Re: A6/DNAME usage guidelines and limits

[Paul Vixie <[e-mail]>]

> I think this is the way to go.  A6 is a good idea, in general.  However,
> having the resolver on a client have to perform multiple queries to get
> a single name resolved is not a good idea.  End-points in the DNS tree
> should be able to ask DNS server/cache systems to provide them an AAAA
> answer to a name.  The server/cache should be able to handle AAAA and
> A6, and convert A6 records to AAAA responses accordingly.  This gives
> the ability for "easy" address renumeration, and a thinner client
> resolver profile.  In short, it's the best of both worlds.

what the world is coming to is that applications (by which i include stub
resolvers) will have to have available to them a far simpler protocol/API
than full DNS.  is that what we all really want?  what it would look like
is "applications make TSIG'd AAAA and A queries of their local recursive
name server" and "recursive name servers will use EDNS0, DNSSEC, and A6."

that doesn't cover the DNAME/PTR issue but that falls out naturally.

that doesn't cover the synthesized AAAA issue but that also falls out
naturally.

the question is, before we go into detail on it, is: have we really and
truly put so much mass into the DNS protocol that most applications (and
most stub resolvers) cannot reasonably be expected to speak it?

once we know that, we know how to proceed.

--
to unsubscribe send a message to [e-mail] with
the word 'unsubscribe' in a single line as the message text body.
archive: <http://ops.ietf.org/lists/namedroppers/>