Re: PAB charters with teeth

Kent Crispin (
Thu, 5 Mar 1998 09:53:59 -0800

On Thu, Mar 05, 1998 at 12:20:42AM +0000, William Allen Simpson wrote:
> I see that there has been a small number of comments on this topic while
> I was out of town, but this is the most cogent and well-formed:

Actually, I thought some of the other comments were rather good,
myself :-)

> > Date: Sat, 28 Feb 1998 22:27:34 -0800
> > From: Kent Crispin <>
> > I would like to back this up a level for just a moment, if I could. I
> > recently mentioned that I have both the and
> > domains. Bill pointed out that I was out of compliance with the
> > definition of the definition fot the .net TLD,
> >...
> > However, it is abundantly clear that I am not alone in my
> > transgression of rfc1591. As I recall, one of the IAHC drafts
> > recommended that the definitions for .net and .org just be dropped,
> > because they were widely ignored. The lesson to be learned from this
> > is that charters without clear definition and *legal* teeth will be
> > ignored.
> >
> I certainly agree.
> The real problem with RFC-1591 was never that things did not have a
> clear definition, but rather that NSI was registry and registrar, judge
> and jury, and there was no way to take the responsibility away from them,
> despite the terrible job they have been doing.

Out of fairness to NSI (did I say that?) there is another factor --
the incredible demand for names meant that people would fudge their
forms. Songbird is an organization, too...

> > Further, the IAHC reports define gTLDs as TLDs with essentially free
> > registration -- the *definition* of a gTLD is that it doesn't impose
> > any but the most general requirements for registration --people chose
> > a particular gTLD because of what it means to them, not because of a
> > structured meaning imposed by the charter.
> >
> I disagree. In such a case, there simply is no meaningful use for
> having multiple _defined_ TLDs, and we could simply have a root where
> everyone registers whatever is "meaningful" to them.
> TLDs are not meaningless labels, like "a", "b" .. "z". That was
> proposed (by Postel) and rejected years ago.

You missed my point, I think (and the IAHC's point): It's not a
qestion of there being a meaning, it's how the meaning is enforced and
defined. The essential characteristic of a gTLD (as opposed to a
sTLD) is that the registrants decide what the string means, not the
registry. The registry picks names originally, based on some idea of
an intended population, but after that it is the meaning of the
string, as perceived by the registrants, that controls who uses it --
a purely market driven approach, that is. For example, while .com, .net,
and .org presently have no enforcement, in fact registrants do
self-select TLDs appropriate to their use. If there were many more
names this self-selection would become dominant.

It's not so important when you have 10 gTLDs. But when you have 100
or 1000 TLDs, all with relatively meaningful names, people will
self-select the ones that are meaningful to them -- even very large
companies are not going to register 1000 domains, if there is a decent
spread of meanings. (Imagine

Bear in mind that as far as DNS is concerned all strings are equally
meaningless -- the only thing that matters is the distribution of
domains across the TLDs. One mechanism to get spread across the TLDs
is to write strict enforcable charters. Another is to have a broad
selection of names that are chosen so that self-selection will cause
the spread. These last are the gTLDs (according to my interpretation
of the IAHC's definition).

I think both mechanisms are valid, and in fact, I think the gTLDs are
really a particular case of the sTLDs -- the case where the charter
basically describes the above-described self-selection rule.

Charters really come into their own when you start talking about
truly specialized TLDs. Cary Karp has for some time advocated the
".museum" TLD. This would be a shared registry -- any registrar
could register names in it. But all registrations would have to be
approved by an organization designated in the charter. (It turns out
that an appropriate organization does exist; it is international in
scope; virtually all museums belong to it.) This is a simple, clear,
and easily enforcable rule; it puts essentially no burden on the
registrars for enforcement.

My favorite example is the .ham TLD -- you have to have a valid
amateur radio license to register a domain in it. This is also an
easily enforcable and completely straightforward rule, international
in scope.

Note that enforcement is purely a registrar/registrant function --
not a registry function. The CORE shared registry could handle sTLDs
just fine.

There are hundreds of realistic possibilites for chartered TLDs, with
clean enforcable charters -- you just have to find some blanket
international group that can do the approvals.

This is why I am lukewarm about writing charters for gTLDs -- or, to
put it another way: in my view the charter for gTLDs is that they are
"open registration" TLDs, where people self-select which ones they
want to use. I want there to be a defined class of TLDs with exactly
this characteristic. (I have tried to change the term to "open TLDs",
or "oTLDs", with little success.)

> Having TLDs for different areas is pretty useful, in that it aids in
> avoiding naming conflicts. As many different folks should be able to
> register as possible (as opposed to a few folks registering the same
> thing as many times as possible). That is what will allow the DNS to
> work!
> > With this definition of a gTLD (the definition I believe the gTLD MoU
> > was based on) there only needs to be one charter for all the gTLDs --
> > a charter that defines this open registration policy.
> >
> IMHO, that road leads to an unfettered free-for-all, and the same
> problems that we already have.
> Indeed, as an ISP, we have already received Unsolicited Commercial
> Electronic Mail (spam), from, urging us to register our
> current domain names in all the new TLDs.

Paul Vixies approach to this, and I have come to agree with him, is to
create a couple thousand TLDs. That will cure people from
registering in every TLD. If there is a fairly open process by which
TLDs get approved (both gTLDs and sTLDs) then it will be hard to add
more than 100 per year -- so this would keep us busy for 20 years.
By then we surely will have DNSng. The trademark interests will
resist for a while, but in the end they will be better off once they
realize that they simply can't search all the names.

> > It is my belief that a charter for a TLD is pointless
> > unless it has some enforcement mechanism defined.
> Again, I agree. We already have such an enforcement mechanism, but are
> lacking in well-defined policy to enforce.
> The mechanism as currently defined is very simple. The registrar assists
> the registrant in selecting the best TLD for their usage. When a
> registrant ignores the advice of the registrar (or the registrar does
> not offer any value added services or gives poor advice), and registers
> in an inappropriate TLD, or in multiple TLDs to shut out others, those
> others challenge the registration through the appeals process.
> Registrants that ignore charters will lose, and will waste time and money.
> Registrars that frequently give bad advice will lose customers and go
> out of business.

Right. This is the "and then a miracle happens" argument, you know
:-) This solution puts everything off on the appeals process, which,
as it happens, is completely voluntary, and besides is specifically
designed to handle trademark conflicts.

So forget the appeals process. It, as currently defined, has no
real enforcement power. Instead concentrate on how the issue would
be resolved in the courts -- nobody will pay any attention to the
appeals process until it gets some court backing. In the case of
trademarks the appeals process was written by trademark lawyers who
have a good idea of how the courts will react. No such work has been
done with general charters, or indeed any other kind of dispute

Kent Crispin, PAB Chair			"No reason to get excited",			the thief he kindly spoke...
PGP fingerprint:   B1 8B 72 ED 55 21 5E 44  61 F4 58 0F 72 10 65 55