Re: .COM Clusters are Not RSCs

[Karl Auerbach <karl at CaveBear.com>]

Keith -- you have are simply too sensible. ;-)

I find your analysis to be highly credible. 

Since this is really a domain name policy issue and not necessarily of
interest to all IETF folks, I am only responding here to:

 a) agree with you

 b) attempt to clarify some of my prior statements regarding a 
    multiplicity of roots.

> > Since when does the IANA or IETF have a right to tell a nameserver operator
> > what they can load in their cache hints file?
> 
> I don't know of anybody who claims the right to control such things.  
> But we all lose if the DNS is fragmented.

I agree with Steve Bellovin.  I, too, will be really ticked off when my
e-mail starts bouncing because of name system fragmentation.  And I don't
want to re-live the pain of usenet style addressing. 

I don't mind if the server *computers* are fragmented.

However, I would very strongly object if the *data* contained in those
servers were fragmented or divergent. 

This is why I support RSCs only with the limitation that they all become
equivalent.

I recognize that today they are *not* equivalent, that they overlap only
with respect to the TLDs offered by the most conservative of the RSCs, our
well-known [a-k]-root-server.net group. 

I'm trusting (perhaps naively) that this equivalence will come about
either through a process of RSC's each trying to out-trump one another
through completeness, or through withering of the TLDs that are less than
universally reachable. 

This would be a process that engenders some pain.  But it has the
characteristic of being something that happens without centralized control
or imposition of some pre-conceived policy of rightness or wrongness.

As for TLDs being fragmented -- That should be anathema.

I would hope that any TLD that allows itself to be subsetted finds itself
quickly without subscribers and quickly on the way to the great TLD farm
in the sky.

But again (and again, perhaps naively), I believe that no coercive process
need be applied to make this happen.  I think the natural sanity of domain
name holders will cause any TLD operator who tries to fragment to quickly
reconsider the error of its ways. 

(I write this with a small voice in my head saying, "Given the enormous
power of extremely well established TLD operators like NSI with respect to
.com, is this a rational thing to believe?"  And another voice says
"Couldn't the largest RSC demand concessions of other RSC's and the TLD 
operator for accepting a new TLD?"  I do not have good answers to these
questions.) 

> Lots of people are talking about various ways of splitting the root.
> You must have missed it.

Multiple root confederations are, to me, OK, but **ONLY** if the following
conditions hold: 

   - Each RSC leads me to a consistent, equivalent version of a given TLD

   - Each RSC contain all "valid" TLDs (I'm assuming that there will
     always be some wannabe TLDs trying to estalish a presence and that
     there will be periods of less than universal coverage for those.[*])

[*] I realize that some RSC's are being used as a means by which some
people are trying to leverage new TLDs into existance in order to reap
registration fees.

However some of these are pursuing their goals by defaming and using legal
means to harass some of our most honored and constructive net citizens.

I abhor such tactics and condemn those using them to the unknown reaches
of /dev/null.

If an RSC wishes to promote a new TLD, it should be free to do so.  But it
should be aware that the new TLD may readily fail should other RSC's not
elect to carry a reference to that TLD.  Customers of that new TLD should
be made aware of that risk (do we need a TLD "prospectus" mechanism?). 

(As I type I just realized that there is a counter-problem: a big RSC
could obtain improper control over an established TLD operator by
threatening to drop that TLD.  Of course that sitation obtains with even
more force today with the "official" single RSC, [a-k]-root-servers.net.) 

[As an aside, I agree with you that having too many TLDs is a technical
problem -- it flattens the DNS hierarchy, defeating scalability, and thus
causing increased root traffic and, even worse, potentially causing cache
overflow and subsequent server thrashing leading to a potential *plosion
of root traffic (I'm not sure whether it's an "im-plosion" or an
"ex-plosion" ;-)  ]

I do not have a good feeling what constitutes "too many" TLDs.

		--karl--