Richard, I am severely disappointed by your critique.

It misses the conclusion, which is that we have a missed opportunity
because SMTP state is not preserved in the Received header field.   This
can change by convincing the relatively small number of MTA software
providers to alter how they construct a Received header field.   It would
facilitate identifying the originating organization while also facilitating
identification of forwarding points.

But to your specific objections to the content:

An overly-broad SPF policy is a problem, but it can be corrected by local
policy that limits acceptable results to a smaller set of source servers.
  This is not much different than using local policy to correct a published
SPF policy from "IP:address" to "IP4:address" to avoid a PermError result.

It is technically true that DKIM has potential for uses other than
authentication, but authentication is the only standardized usage.
 Regardless of how a DKIM signature is being used, a verified signature is
only valid if it correctly represents the signing domain.   When the
signature is applied by a third party agent, the signature is only useful
if the agent has correctly determined that he has been properly authorized
to deploy his agency.   We have a problem with DKIM signing by outbound
gateways because we do not know if they are acting appropriately as
agent for a client, or acting foolishly because they have been duped by an
impostor.    ARC exposed this problem, because we routinely receive ARC
sets which indicate that a message was signed after failing SPF, DKIM, and
DMARC,   We hope that the outbound gateway authenticated the client by some
method other than SPF, DKIM, and DMARC, but we cannot know.   The gateway
service has no method for reporting a proprietary authentication scheme,
and the evaluator has no way to verify the assertion if it could be
provided.

There are other problems with DKIM.

   - Not every message has a DKIM signature.  When there is no signature,
   local policy can only authenticate based on SPF-equivalence, using server
   identity and RFC5321.MailFrom identity.
   - DKIM signatures are vulnerable to DNS timeout problems.   I have
   considered implementing local policy to trust only DKIM on certain
   dual-authenticated domains, but I have been reluctant to do so because a
   DNS timeout will cause a trusted and important message to become distrusted.
   - DKIM signatures can be invalidated by in-transit changes, and a local
   policy to trust an unverified signature will usually depend on trusting the
   SMTP identifier history.

In short, I can authenticate any wanted message using a local policy based
on a combination of server identity, SMTP identity, From identity (and
sometimes Reply-To identity), and this result can be used to distinguish
wanted messages from unwanted messages.   To avoid deception, I need this
information about the origination point and all forwarding points.
 Upgrading the Received header field is the best hope for ensuring that
this information is available when needed.   DKIM2 will add important
details, but believing that it will be a complete solution will require
believing that 100% deployment is achievable.   We need to broaden our
strategy.   ARC++ is not the solution, and DKIM2 does not take us far
enough.,

Doug Foster


On Sun, May 31, 2026 at 5:37 PM Richard Clayton <[email protected]>
wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> In message <[email protected]
> il.com>, Douglas Foster <[email protected]> writes
>
> >Email authentication seeks to establish these things:
> >
> >   - That the originating organization acted properly by transmitting a
> >   message on behalf of the original RFC5321.MailFrom domain.
>
> ..
>
> >The first of these objectives is the most important, because the other two
> >are meaningless unless the first objective has been accomplished.   It is
> >achieved when one of these criteria are satisfied:
> >
> >   - When the message produces SPF Pass, based on the originating
> >   organization's external IP address and the message's original
> >   RFC5321.MailFrom domain.
>
> That's not especially correct.  I regularly see unauthorised email with
> an SPF pass (and a corresponding DKIM pass) ... the common case being
> that people use an includes clause for the IP address space of an entire
> "cloud" and the "cloud" allows all and sundry to send any email that
> they wish ...
>
> >Yet the DKIM component of DMARC implicitly assumes
> >that local policy will trust a signature created by a downstream server as
> >equally valid with a signature created by the originating organization.
> To
> >my understanding, DKIM2 expects to preserve the assumption that a
> >signature's validity is independent of when it is created.
>
> that is correct ... the signature merely demonstrates that a system was
> in possession of the relevant private key and that the entire contents
> of the message was available to it (I suppose strictly, that it is in
> possession of the hash values for the header and body and is able to add
> a header field to the message).  You might like to note that DKIM2
> expresses things in this manner and avoids language suggesting claims of
> trust or ownership
>
> BTW: I don't see why it would matter to anyone that organisations
> regularly outsource the signing of their email. $DAYJOB$ outsources
> sending their daily newsletter to a third party who applies a DKIM1
> signature -- the public key is published under $DAYJOB$.net which is
> entirely under $DAYJOB$'s control and writing to that space is forbidden
> to anyone else.
>
> - --
> richard @ highwayman . com                       "Nothing seems the same
>                           Still you never see the change from day to day
>                                 And no-one notices the customs slip away"
>
> -----BEGIN PGP SIGNATURE-----
> Version: PGPsdk version 1.7.1
>
> iQA/AwUBahypsmHfC/FfW545EQL3FwCffbby9y3QgERPYnfoMr7UPw2HcQ4An39W
> R9UhodJIBmymNPX6L3fILxAI
> =L4Om
> -----END PGP SIGNATURE-----
>
> _______________________________________________
> dmarc mailing list -- [email protected]
> To unsubscribe send an email to [email protected]
>
_______________________________________________
dmarc mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to