If early reject is not a problem, then perhaps this paragraph is not needed
at all.  If it is present, it needs to communicate correctly to everyone,
including the new cybersecurity student.   I see nothing in the language
which limits the warning to this one special case.   I read it as a broad
statement that any policy containing "-all": is likely to produce unwanted
behavior by some evaluators.  That broad statement is wrong, and is
particularly misleading when applied to DMARC.

I don't think the special case of a policy containing only "-ALL" needs any
comment in our document.  I also don't think we can say very much about it
without creating a layer violation, since the special case is not part of
RFC 7208.

Doug


On Fri, Feb 11, 2022 at 8:03 AM Dotzero <[email protected]> wrote:

>
>
> On Fri, Feb 11, 2022 at 7:19 AM Alessandro Vesely <[email protected]> wrote:
>
>> On Fri 11/Feb/2022 08:57:17 +0100 Douglas Foster wrote:
>> > This section implies that publishing SPF -ALL is a risky move, which is
>> made
>> > worse by DMARC.SPF -ALL is a only risk when (a) the message is
>> forwarded
>> > without MAILFROM rewrite and (b) the evaluator does not implement DMARC.
>>
>>
>> My reading of Section 7.1 differs.  I see no risks implied in its wording.
>>
>>
>> > Proposed language for the second paragraph:
>> >
>> >     “By design, DMARC allows a verified and aligned DKIM signature to
>> override
>> >     an unfavorable SPF result, including FAIL.    However, the full
>> message,
>> >     including the DATA section, must be accepted before DMARC
>> participation can
>> >     be determined and DKIM signatures can be evaluated. Consequently,
>> DMARC
>> >     evaluators SHOULD NOT use SPF results to reject a message prior to
>> receipt
>> >     of the entire DATA section.”
>>
>>
>> I like better the current wording.
>>
>>
>> > When this was previously proposed, it was noted that some DMARC
>> evaluators
>> > consider the combination of SPF FAIL with a policy containing only
>> "-ALL" to be
>> > a special case which justifies early reject.  I think it is obvious
>> that if an
>> > evaluator does not wish to allow a DKIM override for that situation,
>> then the
>> > SHOULD NOT can be ignored.
>>
>>
>> SPF clearly suggests that receivers apply whitelisting broadly.  In
>> addition,
>> publishers can explicitly include ?exists:%{ir}.list.dnswl.org or
>> similar
>> reference to public whitelists.  Whitelisted is neither pass nor fail.
>>
>> Of course, messages rejected before any DMARC processing takes place will
>> not
>> contribute to feedback reports.  Operators just have to be aware of it.
>>
>>
>> Best
>> Ale
>> --
>>
>> I agree with Ale. Further, it is not as if we are considering this in a
> vacuum. Since originally being made public, DMARC has been widely
> implemented and it has not been identified that this (early reject on SPF
> -all) has been a significant or even an insignificant problem based on data.
>
> Michael Hammer
>
_______________________________________________
dmarc mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dmarc

Reply via email to