[dmarc-ietf] An alternative proposal to the known intermediary problem

[Dotzero <[e-mail]>]

After reading Dave Crocker's proposal, which I consider fundamentally
broken, I've given some thought to the issues.

1) Many MLM's do not want to change their current practices yet suffer
consequences from DMARC implementations at their subscribers domains.

2) There is a mismatch between the granularity (domain) which DMARC
operates at and individual subscriptions to mailing lists.

3) It would be useful for domains with individual users to show
authorization for messages by individuals that are likely to flow through
intermediaries such as MLMs.

Without formulating all the details, what I propose for the groups
consideration is a new field which I will call "Intermediary" as a stalking
horse. This field would be DKIM signed separately from the current DKIM
signing. This would ensure more robust survivability in passing through the
intermediary. There would likely need to be some additional elements signed
to mitigate against replay attacks using the second signature.

Variations on this approach might include:

A) Generic signing of "Intermediary". This seems as risky or perhaps more
risky to me as signing using the "*l*=" tag. Presumably generic signing
should not be used in conjunction with reject.

B) Specifying the specific Intermediary in the Intermediary Field. This
would indicate that the users domain recognizes that the user uses the
intermediary and by policy exempts this use even though it breaks both DKIM
and SPF validation. The receiving domain would need to recognize some
potential risk of malicious modifications or additions to the message.

Benefits of this proposal include:

1) This is an addition to DMARC that does not change the underlying
functionality of DMARC for those domains not interested in participating in
this extended functionality.

2) It resolves the mismatch in granularity between the domain level and the
user level as far as authentication/authorization.

3) It would require no effort on the part of known intermediaries other
than not breaking the second signature.

Issues:

1) It increases administrative complexity for the originating domain in
that it requires the domain to either track user:intermediary relationships
or enable users to self approve such relationships ad hoc.

2) A more detailed specification would need to be fleshed out to determine
whether this approach would survive likely modifications by known
intermediaries.

3) This does not address the issue of modifications by intermediaries not
identified in advance and authorized.

Feel free to start pulling this apart. This is a proposal for an approach,
not a complete proposed solution.

Michael Hammer