Date:        Sun, 09 Jun 2002 19:49:34 -0400
    From:        Allison Mankin <[EMAIL PROTECTED]>
    Message-ID:  <[EMAIL PROTECTED]>

  | I would question whether this is well-understood and DNS servers
  | are ready to select which AAAA records to reply with depending
  | on the address the query was sent to.  Do current servers implement
  | this, including caching servers?

It is much worse than that - just imagine a DNS server running on a
site border router (which is much more likely in some environments than
others - but in those ones, it is quite plausible).   How such a server
figures out which particular site local AAAA records it should forward
into which site, even assuming that the DNS server is actually taught
about sites, is mind blowing.

This all could be kind of managed if Paul Francis' NUSLA proposal were
adopted (see the message I just sent) but without that, I thought that we
had pretty much concluded ages ago that putting site local addresses in
the DNS (other than for people using site locals for the purpose of
running a disconnected net, where those are the only addresses that exist)
was a total write off as an idea.   Sure it can be made to work in
particular cases with special care (just as sticking 1918 addresses in
the DNS can be coerced into working) - but it isn't something that anyone
rational would ever actually suggest as SOP.

What we need is a way to decide when to use site local addresses, and
to discover what the appropriate address is, that does not use the DNS.
I think that's entirely possible, and certainly don't believe that the
current lack of a specification for how to do it warrants discarding site
local from the architecture (though it probably warrants suggesting
extreme caution in their use for now).

kre

ps: Aside: note we have no process problem advancing the specs we have
with site local remaining in them - we have enough implementations of
site local, which work as much as the spec says they should work, which
isn't very much, for the specs to be able to advance as they are.

--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page:                      http://playground.sun.com/ipng
FTP archive:                      ftp://playground.sun.com/pub/ipng
Direct all administrative requests to [EMAIL PROTECTED]
--------------------------------------------------------------------

Reply via email to