At 02:59 PM 8/10/2004 -0700, Rand Wacker wrote:
I believe that the working group made no objection to sub-domaining
records (publishing TXT in _marid.example.com), so if we go down the path
of changing the version number or even the format we should take that in
to consideration.
I remember that one problem with _marid.example.com was that one lost wildcard
behavior; it isn't possible to have _marid.*.example.com to resolve requests
for sub.example.com. But it was less clear to me whether one could publish a
record for *.example.com and have it work "as desired" for messages from
sub.example.com.
I have been a bit concerned about the loss of an outgoing counterpart to
wildcard MX record behavior. In other words, if you can have wildcard MX
records to accept mail for *.example.com, you should be able to have something
that indicates the outgoing mail servers for *.example.com. This might be an
area where a new RR has an advantage over _marid TXT records.
Short of wildcarding, here are a couple of options:
1. The recipient tries the domain in the PRA, and if that fails starts walking
up the tree (first sub.example.com, then example.com, then (hopefully not) com)
looking for a MARID record.
2. The originator includes a message header indicating the appropriate level of
abstraction for the MARID record. So a message from
jdoe(_at_)sub(_dot_)example(_dot_)com might have a header like "MARID-Domain:
example.com". I suspect this runs counter to the MARID approach of not
changing the 2822 headers. In the MASS world, Identified Internet Mail and
DomainKeys both include this with the signature in order to find their keys or
key authorization data.
Do either of these seem workable? I'm concerned that the wildcards won't do
the Right Thing.
-Jim