On Mon, Jul 10, 2017 at 2:46 PM, Gene Shuman <[email protected]> wrote:

> (resending from my personal account to avoid the spam filter)
>
> I think there's still something missing from the draft wrt fail/invalid.
> In section 5.2.2, it says that gross violations MUST be capped in the
> manner specified. This seems to only encompass what we were previously
> considering cv=invalid.  Does it say somewhere that cv=fail should be
> handled in this fashion as well?  Or does what was previously cv=fail still
> sign all arc sets?  Are we handling these differently?  I'm not necessarily
> sure we should.  Also, the language here is MUST.  Shouldn't this be
> optional, as we'd discussed?
>

I still think (and find it compatible with the tenor of other
conversational threads) that MUST is appropriate. As usual, people can
choose to not follow the spec :-(

To your other points:

   1. Section 6.1 para 2 says:  "Until the chain is determined to be
   failed, and marked with an ARC set bearing the "cv=fail" indication, each
   participating ADMD SHOULD apply their own seal."
   2. Section 5.2.2 para 2 says: "Because the violations can not be readily
   enumerated, the header fields signed by the AS header field in the case of
   a major violation MUST be only the matching 'i=' instance headers created
   by the MTA which detected the malformed chain, as if this newest ARC set
   was the only set present."
   I see now that I did not properly expand the scope of that directive to
   cover all cases where a cv=fail verdict is determined to be necessary, but
   I think that is the right approach. cv=fail --> sign only the current
   (latest) ARC set (regardless of the failure reason)

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

Reply via email to