[vip] Educational session on existing variant practices

[Andrew Sullivan]

Dear colleagues,

I think this is an excellent suggestion worthy of pursuit.  I have one
remark, however.

On Mon, Jul 04, 2011 at 11:04:07AM +0100, Vladimir Shadrunov wrote:

> If someone active in the IETF community would be willing to give a
> presentation on how existing DNS standards and related RFC handle string
> similarity that would also be great. Additionally some feedback from the IP
> lawyers would be appreciated as to whether the variant problem exists in the
> IP realm.

I can certainly provide an overview of how various DNS technologies
can be used to help with label-string similarity issues.  There are
some tricks one can use in the DNS that make some of these issues in
some ways easier, and in some ways trickier, to deal with.

But the short answer to the question as phrased above is much more
blunt, and I want to put it here on the list so that it is not lost.
The existing DNS standards and related RFCs do nothing at all about
string similarity.  DNS is an exact-match technology.  You send a
query for a QNAME, QCLASS, and QTYPE.  If there is an entry in the
authoritative name server you happen to talk to, or some in
intermediate cache, that matches _exactly_ your requested combination,
you get back an answer.  If there is the same name but something else
doesn't match, you get back an empty answer with no error (or,
sometimes, a redirection).  And if there is no such name in the
authoritative servers, you get back "NXDOMAIN" (that is, RCODE=3 Name
Error).

This is a deep and fundamental part of the DNS, and it is important we
not lose sight of it; it is not something we can change without
effectively replacing the DNS itself.  Anything that we cannot
ultimately simulate by adding to the number of exact matches in the
global DNS -- that is, making the DNS bigger -- is just not a policy
that can ever be deployed.

Best regards,

A

-- 
Andrew Sullivan
ajs at anvilwalrusden.com