Re: [stir] Benjamin Kaduk's Discuss on draft-ietf-stir-enhance-rfc8226-03: (with DISCUSS and COMMENT)

[Russ Housley <[e-mail]>]

Ben:

>> For now, just responding to the DISCUSS ...
>> 
>>> ----------------------------------------------------------------------
>>> DISCUSS:
>>> ----------------------------------------------------------------------
>>> 
>>> Let's discuss whether we should have content in this document discussing
>>> the relationship between this new certificate extension and the
>>> extension defined by RFC 8226.  In paticular, whether it is
>>> permitted/expected for both extensions to appear in the same
>>> certificate, and whether any specific processing is required in that
>>> case.  (If no such processing is specified, we could end up with
>>> interesting edge cases where a given PASSporT is handled differently
>>> depending on which extension(s) are supported by the recipient.)
>> 
>> I see no reason why both extensions would ever appear in the same certificate.
> 
> There was an open question for me about backwards compatibility...
> 
>> If there is a mustExclude, then the constraints cannot be expressed in the extension defined in RFC 8226, so I would only expect the on in this document to be present.
> 
> ... specifically, if the desired constraints include both mustExclude and
> one of the preexisting restrictions, does "something is better than
> nothing" apply when the recipient might not implement the new extension,
> prompting a desire to include both extensions so that the recipient will
> apply those constraints it does know about?

No.  It is important that all of the constraints are met for the use cases where mustExclude come into play.

> If there is a desire to provide this backwards compatibility, is it better
> to duplicate the constraint in both extensions or rely on both extensions
> being processed and only put mustExclude in the new extension?  If the
> latter, why do we bother allowing the mustInclude and permittedValues forms
> in the new extension?

Indeed,  If that were the case, I think a different design would be preferable.

>> If there is not a mustExclude, the the constraints can be expressed in either extension equally well, so I would expect only the extension defined in RFC 8226, with the presumption that it is implemented more widely.
>> 
>> I can add this advice to the document if it resolves your concern.
> 
> I think at least the last bit makes sense to mention in this document.  I'm
> still not entirely clear on the full picture, though.

I'll craft some text to address this, and then I will turn to your other comments.

Russ