--On Saturday, 31 March, 2007 06:50 -0400 Dave Crocker
Tony Finch wrote:
This is a contradiction. The fix, in line with current
practice, is to downgrade the MUST NOT to a SHOULD NOT - or
delete it altogether.
Given the complexities, controversies, and resulting
instability of community consensus about abuse-related reasons
for rejecting mail, anything that touches that space should be
omitted from the document.
Drop the text.
For the record, as editor, I'll do whatever consensus emerges
that I should do.
However... The specific name != address prohibition originated,
I believe, with 1123 after significant discussion. It was
discussed again in DRUMS and reaffirmed. The conclusion both
times, IIR, was that this particular test could lead to a lot of
trouble given special cases and sloppy DNS configurations.
Tests involving "uses my domain name but isn't one of my hosts"
or "uses internal address, but comes from outside" have never
been prohibited by that restriction. I'm a little reluctant to
see the restriction changed on the basis of these sort of
discussion, and more reluctant to see the text dropped entirely.
The general observation that appears to contradict it was
inserted during the DRUMS effort to clarify that, as far as
rejecting or refusing to accept mail traffic was concerned, it
was acceptable to have an MTA configured to be much more
restrictive than limiting restrictions specifically identified
in the main body of text. That had been obvious to everyone for
years, but it seemed wise (for reasons I no longer remember) to
be explicit about it while DRUMS was being developed.
(1) The minimum-change way to eliminate the apparent
contradiction would be to insert an "unless specified elsewhere
in this document" phrase in the general "can reject for any
reason" text. My personal opinion, given this discussion, is
that this change should be made regardless of what else is done.
(2) The second-most-minimal change would be to change the MUST
NOT to SHOULD NOT in the specific name != address case. My
personal preference at the moment is to not do this, but I don't
feel strongly about it and am easily persuaded (or outvoted).
(3) Dropping the text entirely seems to me like overkill given
the history of decisions to put that text in and the observation
that I've yet to hear anyone make a strong case that mentioning
that the name != address test doesn't work particularly well is
a bad idea.