-- 
Alex Brotman
Sr. Engineer, Anti-Abuse & Messaging Policy
Comcast
 

> -----Original Message-----
> From: Alessandro Vesely <[email protected]>
> Sent: Monday, January 13, 2025 1:58 PM
> To: Murray S. Kucherawy <[email protected]>
> Cc: [email protected]
> Subject: [dmarc-ietf] Re: I-D Action: draft-ietf-dmarc-aggregate-reporting-
> 25.txt
> 
> On Mon 13/Jan/2025 16:36:43 +0100 Murray S. Kucherawy wrote:
> > Would these be more helpful submitted as PRs?  Seems like we're
> > creating a middle step here that blocks on Alex re-typing all of this.
> 
> 
> I'm not sure what is going to be accepted.
> 
> 
> > On Mon, Jan 13, 2025 at 4:00 AM Alessandro Vesely <[email protected]>
> wrote:
> >
> >> In Section 2.1.1.4, the criterion for defining periods is still not
> >> defined.
> >> I'd re-propose what I wrote last December:
> >>
> >>      The start and end of reporting periods are chosen to obtain a 
> >> partition
> >>      of time, such that each instant of time belongs to exactly one period.
> >>      Empty reports must not be sent.
> >
> > This seems like a wordy way of saying "Reporting periods MUST NOT
> > overlap, and reports MUST NOT be empty."
> 
> 
> Yes, and also that they should form a continuum (as opposite to be the
> min/max time of the reported messages).

There was a PR for this, and not everyone agreed with your PR.  If you want to 
re-open that, please create a new thread on the list for discussion, pointing 
at your PR.

> 
> 
> >> In Section 2.1.1.11, *If validation is attempted for any DKIM signature,
> >> the results MUST be included* is an excessive requirement.
> >
> > Why?  Are you arguing for "SHOULD"?  If so, what's a situation where you
> > might not comply?
> 
> 
> Even SHOULD is a bit too strong.
> 
> Depending on the method used, there can be practical limits to the number of
> signatures one can report.  For example, I use a SQL statement that LEFT JOINs
> various queries, four of which are for DKIM signatures.  I could put five or
> six of them, thereby slightly lengthening the overall query; anyway, the
> number
> of signature (columns) in the result is fixed[*].  (LEFT JOIN provides for 
> NULL
> values if there are less signatures.)

This sounds like an implementation concern?  If someone else makes a design 
choice and can only retrieve aligned signatures, should we say that only those 
should be reported?

> 
> To know if a signature verifies, one has to /attempt/ validation.  Then, some
> signatures, especially failed ones, will be ordered after the max number of
> signatures.
> 
> 
> Best
> Ale
> --
> 
> [*]https://urldefense.com/v3/__https://dba.stackexchange.com/questions/2
> 89374/can-cte-simplify-repeated-and-possibly-recursive-
> joins__;!!CQl3mcHX2A!FAuTDmxkaMle6dueyCM-
> V7PiknoYC7f1wUcwesEykeEUu4GbYvSuMCsekXljn97FWHGpTwBTLehE-
> b7QR9ibriT0$
> 
> 
> 
> _______________________________________________
> 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