On Fri, Apr 10, 2009 at 09:53:46AM -0400, Edward Lewis wrote:
> But the draft really isn't about DNS. It's about SMTP.
I'd say this is not either/or, it's the use of DNS in another protocol.
It has to be addressed in a cross-wg or cross-area effort. Developing
a DNS view here is fine.
/* hat off */
The "solution" at hand has protocol ("additional section processing") and
operational implications but also touches upon general protocol design,
in particular how to interpret the absence of an RR or - from another angle -
how to specify absense of a service. Someone already quoted
<http://www.ops.ietf.org/lists/namedroppers/namedroppers.2005/msg00944.html>.
Fallback for RR types (try MX, then try A) looks attractive during the "phase
in"
but increasingly bad later, because of exactly the problem that people are
facing with MX today: exactly that fallback makes it hard to tell whether
there should be _no_ mail service for that domain name. Unfortunately
during the approval of RFC 5321, the explicit addition of AAAA fallback did
not really improve the situation.
The clean solution would involve some measurement regarding the volume of
non-spam (yeah, rathole) that is delivered through A-without-MX and some
willingness to move away from the fallback. Of course, one might also
believe that redefining part of the ancient MX RR is much easier, especially
if you don't have to carry the DNS operational burden.
-Peter
_______________________________________________
DNSOP mailing list
DNSOP@ietf.org
https://www.ietf.org/mailman/listinfo/dnsop
