On 09/01/2013 08:29 PM, Steven M Jones wrote:
> On 08/31/2013 Raman Gupta wrote:
>> No, the hostname of the mail server is not a subdomain of the From
>> domain. But the mail server is a valid SPF sender for that domain.
> 
> This makes it sound as if you have a host in domain bar.com,
> sending/receiving mail for the domain foo.com. The host is duly
> authorized in the SPF record for foo.com.

Correct.

> However if the host in bar.com generates a message with a null
> reverse-path and an RFC5322.From in foo.com, the receiving system will
> build an SPF identifier of [email protected] and check that against
> the SPF record for foo.com. That will pass.

Correct, that is indeed passing.

> As was mentioned, DMARC requires an extra "alignment" between these
> identifiers. It gets an SPF pass using the identifier bar.com, but an
> RFC5322.From of foo.com - those simply will not match, and in the
> absence of a DKIM pass (with identifier alignment), you'll get a DMARC
> fail.

These messages are DKIM signed, and therefore did pass DMARC. It was
simply my perfectionist nature wanting the DMARC SPF to pass as well.

> On 09/01/2013 Raman Gupta wrote:
>> But this would imply that my mail server should be configured to send
>> "HELO mydomain.com" instead of "HELO mail.mydomain.com" where "mail"
>> is the hostname of the mail server.
> 
> Won't help if the situation is what I described above with this server
> generating messages for the other domain(s). But it's possible I've
> gotten that all wrong - please confirm.

You've gotten it right. I now understand the alignment issue is with
the sender domain vs the RFC5322.From domain. I should have seen that
before.

>> It seems to me the DMARC spec should take into account that, for empty
>> Return-Path's, the HELO/EHLO may be (and probably will be) a fully
>> qualified hostname, not a domain, and adjust the rules for SPF
>> identity alignment matching accordingly.
> 
> DMARC allows for FQDN vs. DN - unless you override the default and add
> "aspf=s" to your DMARC policy.

I'm in relaxed mode. But based on the information above, I now realize
the subdomain was not the issue with alignment, and my statement was
in error.

> I think you may be referring to a cross-domain authorization issue to
> handle messages generated with a null reverse-path on a host not in
> the same domain as the RFC5322.From.

Correct.

> DMARC is layered atop SPF, so it doesn't go poking around in SPF
> records. Because of the problems DMARC is trying to solve, it
> imposes additional alignment checks that SPF doesn't, after SPF has
> rendered its verdict.

That makes sense, usually.

> IMO, DMARC is not designed for cases where Sieve scripts are
> generating messages with a null reverse-path on a host in a different
> domain. That's a design issue vs. an operational issue, and should be
> hashed out on the IETF dmarc list
> (https://www.ietf.org/mailman/listinfo/dmarc
> <https://www.ietf.org/mailman/listinfo/dmarc>). Feel free to put
> forward a use case there.

Yes, I think this is a corner case in which DMARC should not impose
the additional SPF alignment check. But given that DKIM still results
in a DMARC pass, this could just be my perfectionist nature asserting
itself. Regardless, I probably will post this use case to the IETF
dmarc list.

> If I've misunderstood or mis-characterized the situation you're
> dealing with, my apologies. In that case it might be good to describe
> the scenario step-by-step as a message is received and a response
> generated by Sieve, etc.

I think you've understood it, and clarified where I was wrong in my
thinking, as well as validated my original belief that the DMARC
design does not handle this situation "perfectly". Thanks.

Regards,
Raman
_______________________________________________
dmarc-discuss mailing list
[email protected]
http://www.dmarc.org/mailman/listinfo/dmarc-discuss

NOTE: Participating in this list means you agree to the DMARC Note Well terms 
(http://www.dmarc.org/note_well.html)

Reply via email to