Re: [Sipbrandy] WGLC: draft-ietf-sipbrandy-rtpsec-03

[Russ Housley <[e-mail]>]

> On Jan 11, 2018, at 12:03 PM, Peterson, Jon <[e-mail]> wrote:
> 
> 
> Thanks for these notes Russ:
> 
>> Major:
>> 
>> In Section 4.3, I do not know what is meant by "STIR Profile for media
>> confidentiality".
> 
> I think we've been using that language as the name for this entire effort:
> as a way to leverage STIR through a particular extension with its
> incumbent behavior to get media confidentiality. Since media
> confidentiality was definitely not a requirement of STIR initially, this
> is effectively a profile, a narrow use case applying STIR to something
> slightly different. I think SHAKEN set the precedent for profiling STIR
> through an extension, and we are just following that. I'm not sure how to
> make that more clear, though.

I think we need to separate the authentication provided the signature in STIR and the ability for some other mechanism to leverage that signature to provide confidentiality.

> 
>> 
>> Minor:
>> 
>> I think the discussion in the first paragraph of Section 3.1 should be
>> slightly expanded to be clear that we are not talking about integrity
>> protection of SDP k= records.  I'd prefer a MUST NOT statement.
> 
> Fine by me. 
> 
>> 
>> In Section 4.1: s/SIP endpoints must acquire/SIP endpoints MUST acquire/
> 
> So... I don't think this language is stipulating implementation behavior
> as such - it is expressing a practical necessity rather than something we
> would test conformance with, I think? The spec then goes on to discuss the
> actual protocol mechanisms you might use to actually acquire credentials.
> If the "must" is annoying I can paraphrase to "as a matter of practical
> necessity need to". Or maybe I need more RFC8174 sensitivity training.

That change would help me.

> 
>> 
>> In Section 4.2: 
>>  s/should be discarded/SHOULD be discarded/
>>  s/should be generated/SHOULD be generated/
> 
> These ones however I think you're right, we are talking about
> implementation behavior.

Thanks.

Russ