Thank you all for the feedback so far. Firstly, I wanted to clarify that we’re aware that none of the data is ever really removed from CCADB. The mentions of deleting items in my original post referred to the CCADB option to mark them as deleted.
In response to Clint: >I think it’s worth noting that 1 & 3 are, I believe, mostly the same; that is, Policy Documents marked as “Deleted” in the CCADB are not removed from the database. The difference between these options that I was attempting to convey is that, in option 1, “old” CPSes are marked as deleted; whereas in option 3, they are not. With that clarification, we take note that Apple sees option 1 as the preferred option at this moment. Since Chris has expressed the same preference on behalf of the Chrome root program, would it be appropriate to declare that option 1 is the requirement? >Regarding the “change log” sections of Policy Documents, I agree there’s not much specific guidance on what is desired or expected here. Both a summary of the changes and a list of sections in which changes occurred seem particularly valuable to me; are there any other suggestions or ideas from the community on this? I think the recent changes in the Chrome Root Program Policy are a good step forward in resolving this. Specifically: “ *To promote simplicity and clarity, these CA policy documents SHOULD be:…..* *available in Markdown.*” I presume this was added partially to aid in comparing different versions. In response to Ben: >I think the middle option is the best choice, and the first bulleted item is a second-best choice, although any of them are appropriate. It seems Mozilla has a different preferred approach (Option #2), but does not set a hard requirement on this. In response to Mike: > I think the most common practice is 1, though? Both options 1 and 2 seem to be common at present. > = is accompanied by a CPSHash property in the cert The CPSURI is not a required field for end-entity certificates, so I’m not sure any new CPSHash property would be a required field in the future. > a CPS is marked “obsolete” or “expired” or whatever the definitions WG comes up with once there are no valid certs left that reference the CPS in question While such an approach *might* work for the TLS WebPKI (though I’m not a fan of the approach), it might cause issues when looking at S/MIME and Code Signing, in which certificates are still useful after expiration. Different policies for TLS compared to other certificates seems like it would open the door to incorrect executions down the line. A unified approach here would be strongly preferred. As some final thoughts to all: It remains unclear if any of the root programs would categorise any of the 3 options as either MUST or MUST NOT. Currently, I’d summarise it as preferred approaches. Since the preferred approaches also do not fully align, would the CCADB Steering Committee be willing to discuss a singular, unified approach, and clarify the requirement in a future CCADB policy update? Document archival for “deleted” CP/CPSes is only available for Root Certificates in CCADB, as far as we can tell. Many Intermediate Certificate records have “Same as Parent” set, but that is not the case for all. If “Same as Parent” is not set, the applicable CP/CPSes are entered via freeform text fields in the Intermediate Certificate record. As such, we currently understand that these fields should always have only the last published version listed (within 7 days after publication). Regards, Martijn Op zaterdag 17 augustus 2024 om 11:03:02 UTC+2 schreef Mike Shaver: > On Fri, Aug 16, 2024 at 7:15 PM Chris Clements <[email protected]> wrote: > >> (e.g., the in-force CP for a given PKI hierarchy is incremented from >> Version 1.1 to Version 1.2), it should be replaced with data that is >> accurate. >> >> >>> 1. The new CPS should be added, and the old CPS should be deleted as >>> it is no longer in effect for new certificate issuance. >>> 2. The new CPS should be added, but the old CPS should be kept in >>> place as long as there are unexpired certificates under its policy. >>> 3. The new CPS should be added. Older entries should be kept >>> indefinitely to serve as an archive overview. >>> >>> >> The Chrome Root Program prefers CA Owners behave similar to #1, but as >> described above, “delete” is not the most appropriate term for the >> non-current versions of the policy documents. In actuality, the system >> behavior is more consistent with #3 due to document archive functionality. >> > > OK, I may be confused here and talking nonsense, but I don’t think that > there is necessarily (or even likely) a single in-force CP for a given PKI > hierarchy at a given time. A policy is bound to a cert via the CPSUri when > that cert is issued, and the policy relating to that cert is then immutably > fixed. That’s why we’ve seen revocations necessary for certs issued in ways > that didn’t match the contents of CPSUri at the time that the cert was > issued, since the only way to remedy that is to issue an new cert bound to > the updated policy. > > At the moment that a new CPS comes into effect, it actually relates to > _none_ of the live certs in the hierarchy, and depending on the update > schedule for the CPS, it might not at any point even apply to a plurality > of such live certs. CCADB should at least have all the CPS documents that > apply to any live certs under a registered root, IMO, and they should all > be treated as equally “active”. > > Unfortunately, this weak binding (requires that a human go to a URL and > pick the one that applies to the date range and cert type) sort of sucks in > terms of clarity and garbage collection. > > If I were designing this system now, I would probably make the rules be: > > - CPS documents are markdown-enhanced text files > - each CPS gets a unique URL, ideally with a human-readable/collatable > version number as a component > = either a content-addressable hash-carrying URL hierarchy hosted under > ccadb.org, or > = is accompanied by a CPSHash property in the cert > - a CPS is marked “obsolete” or “expired” or whatever the definitions WG > comes up with once there are no valid certs left that reference the CPS in > question > > That would let the matching of applicable CPS to cert be mechanically > performable, rather than the gross version-walking-in-an-HTML-index that > has to happen now. Making it be markdown would mean fewer changes that are > due to formatting, and improve the ease of mechanical analysis. > > I’m squinting through some meaty jet lag so that proposed system probably > has some issues, but I think that it’s important that we establish whether > a CPS is per-cert, or per-hierarchy; if the latter, then maybe move it into > the roots themselves and save some handshake bytes? > > Mike > > > > -- 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 on the web visit https://groups.google.com/a/ccadb.org/d/msgid/public/d71c388f-8853-4a4b-9d49-a9aa4635ef50n%40ccadb.org.
