It has been brought to my attention I did not explicitly declare that the 
comments, requests for clarifications, and opinions I shared on this thread 
were in a personal capacity, this was an oversight.

Ryan Hurst
(Personal)

On Tuesday, August 3, 2021 at 10:26:50 AM UTC-7 Ryan Hurst wrote:

> Hi Kathleen,
>
> I think your language works but I believe we have some language in other 
> WebPKI related documentation that refers to the concept of "TLS capable" 
> certificates. Adopting this language might be sufficient and be consistent 
> with other wording in use. I believe this would make it clear that 
> certificates like SGX that are "not TLS certificates" but are "TLS capable" 
> are explicitly within scope.
>
> As for your questions regarding the reasonableness of providing the full 
> CRLs and/or the JSON array of all CRLs; I personally think it's doable as 
> framed. 
>
> Some thoughts on that since you asked..... providing full CRLs is 
> problematic, especially when revoke/replace is used because the CRLs grow 
> very quickly taxing all relying parties' bandwidth. Rolling issuing CAs 
> regularly can create implicit creating CRL shards but is taxing for the CA 
> to maintain due to manual nature of that process. This leads CAs down the 
> path of CRL sharding, and the JSON array approach accommodates this pattern 
> just fine. I would personally rather see a well-known location be published 
> and that location be incorporated into CCADB vs putting the JSON itself in 
> CCADB in that it complicates the rollout for CAs because they now maintain 
> an online dependency on CCADB but the current proposal is workable.
>
> Regarding your goal of non-overidable errors for a subset of reason codes; 
> I think this is a laudable goal but reason codes have a number of problems 
> I struggle with. 
>
> The first in my experience the subscriber generally do not provide this 
> information, unless forced to. The problem with this is that they don't 
> necessarily understand what each means and may provide the wrong value and 
> possibly choose a value that is not included in the enforced set of reasons 
> and not get the expected behavior. A few years ago I wrote about this here: 
> https://unmitigatedrisk.com/?p=583.
>
> The second reason is that these codes in a blocking fashion introduce a 
> user tax that impacts those that pay for internet based on usage and those 
> that have slow internet. With that said the inclusion of reason codes is 
> already needed for the mega CRLs that are deployed already so I do not 
> worry about the impact of this requirement on CAs.
>
> So to your specific questions:
> 1. Do CAs currently use those reason codes when applicable?
> It is my belief that when provided to them they do include this 
> information but it is unclear if they force providing this information.
>
> 2. Do you think the BRs need to further specify when those reason codes 
> must and must not be used?
> I do.
>
> Ryan Hurst
>
>
> On Tuesday, August 3, 2021 at 9:22:27 AM UTC-7 [email protected] wrote:
>
>> Hi Ryan,
>>
>> Thank you for your feedback.
>>
>> How about ‘Pertaining to Certificates Issued by this CA that Cannot be 
>> used for TLS Server Authentication’?
>> If needed, we can add text to the top of that section explaining more. 
>> Just let me know what you think is needed to fully clarify.
>>
>> You are correct... The non-TLS limitation is currently intended to be the 
>> first phase of this requirement. While root store operators can collect TLS 
>> CRL information from CT, it will be much better to have CAs provide it 
>> directly. So, as browsers improve their revocation checking, I expect our 
>> requirements in this area to also change.
>>
>> I have a few questions for CAs about that...
>>
>> 1) Is it reasonable to have CAs provide full CRLs (or JSON arrays) for 
>> TLS-Server-Auth certificates that is separate from the full CRLs for 
>> everything else?
>> i.e. via different fields in the CCADB?
>>
>> 2) I would like to have non-overidable errors for TLS-Server-Auth 
>> certificates that have been revoked for the keyCompromise, cACompromise, 
>> and affiliationChanged CRL revocation reason codes. Currently CRL reason 
>> codes for TLS end-entity certs "SHOULD" be provided, but I would like the 
>> reason codes to be required under the applicable circumstances.
>> Do CAs currently use those reason codes when applicable? 
>> Do you think the BRs need to further specify when those reason codes must 
>> and must not be used? 
>>
>> Thanks,
>> Kathleen
>>
>>

-- 
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 on the web visit 
https://groups.google.com/a/mozilla.org/d/msgid/dev-security-policy/bd6e884c-9c34-494a-9f20-002380280e12n%40mozilla.org.

Reply via email to