Re: .COM Clusters are Not RSCs

[Karl Auerbach <karl at CaveBear.com>]

> In my opinion, "root server clusters" -- and dressed up in any name
> at all -- would be the single worst thing that could happen to the
> Net.  It would completely destroy the global namespace of the
> Internet.

The RSC's just point to the normal TLD (Top Level Domain, i.e. .com, .org
...) servers. (There's only a problem of ambiguity where the is a dispute
over who runs a TLD, as in the case of .web.) 
 
> It would no longer be possible for me to advertise my mail server
> as being research.att.com.

No, you could still have the same, unambiguous name for the mail server.
The only thing that would happen is that when someone's DNS resolver says
"hmmm, where can I find a server for .com?" the packet goes to an RSC
rather than to the existing [a-h]-root-servers.net machines.  The response
packet will point to the real and true set of servers that know the
contents of .com.

So the server that figures out what "att" means in the context of .com is
the same in either case.

There is a separate concept of breaking up the TLD servers in a geographic
sense -- replicating the content -- so that one can select a set of root
servers which will give you references to close by TLD servers.  This
could risk some ambiguity if the replication is less than perfect.

I've been running using the AURSC as a root for some weeks now on a server
that provides name resolution for a couple of small, but network-active,
companies and have not noticed a single glitch or ambiguity (other than I
can resolve some additional TLDs.)   And response time isn't a problem
since my server has long since cached all the TLD pointers.  (Nor was it
ever a problem since my path to Austrailia from Santa Cruz isn't much
longer than it is to the east coast of the US.)

Before that I tried out the "Grass Roots" tool to generate my own root
zone files and essentially run with me as my own root.  This also ran
quite well, but required me to periodically update my root zone to reflect
changes.

All this is to say that, yes, there are risks of ambiguity, but not due to
RSC.  The risk comes when there are disputed TLDs.

		--karl--