On Tue, Feb 7, 2017 at 7:22 PM, Brandon Long <[email protected]> wrote:
>
> On Mon, Jan 30, 2017 at 2:59 PM, Kurt Andersen <[email protected]>
> wrote:
>
>> On Mon, Jan 30, 2017 at 10:24 AM, Gene Shuman <[email protected]> wrote:
>>
>>> Extricating this discussion from the one about key sizes.  We left off
>>> at discussing whether we should implement something like a cv=invalid
>>> for arc chains that are no longer well defined.
>>>
>>> Brandon, I think you had the last response, and suggested that
>>> receivers just refuse to continue to process invalid chains?  This
>>> seems plausible, and might be simpler.  I have no strong opinions one
>>> way or the other, although I think I slightly prefer the cv=invalid
>>> state, as it strictly defines what receivers are to do with invalid
>>> chains.
>>>
>>> Regardless, I still stand by the fact that section 5.2.1 needs to go &
>>> that we should probably explicitly codify what constitutes an invalid
>>> arc chain.
>>>
>>
>> I don't think that 5.2.1 needs to be deleted, just significantly
>> refactored (which I plan to do next weekend).
>>
>> I think that having a participating receiver mark an chain as "broken
>> beyond repair" (aka = invalid) is a beneficial terminal state beyond which
>> no other mediators should perform any further ARC machinations. It's
>> certainly possible that a malicious mediator could essentially break every
>> ARC chain that it sees, but that's no different than what can happen today
>> and also is a situation that we are not trying to solve.
>>
>
> One challenge is how to specify that a chain is cv=invalid.  It implies
> being able to sign an invalid chain.
>

What is expected workflow in this case?

Say I'm alumni-forwarder.edu, and I receive a message with an ARC set (and
for the sake of this example, these are forgeries). After checking the
signitures, I consider the ARC set to be invalid. Section 6.4.2 suggests I
inform the sender of a message integrity failure. However, presume (given
no other abuse or spam signals) I want to forward all the mail receive on
and ARC failure alone is an insufficent reason to stop my forwarding.

How do I proceed? Stip the existing ARC set and start over with i=1 ?  Set
cv=fail?

A downstream receiver still might want to trust the
[ARC-]Authentication-Results
appended by alumni-forwarder.edu, but I think currently would be advised
against it in the case of cv=fail.

- Jacob
_______________________________________________
dmarc mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dmarc

Reply via email to