I think it's fine for Sections 9.6.3 and 9.6.4 to contain words like "should" and "must", because those sections are directed at subscribers and relying parties. However, it is important to note that these sections are not *binding* on those parties: the CA places requirements on subscribers via its Subscriber Agreement, not its CPS; and it doesn't have leverage to place restrictions on relying parties at all. So I think that using "should" and "must" in those sections, especially RFC 2119 all-caps versions of those words, is at best wishful thinking and at worst actively misleading.
Aaron On Thu, Aug 13, 2026 at 8:19 AM 'Ben Wilson' via [email protected] <[email protected]> wrote: > Forwarding to the list > > ---------- Forwarded message --------- > From: <[email protected]> > Date: Thu, Aug 13, 2026 at 2:41 AM > Subject: AW: Descriptive vs. normative language in CP/CPS documentation > To: <[email protected]>, <[email protected]> > Cc: <[email protected]> > > > Hi Ben, Aaron, > > > > from the practical experience we are currently having, as we are > rebuilding our CP/CPS structure from an overarching CP (with lots of RFC > 2119 keywords ) and specific CPS to single purpose CP/CPS, I agree with > Aaron, that a combined CP/CPS is nearly identical to a standalone CPS > without RFC 2119 keywords. > > > > Our draft of the “CP portion” in Section 1.1 is currently something like > this: > > > > *This document is the *CP/CPS TLS* ... * > > *In the structure of the [RFC3647], it describes the implementation of all > relevant requirements of <List of all relevant laws and specifications, in > our case EU and German law, ETSI, CA/Browser Forum, Root Stores, CCADB>* > > *… confirms compliance with all relevant requirements of the current > version of the above-mentioned documents. In the event of a conflict > between this document and the above documents, the provisions of the above > documents shall prevail …* > > > > But in the following, the language is generally descriptive rather than > normative. There may be a few exceptions, e.g., in chapter 9.6.3 and 9.6.4, > where subscribers and relying parties are obliged to comply with certain > requirements and thus the language can be, e.g., "subscribers must > guarantee..." or "relying parties should...". But I'm not sure about that > yet, as it could also be descriptive, e.g. "The requirements... are > described in the Terms of Use" > > > > Kind regards > > > > Stefan > > > > *Von:* 'Ben Wilson' via [email protected] < > [email protected]> > *Gesendet:* Donnerstag, 13. August 2026 06:26 > *An:* Aaron Gable <[email protected]> > *Cc:* [email protected] <[email protected]> > *Betreff:* Re: Descriptive vs. normative language in CP/CPS documentation > > > > Thanks, Aaron. > > I agree that the absence of an RFC 2119 keyword does not make an otherwise > applicable CPS statement nonbinding. A CPS must describe the CA’s > practices, and a CA is expected to operate consistently with such > descriptions. > > Your recommendation that CPS documents contain no RFC 2119 keywords—and > that the CP portion of a combined CP/CPS be limited to identifying the > external policies with which the CA complies—is a broader proposal. I would > be interested in hearing whether other CA operators, auditors, root store > operators, and community members agree with that approach. > > I also welcome everyone's thoughts on how Mozilla should evaluate whether > any particular CPS provision constitutes an implementation commitment. > > Thanks, > > Ben > > > > > > On Wed, Aug 12, 2026 at 6:41 PM Aaron Gable <[email protected]> wrote: > > Hi all, > > > > I'll reiterate my previous position: Policies are set by the PKI in which > the CA operates; practices are described by the CA. Having each CA write > its own CP has always been a category error. CP documents are > *prescriptive*, and rely extensively on RFC 2119 keywords. CPS documents > are *descriptive*, and should contain zero uses of the RFC 2119 keywords. > > > > In my opinion, a combined CP/CPS should be nearly identical to a > standalone CPS, and contain no uses of RFC 2119 keywords. The CP portion of > a combined document is just the paragraph at the top stating conformance to > the Baseline Requirements (as required by BRs Section 2.2 > <https://github.com/cabforum/servercert/blob/main/docs/BR.md#22-publication-of-information>) > and other root program policies (as required by the Chrome Root Program > Policy Section 1.1.3 > <https://googlechrome.github.io/chromerootprogram/#113-chrome-root-program-participant-policies>, > among others). > > > > With that context, my responses to Ben's specific questions are inline: > > > > On Sun, Aug 9, 2026 at 11:11 AM 'Ben Wilson' via > [email protected] <[email protected]> wrote: > > Does a CPS statement commit a CA to a practice even if it does not use > words such as MUST, each, or every? > > > > It is my opinion that a CPS or combined CP/CPS statement* definitely does* > commit > a CA to a practice even when the statement does not use MUST / SHALL / etc. > > > > Other words like "each" or "every" are more ambiguous. That gets into your > other questions about ordinary meaning, drafter's intent, and genuinely > unclear statements. > > > > What should be Mozilla's process or criteria when a CPS provision is > identified as ambiguous or unclear? (E.g. ordinary meaning, the > surrounding text, certificate profiles, applicable requirements, actual > issuance practices, or a CA's or drafter's intent.) > > > > I think that Mozilla should make a public judgement call as to whether the > ambiguity is sufficient to require an incident report. I don't think > there's a great objective scale against which to make this judgement. One > can imagine things like "a reasonable reader" (similar to the US legal > system's "reasonable person") being invoked, but I don't know exactly how > to structure that. At the end of the day, Mozilla is the entity that has > the ability to close Bugzilla tickets, so Mozilla is the entity that has to > decide -- in public -- whether the ticket gets to be closed or not. > > > > Should Mozilla’s CP/CPS guidance clarify the situations in which > descriptive statements would be considered commitments, and should we > explain how to handle genuinely unclear statements? > > > > I think that Mozilla should require that CPS documents contain no RFC 2119 > keywords, and that combined CP/CPS documents be formatted as I described > above: as a CPS, plus an additional paragraph stating adherence to external > CPs. This will make it abundantly clear that even descriptive statements > are binding, because descriptive statements are the only kind that will be > present. > > > > Aaron > > -- > You received this message because you are subscribed to the Google Groups " > [email protected]" 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/mozilla.org/d/msgid/dev-security-policy/CA%2B1gtabDQEJQ42gkPFMrqSeLFj56yqQzOV43hUkJwxGrUFXiPA%40mail.gmail.com > <https://groups.google.com/a/mozilla.org/d/msgid/dev-security-policy/CA%2B1gtabDQEJQ42gkPFMrqSeLFj56yqQzOV43hUkJwxGrUFXiPA%40mail.gmail.com?utm_medium=email&utm_source=footer> > . > > -- > You received this message because you are subscribed to the Google Groups " > [email protected]" 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/mozilla.org/d/msgid/dev-security-policy/CA%2B1gtaZ%2B9%2Bt56fNxHte6pdSEoG5DG5Qga8AO3aYDgMsR-EZSPA%40mail.gmail.com > <https://groups.google.com/a/mozilla.org/d/msgid/dev-security-policy/CA%2B1gtaZ%2B9%2Bt56fNxHte6pdSEoG5DG5Qga8AO3aYDgMsR-EZSPA%40mail.gmail.com?utm_medium=email&utm_source=footer> > . > -- You received this message because you are subscribed to the Google Groups "[email protected]" 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/mozilla.org/d/msgid/dev-security-policy/CAEmnErctBXKiePFqirbDuoAQJsGAytCGaL9qoUCYftuY1BjtTA%40mail.gmail.com.
