<sarcasm>
gee, I don't know... if current practice is to require that senders have
A or MX records, maybe we ought to insist on it in the standard. I
mean, isn't the purpose of the standard to dictate current practice?
and isn't this the sort of change that should cause a recycle to proposed?
</sarcasm>
seriously, why is this the sort of bug that should dictate a change to
the standard, whereas the problems associated with fallback to address
records are the sort of bug that should be encouraged by the standard?
Keith
Arnt Gulbrandsen wrote:
I agree that the wording in -10 reflects the rough consensus.
However, I think 2821bis needs a little extra text to notify operators
that they may need to upgrade or reconfigure to comply with 2821bis. For
example bs.jck.com currently refuses mail from AAAA-only addresses:
$ vrfy -vv -S test(_at_)ipv6(_dot_)l(_dot_)google(_dot_)com
jklensin(_at_)jck(_dot_)com
rcpt 'jklensin(_at_)jck(_dot_)com' at 'bs.jck.com'
connecting to bs.jck.com (209.187.148.211) port 25
<<< 220 bs.jck.com ESMTP Exim 4.34 Tue, 15 Apr 2008 12:39:35 -0400
EHLO libertango.oryx.com
<<< 250-bs.jck.com Hello libertango.oryx.com [195.30.37.9]
<<< 250-SIZE 52428800
<<< 250-8BITMIME
<<< 250-PIPELINING
<<< 250-AUTH CRAM-MD5
<<< 250 HELP
MAIL From:<test(_at_)ipv6(_dot_)l(_dot_)google(_dot_)com>
<<< 250 OK
RCPT To:<jklensin(_at_)jck(_dot_)com>
<<< 550-Verification failed for <test(_at_)ipv6(_dot_)l(_dot_)google(_dot_)com>
...
ipv6.l.google.com has only an AAAA RR. The same test works when I use
test(_at_)www(_dot_)l(_dot_)google(_dot_)com as sender address.
(I tried some other recent posters. Some accept mail from test(_at_)ipv6,
including Dave, some others don't, including you and me.)
Exim users can comply by disabling the sender_verify variable, postfix
users can disable reject_unknown_sender_domain, I don't know the
relevant variable for any other MTAs.
Arnt