Re: [dhcwg] AD review of draft-ietf-dhc-sedhcpv6-05

[Sheng Jiang <[e-mail]>]

>On Feb 12, 2015, at 12:57 PM, 神明達哉 <[e-mail]> wrote:
>> To be clear: my current position is still against this option.  I'm
>> always in the camp of "problem first, then solution".  And that's
>> also the case of this document since around May 2014:
>> http://www.ietf.org/mail-archive/web/dhcwg/current/msg15520.html
>
>This is a fine idea in theory.   However, section 7 does not represent a valid
>problem statement.
>
>> But at least around last May our consensus
>> should have been against that approach, hence the current organization
>> of the draft.  Just because it triggers a hot discussion doesn't
>> automatically dismiss the original concern and justify ignoring it.
>
>It wasn't clear to me that there was a consensus to add the speculative use
>cases to the document, but I will leave that to the chairs to decide.
>
>> Unless and until we see a clear wg consensus that we now allow it to
>> ship even if no one can explain what kind of problem it will solve, I
>> object to moving forward with just removing this section.  And, IMO,
>> this consensus change will require another round of WGLC.
>
>That's a reasonable request.
>
>But that said, let's be clear: we do actually have a problem.   The problem is
>that the authentication scheme described in RFC 3315 is not deployable,
>because it relies on a key distribution model that doesn't work with a typical
>DHCP deployment.   This document uses a key distribution model that would
>work better.   So this is definitely an improvement over RFC 3315.
>
>It is however true that we do not know who would use this.   This document
>is very speculative.   And I can imagine other ways to solve this problem--for
>example, doing a shared secret exchange using public keys.   But these are
>problems that the working group has been aware of for over a decade, and
>we've made no forward progress.
>
>The reason I support issuing this document as an RFC is that I think it would
>work, and would work better than what we have now, and I'd like to see if
>anybody will implement it.   Until there is a document to implement, clearly
>nobody is going to implement anything.
>
>So that's why despite the lack of a use case document or a list of known use
>cases other than section 7, I think we should publish the document.

There are two changes feedback from AD review, that would require another WGLC in my opinion:

1) remove the section that descript notional use cases, which does not represent a valid problem statement. It will result no use case description in the document. But the original problem, that no mechanism exists for security DHCP messages between servers and clients that does not have a serious key distribution problem, should be sufficient to motivate the proposed mechanism.
2) remove the limitation that only clients use PKI- certificates and only servers use public keys. The new text would allow clients use public keys and servers use PKI-certificates.

I will apply the above two changes and submit an update version today. Then DHC chairs may launch a short WGLC to see whether the WG could reach consensus on this. It would form the follow-up action.

Best regards,

Sheng