I agree with Aaron. CAA failures should be treated as requiring a 24 hour
revocation deadline, but the current language does not support this. We
should fix it by changing 4.9.1.1 to include CAA checking failures as a
reason for 24 hour revocation.

On Tue, Apr 1, 2025 at 12:32 PM 'Aaron Gable' via CCADB Public <
[email protected]> wrote:

> Thanks for bringing this conversation to a better venue!
>
> For the record, I agree that failure to properly check CAA *should* result
> in a 24-hour revocation timeline. CAA checking is a critical, and easily
> automated, part of the issuance process.
>
> However, I disagree that this is what the BRs actually say today.
>
> A failure to conduct CAA does not fall under BRs 4.9.1.1(2), because the
> Subscriber has not notified the CA of anything. This is especially true in
> cases where the CA discovers that they have been failing to correctly check
> CAA, but it turns out that correctly checking CAA wouldn't have resulted in
> a different issuance decision.
>
> And as I argued in the Bugzilla ticket, a failure to check CAA does not
> fall under BRs 4.9.1.1(5) because CAA is part of the issuance process, not
> part of the validation of domain authorization or control.
>
> In that Bugzilla ticket, Amir said:
> > From my interpretation of the BRs, domain issuance can be summed up as:
> "Always issue the certificate, except in ${list_of_scenarios}".
>
> I strongly disagree with this interpretation. The BRs, in my opinion, say
> that a CA should *never* issue unless validation has succeeded. I think it
> is relevant and accurate to think of the BRs as first a positive filter
> (DCV, identity validation) that must be satisfied before a CA can even
> consider issuing, followed by a set of negative filters (caa, linting,
> mpic) that may still prevent issuance but could never allow issuance in the
> first place.
>
> Amir also said:
> > CAA (Certificate Authority Authorization) should be considered in the
> authorization part of domain authorization and control.
>
> Unfortunately this interpretation is ahistorical. The "authorization" in
> CAA is about whether the CA is authorized to issue for the domain. The
> "authorization" in "domain authorization and control" is about whether the
> *applicant* is authorized to request issuance for the domain. These are
> fundamentally different things, despite using the same word.
>
> Potential changes to the BRs that would resolve this apparent gap include:
> - changing 4.9.1.1(2) to say "the CA learns that the request was not
> authorized", rather than requiring proactive notification (though this has
> the disadvantage that CAA failures now have two different revocation
> timelines, depending on whether the CA failure to check a record that
> happened to allow issuance vs one that happened to disallow it)
> - changing 4.9.1.1(5) to list CAA alongside DCV
> - adding a new list item to 4.9.1.1 explicitly addressing CAA
>
> Aaron
>
> P.S. apologies that this is likely my only contribution to this thread, as
> I leave for two weeks of vacation today.
>
>
> On Tue, Apr 1, 2025, 08:20 'Amir Omidi (aaomidi)' via CCADB Public <
> [email protected]> wrote:
>
>> Thank you for putting this up CRP.
>>
>> Based on the commentary I've seen through
>> https://bugzilla.mozilla.org/show_bug.cgi?id=1951415, I do think that we
>> should be treating violations that happen in the path of issuance as a
>> misissuance that requires a 24 hour revocation window. I've always used,
>> and interpreted CAA as a security measure. This is also asserted in the
>> RFC <https://datatracker.ietf.org/doc/html/rfc8659#section-5> itself:
>> "CAA records assert a security policy that the holder of an FQDN wishes to
>> be observed by Issuers."
>>
>> > We recognize that a request not being authorized and issuance not
>> being authorized are indeed distinct, but from our view they appear to
>> communicate the same conclusion in that from the subscriber’s perspective
>> issuance should not have taken place.
>>
>> Agreed, I don't think reinterpreting this as anything else is useful for
>> anyone (including the CAs). Having a unified approach to revocation
>> simplifies operations by removing the judgement that is necessary today to
>> ask "is this a 24 hour or 120 hour event?"
>>
>> Cheers,
>> Amir
>> On Tuesday, April 1, 2025 at 11:02:07 AM UTC-4 Ryan Dickson wrote:
>>
>>> Hi all,
>>>
>>> We were curious for community feedback on the applicability of TLS BR 
>>> revocation
>>> timelines
>>> <https://cabforum.org/working-groups/server/baseline-requirements/requirements/#4911-reasons-for-revoking-a-subscriber-certificate>
>>> when a publicly-trusted CA Owner has violated CAA expectations.
>>>
>>> Studying CAA-related incident reports disclosed to Bugzilla over the
>>> past ~5 years, it appears there *might* be mixed interpretations of the
