Re: [wg-c] http://www.cavebear.com/cavebear/growl/issue_2.htm#multiple_roots

[Karl Auerbach <karl@CaveBear.com>]

> At 08:08 PM 8/30/99 , Karl Auerbach wrote:
> >The value of this paper is that it gives this Working Group an answer to
> >its current stasis.  The paper indicates a means that this WG can use to
> >avoid having answering the hard question of whether to add new TLDs or
> >not, and if so, which ones and when?
 
> > http://www.cavebear.com/cavebear/growl/issue_2.htm#multiple_roots
 
> With your approach, how are ambiguities (double assignments) and 
> non-overlaps (partitioning) avoided?

This is described in the paper.

The operator of a root system selects which TLD's to put into his/her
"inventory" of TLDs.

If there are two (or more) possible versions of a TLD, then the operator
makes a choice based on whatever criteria the operator wishes to use. As a
practical matter, TLDs which are subject to dispute are not very saleable
items for a root system operator to put into its inventory because they
will lead to troublesome resolutions.

Thus, for example, root system operators will probably put into their
inventory all the standard gTLDs and ccTLDs with which we are familiar
today.  They may also include some of the new ones.  However, to take a
case with which we are all somewhat familiar, few root system operators
may incorporate .web until its status becomes more clear.

Thus, the root system operators will tend to take each TLD as an
unfragmented chunk, i.e. *all* of .com, *all* of .nu, etc.

So the main unit of problems, if they were to arise, would be occur on the
basis of entire TLDs either working or not working rather than on the
basis of individual names within a TLD.

If the root operators clients find that the root service isn't leading to
the results they want, they will take their custom elsewhere.  That
potential loss of business will nudge root system operators to avoid
making decisions that will create name resolution paths that don't lead to
where people expect them to lead.

As for the contents of the TLD's themselves, that's up to the TLD
operators (registrars/registries), just as it is today.

We can see this kind of behaviour in the non-ICANN root systems that exist
today.  All of them of which I am aware include the standard gTLDs and
ccTLDs and work exactly as expected.  They add new TLDs in hopes of
promoting those TLDs or because some TLD operator has paid to have the TLD
put into the root system.

Anyway, the notion is that the system will tend towards convergence
because the self-interest of all the parties will be to improve
convergence.  This will be emphasised by fact that it is pretty easy,
inexpensive, and painless for users (or their ISP's or corporate DNS
administrators acting in their behalf) to switch to a new root server
system if the one they are using doesn't give them the results they want.

I don't want to imply that all of these decisions to include or exclude
TLDs won't be the subject of financial quid-pro-quos.  But there's nothing
wrong with letting a little free-market economics into the DNS.  It
certainly beats us spending years arguing with one another. ;-)

Thus, to answer your intitial question, there can be cases where depending
on which root system one uses a name may or may not resolve to what one
expects.  But the self-interest of all involved will drive operators to
prevent that kind of thing from happening to any significant degree.

		--karl--