with rules for selecting operators primarily,
which is problematic to begin with.] There is effectively no rate
limit in the rules model -- new TLDs are added as fast as the
contracts can be inked -- unless you specify arbitrary limit rule. 
This has the unfortunate consequence of fostering a goldrush
atmosphere, something that might be avoided with a more deliberate
approach to tld approval. 

You mention FCFS as an allocation mechanism.  This in my opinion is a
particularly bad idea, because it really encourages a speculation
driven gold-rush.  You can control that by making the barriers to
entry very high, but then you discriminate unjustifiably against
small players.  Even more, Internet access is not distributed evenly
around the globe -- it is not a level playing field.  FCFS inevitably
discriminates against developing countries, by giving all the "good
territory" away in the beginning.

A lottery is somewhat better as an allocation mechanism, because you 
can therefore set arbitrary limits like "we will approve 5 TLDs this 
year".  But the experience of the IAHC was that the lottery idea was 
almost uniformly detested, and arbitrary limits on the number of 
tlds obstruct the possibility of related groups of tlds.

5) Any set of hard-edged rules will need an appeals process, and
someone to interpret the rules.  There will be those who analyze the
rules for any possible loophole and try to game them, and
consequently present puzzling cases for interpretation.  You have the
issue of who should decide -- the ICANN Board, the NC, an appointed
"official", a separate elected panel, etc.  You also need to deal 
with challenges.

The WG process, to a large extent, is its own appeals process.  It 
is more difficult to get a name approved; those that do make it are 
guaranteed to have substantial support.

-- 
Kent Crispin                               "Do good, and you'll be
kent@songbird.com                           lonesome." -- Mark Twain