>>> required revocation timelines when CAA is violated. The evaluations below
>>> were based on a quick, non-comprehensive scan of CAA-related incident
>>> reports disclosed to Bugzilla, where we attempted to interpret the
>>> understood revocation timelines described in the incident reports based on
>>> observed revocation behavior, which wasn’t always clear.
>>>
>>>
>>>    -
>>>
>>>    SSL.com: CAA Empty set handling results in Wildcard issuance
>>>    <https://bugzilla.mozilla.org/show_bug.cgi?id=1932973> (certificates
>>>    revoked within 24 hours)
>>>    -
>>>
>>>    SSL.com: Failure to process CAA records from one SubCA
>>>    <https://bugzilla.mozilla.org/show_bug.cgi?id=1938236> (certificates
>>>    revoked within 24 hours)
>>>    -
>>>
>>>    DigiCert: Unclear Disclosure of CAA Issuer Domain Names
>>>    <https://bugzilla.mozilla.org/show_bug.cgi?id=1914911> (certificates
>>>    revoked within 5 days)
>>>    -
>>>
>>>    GoDaddy: CAA checks passed when records contained incorrect variants
>>>    of godaddy.com or starfieldtech.com
>>>    <https://bugzilla.mozilla.org/show_bug.cgi?id=1904749>] (certificates
>>>    revoked within 5 days)
>>>    -
>>>
>>>    GoDaddy: CAA checks did not properly handle issuewild tag allowing
>>>    FQDN SANs to be added to wildcard certs
>>>    <https://bugzilla.mozilla.org/show_bug.cgi?id=1904748> (certificates
>>>    revoked within 5 days)
>>>    -
>>>
>>>    Google Trust Services: SXG certificates issued without correctly
>>>    checking CAA restrictions
>>>    <https://bugzilla.mozilla.org/show_bug.cgi?id=1902670> (certificates
>>>    revoked within 24 hours)
>>>    -
>>>
>>>    Apple: TLS certificates issued outside the TTL of the CAA record
>>>    <https://bugzilla.mozilla.org/show_bug.cgi?id=1841534> (certificates
>>>    revoked within 5 days)
>>>    -
>>>
>>>    GlobalSign: Certificate issued to FQDN with malformed CAA
>>>    <https://bugzilla.mozilla.org/show_bug.cgi?id=1759854> (certificates
>>>    revoked within 24 hours)
>>>    -
>>>
>>>    Amazon Trust Services: Missing CAA Check For Test Website
>>>    Certificates <https://bugzilla.mozilla.org/show_bug.cgi?id=1746945>
>>>    (certificates revoked within 24 hours)
>>>    -
>>>
>>>    Let's Encrypt: CAA Rechecking bug
>>>    <https://bugzilla.mozilla.org/show_bug.cgi?id=1619047> (certificates
>>>    revoked within 5 days)
>>>
>>>
>>> From our view, it seems a 24-hour revocation window is more appropriate
>>> than a 5-day window. If CAA is to be interpreted as a security function
>>> explicitly enabled by a domain owner that is intended to prevent
>>> certificate issuance by unauthorized CAs, it feels odd to treat this
>>> differently than TLS BR 4.9.1.1 (2), which states “The Subscriber
>>> notifies the CA that the original certificate request was not authorized
>>> and does not retroactively grant authorization (CRLReason #9,
>>> privilegeWithdrawn);”. We recognize that a request not being authorized
>>> and issuance not being authorized are indeed distinct, but from our
>>> view they appear to communicate the same conclusion in that from the
>>> subscriber’s perspective issuance should not have taken place.
>>>
>>> We were hoping to collect opinions from this group to ultimately inform
>>> next steps, which might include clarifying expectations within the
>>> CA/Browser Forum TLS BRs.
>>>
>>> As always, additional discussion is welcome and we’re open to changing
>>> our perspective.
>>>
>>> Thanks for your consideration in participating in this discussion!
>>>
>>> - Ryan, on behalf of the Chrome Root Program
>>>
>>> --
>> You received this message because you are subscribed to the Google Groups
>> "CCADB Public" group.
>> To unsubscribe from this group and stop receiving emails from it, send an
>> email to [email protected].
>> To view this discussion visit
>> https://groups.google.com/a/ccadb.org/d/msgid/public/1760a375-039d-47e9-8bb5-e5a05b7d28afn%40ccadb.org
>> <https://groups.google.com/a/ccadb.org/d/msgid/public/1760a375-039d-47e9-8bb5-e5a05b7d28afn%40ccadb.org?utm_medium=email&utm_source=footer>
>> .
>>
> --
> You received this message because you are subscribed to the Google Groups
> "CCADB Public" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to [email protected].
> To view this discussion visit
> https://groups.google.com/a/ccadb.org/d/msgid/public/CAEmnErez62LfyZxsOu44BtZnb088Q7mvnTtKJ7yv%3DZi%2Bb0QmFQ%40mail.gmail.com
> <https://groups.google.com/a/ccadb.org/d/msgid/public/CAEmnErez62LfyZxsOu44BtZnb088Q7mvnTtKJ7yv%3DZi%2Bb0QmFQ%40mail.gmail.com?utm_medium=email&utm_source=footer>
> .
>

-- 
You received this message because you are subscribed to the Google Groups 
"CCADB Public" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion visit 
https://groups.google.com/a/ccadb.org/d/msgid/public/CAFK%3DoS9VeVpwgqhdquqwk%2BX2Ms6Y3bSPw-95UADuZwWOBA3fDg%40mail.gmail.com.

Reply via email to