[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]


To: "David R. Conrad" <david.conrad@nominum.com>
cc: "D. J. Bernstein" <djb@cr.yp.to>, ngtrans@sunroof.eng.sun.com, namedroppers@ops.ietf.org, ipng@sunroof.eng.sun.com, dnsop@cafax.se
From: Keith Moore <moore@cs.utk.edu>
Date: Tue, 07 Aug 2001 21:30:57 -0400
In-reply-to: Your message of "Tue, 07 Aug 2001 17:31:59 PDT." <5.0.2.1.2.20010807172339.0301adb0@localhost>
Sender: owner-dnsop@cafax.se
Subject: Re: (ngtrans) Joint DNSEXT & NGTRANS summary

> P.S. Sorry if this is considered heretical.

I don't think it's heretical at all.  I don't take it as an absolute
that people would inherently rather renumber than use NAT, but I think
it's reasonable to assume that large numbers of people will follow the
path of least resistance, or perhaps, of least apparent risk.  If at 
the time the decision is made, NAT looks easier than renumbering, 
that's what they'll do.

So if we want to avoid NAT in IPv6, we need to make renumbering *much*
easier than it is now.

Problem is, A6 in its current form doesn't really do this.  Even if it 
works as advertised, A6 only addresses one of many aspects of renumbering.  
And DNS is almost certainly not the mechanism that you want to use to 
advertise address changes in your local network.  And people have a 
difficult enough time configuring SOA, NS, A, CNAME, PTR and MX records 
correctly, without having to also sort out A6 and DNAME.  To me it 
appears that A6 could actually increase the difficulty of renumbering
without disruption.

OTOH, I can see how A6 (or some subset thereof) might end up being part of
an overall renumbering strategy.  For instance, I can see how a DNS-like 
identifier might be assigned to name the set of global address prefixes 
by which a given network might be reached, and how a DNS server for 
a domain might want to act as a cache for those name-to-prefix bindings,
and how those bindings might be suitably represented by A6 records.
I can even see how it might be desirable to expose those A6 records
outside of your local net if you are signing your records with DNSSEC.
I can also see how it would be desirable to just sign the AAAA records.

The presumption behind A6 seems to be that the party who signs your
prefix record is likely to be different than the party who signs the
suffixes.  I don't buy that at all, because I believe that renumberings,
no matter how easy they become, are still going to need to be explicitly 
managed events.  You don't want your ISP asserting that you have a new
prefix before your firewalls, routers, hosts, and applications know how 
to deal with it.  Re-signing the records in a zone is the least of the 
problems that must be solved in order to do the renumbering.

Keith

Home | Date list | Subject list