We need to remember the two uses of NP.

Non-existent organizations are inherently untrusted.   A knowledgeable
administrator should block them whenever they are detected, with or without
a DMARC policy, with or without an NP term in a PSD policy.   There seem to
be political problems with saying as much in this document, but it is the
right action.  Feedback reporting will only occur if there is a PSD policy,
and every PSD policy SHOULD include an NP=reject term.   If the PSD policy
lacks an NP term, and NXDomain is the reason a message is blocked, then the
disposition reason is local policy.   Perhaps we need a local policy code
for "blocked because of NXDomain without an NP clause", but this situation
should be rare.

A non-existent subdomain is not a problem unless the domain owner says that
it is a problem in the organization's DMARC policy.  Non-existent
subdomains are used on a daily basis for legitimate mail.  Normal DMARC
rules allow non-existent subdomains to be authenticated by parent or
sibling domains.  We should not be changing this.   If an administrator
chooses to block for non-existent subdomain without domain owner direction,
this would also be a local policy disposition.

I am intrigued by the idea of using inbound A-R data to provide a permanent
record that could be used to validate outbound reporting data.   But are
changes really needed to do so, and would they be changes to our document
or changes to A-R?

Doug Foster


On Fri, Dec 30, 2022 at 9:03 AM Alessandro Vesely <[email protected]> wrote:

> On Wed 28/Dec/2022 17:39:34 +0100 Scott Kitterman wrote:
> > On Wednesday, December 28, 2022 11:19:46 AM EST Alessandro Vesely wrote:
> >> Appendix A5 in the I-D describes "Issues with ADSP in Operation".  This
> >> appendix existed in RFC 7489 (March 2015), when ADSP was already set to
> >> Historic (November 2013).
> >>
> >> Bullet #2 in that appendix says "Nonexistent subdomains are explicitly
> out
> >> of scope in ADSP."  DNARC, instead has an apposite np= policy.
> >>
> >> However, in Authentication-Results one can write dkim-adsp=nxdomain.  I
> >> found no equivalent result for dmarc=.  Shouldn't there be one?
> >
> > nxdomain isn't a DMARC result, so from that perspective, no.
>
>
> Why not add it?  We'd write dmarc=temperror on a DNS hiccup, correct?  The
> result of DNS lookup strongly affects DMARC results.
>
> It is important to know whether the From: domain exists.  The only
> standard way
> to report it is to write dkim-adsp=nxdomain, which is a nuisance.
>
>
> > I think a better question is should the A-R header field indicate which
> tag was
> > used for policy determination (p=, sp=, np=).  I think the answer is,
> again,
> > no.
>
>
> The draft has a polrec.p= tag to report the policy found.  It is tricky to
> tweak that into, say, polrec.np=.  Would it mean that the polrec.domain
> had an
> np= tag, or that the receiver got nxdomain and hence determined to use an
> np=
> policy, irrespective of what was actually written in the policy record?
>
>
> > Is that currently captured in aggregate reporting?  If we're going to
> indicate
> > it anywhere, I think that's the right place.
>
>
> I think having a strict match between A-R lines and aggregate reports
> would be
> a good thing.  To wit, one could debug emitted reports by looking at A-Rs.
>
>
>
> Best
> Ale
> --
>
>
>
>
>
>
> _______________________________________________
> dmarc mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/dmarc
>
_______________________________________________
dmarc mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dmarc

Reply via email to