Re: [Acme] Optional "Wildcard" authorization field

[Brad Warren <[e-mail]>]

In a similar vein, another small but real world example where this being standardized would be useful is Certbot has the flag —allow-subset-of-names that causes it to not treat it as a failure if you cannot complete all authorizations and instead obtain a certificate only for the identifiers you were able to successfully obtain valid authorizations for. We currently don’t support this wildcards because we’re unable to differentiate an authorization for a wildcard from one for the base domain.
 
> On Mar 2, 2018, at 12:24 PM, Richard Körber <[e-mail]> wrote:
> 
> Hi!
> 
> I also see the problem that clients might get confused because there are
> seemingly two authentications for the very same domain ("example.com").
> Without a proper flag, they could only be distinguished by the different set
> of challenges, but that would assume knowledge of server internas.
> 
> Also, the client could actively validate the authorization having
> wildcard=true first, and then could check if it is still necessary to do the
> other authorization. Depending on the server implementation, if verifying the
> wildcard domain also validates the domain itself, it could save an unnecessary
> challenge round trip.
> 
> -- 
> Richard Körber
> 
> 
> Am 02.03.2018 um 19:28 schrieb Felipe Gasper:
>> One (fairly) obvious use of the “wildcard” flag is for status reporting without the context of the original newOrder. The client can thus more easily say:
>> 
>> Authorization for “*.example.com”: $message
>> 
>> … without having to correlate the authz object with the order.
>> 
>> -FG
>> 
>>> On Mar 2, 2018, at 12:32 PM, Daniel McCarney <[e-mail]> wrote:
>>> 
>>> Richard: That's up to the client and the situation. In the linked Certbot issues there were questions about error output/UX. In this case if the client saw an error attached to an authorization with the identifier `{ "type": "dns", "value": "example.com"}` and the authorization had `wildcard: true` the client could say "Failed to authorize *.example.com: blah blah blah" or otherwise use the knowledge to inform their actions (whatever they may be).
>>> 
>>> In general I think there will be reason for client developers will want to have a standardized way of understanding if an authorization corresponds to a wildcard identifier or not. I'm hopeful some client developers will chime in with more concrete examples, I'm a server-side grunt.
>>> 
>>> - Daniel / cpu
>>> 
>>> On Fri, Mar 2, 2018 at 12:29 PM, Richard Barnes <[e-mail]> wrote:
>>> Daniel: I don't have a strong objection here, but could you elaborate what the client is expected to do differently based on this flag?
>>> 
>>> On Fri, Mar 2, 2018 at 12:22 PM, Daniel McCarney <[e-mail]> wrote:
>>> Hi folks,
>>> 
>>> There is a slight disconnect with the current specification between identifiers in newOrder/newAuthz requests and identifiers in authorization objects. The former is allowed to include wildcard domains in the value of DNS type identifiers while the latter is forbidden. 
>>> 
>>> Let's Encrypt's implementation of ACME wildcard issuance guessed this might lead to confusion and introduced a non-standardized "Wildcard" boolean field in authorization objects. If true, then the identifier value in the authorization identifier is known to be the base domain corresponding to a wildcard identifier from the newOrder/newAuthz request.
>>> 
>>> I think it would be beneficial to the entire ecosystem if this optional "wildcard" authz field could be standardized so I've sent a small PR: https://github.com/ietf-wg-acme/acme/pull/402 Both Certbot and ACME4J have independently bumped into this disconnect, which helps justify the need.
>>> 
>>> - Daniel / cpu
>>> 
>>> 
>>> 
>>> 
>>> _______________________________________________
>>> Acme mailing list
>>> [e-mail]
>>> https://www.ietf.org/mailman/listinfo/acme
>>> 
>>> 
>>> 
>>> _______________________________________________
>>> Acme mailing list
>>> [e-mail]
>>> https://www.ietf.org/mailman/listinfo/acme
>> 
>> _______________________________________________
>> Acme mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/acme
>> 
> 
> _______________________________________________
> Acme mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/acme