The criteria for a good naming system

[Brad Templeton <brad@alto.clari.net>]

Reviewing the matter more, I offer what I think are the criteria for a
good DNS

1) Memorability and typability.  In the end, that's what people want
(along with the motherhood issues of efficiency.)  They want to be able
to tell people a name so that they can remember it, and they want it to
be easy to type in when you have to type it in.   Even when it appears in
a URL often somebody had to remember it and type it in.

Any scheme involving random strings or long convoluted names to meet other
goals fails this most important goal.  Of course we want names to be as
simple as possible, but no simpler -- and thus the other goals.

2) Efficiency of operation.  Names should be dereferenced quickly and
reliably.  Systems must be in place to fund this goal.  This goal
suggests that business motives should be involved rather than the
style of a public trust.

3) Ease of administration.   This seems to have been the IAHC's main issue,
but truth is people doing business under a name would rather meet goal #1
at the cost of more administrative hassle.   This goal has two components:
	a) Efficient and customer responsive operation of registries,
	   and the biggie
	b) Resolution of conflicts.

Things like a 60 day waiting period detract greatly from (a) to aid (b).

4) Guessability of names.  People want this because it was possible in
the days of a smaller net, and is still possible for those with unique
names in the outside world, but it is not an easily attained goal for
those without unique names.  Memorability is better.

Instead, the issue of guessability should be supported through directory
services.  An efficient white pages protocol should be developed, perhaps
served by DNS like methods so that programs where people type in names
(web browsers, mailers, html editors, ftp and other clients) can make
quick efficient queries.  A user should be able to type in
"ABC" to her mailer and get a quick menu of the Australian and American
Broadcasting Companies and other ABC businesses and people on the net, and
then pick one or refine the query.   A better, faster, more distributed
whois.  (The service might also know to offer local matches before
distant ones, bigger firms before smaller, etc.  Hell, companies might
buy placement in the order)  There are problems with doing it offline,
but a service like this might remove much of the conflict over "premium,
guessable" names.

There will always be disputes over trademarks but they will become very
similar to they way they were before the net.  The net has given people
the illusion that they "must" have a domain name of <trademark>.com because
it can be guessed (and thus is supremely memorable.)  It has made them
forget that most trademarks don't exist outside of context.

Sure, you need to stop people from registering names just to sell them,
but there are better ways to do it than delays and the other proposals.
It is not necessary to prevent misuse with the system, all that is necessary
is that misuse have some negative consequences.  A pledge of good intent
or not-for-resale in the application is sufficient.

I would propose:
	a) gTLDs for domain registries that apply, however the names
	must not have a unique meaning with no equivalently powerful
	synonym.  Ie. do not allow "www", "corp", "inc", "ltd",
	or any gTLD where the name is "the ideal choice" for the
	proposed registry.  Names should be new strings that the owners
	plan to *make* meaningful.  Ie. D&B might well want to make
	a corporate registry, but let then call it ".Dun" after themselves,
	not ".corp".  gTLDs for registries only, however, not for
	private use, to keep the number down.

	b) For the gTLDs that have a meaning, let those be a public trust
	if you wish, with multiple registries. 

	c) 2LDs as the registries want to do them, with the main rule being
	that applicants should submit a statement of proper intent.
	But alternately, just leave this to the gTLD registries.  If they
	have to bear the problems they will quickly gravitate to solutions.
	gTLD registries can easily stop sale of domain names on their
	own, now that they know the problems.  It is not a TM violation
	to have an entry in a directory, from what I know, it's a violation
	to do business under a TM and confuse the public.  NSI got
	blindsided by some speculators but the truth is that now that we
	know the lessons of that, the problem is not that hard to solve.