On Wed, Sep 27, 2006 at 11:04:30AM +0200, Mark Townsley wrote:
> 
> I'm copying Donald, as one of the authors of RFC 4086 for advice here. 
> Donald, this is in regard to draft-ietf-pana-pana-12.txt
> 
> This recommendation on selection of the Cookie:
> 
> > In order to do that, the cookie MUST be computed in such a way that
> > it does not require any per-session state maintenance on the PAA in
> > order to verify the cookie returned in the PANA-Start-Answer message.
> > The handshake phase that takes advantage of cookies is called
> > "stateless handshake".  The exact algorithms and syntax used by the
> > PAA to generate cookies does not affect interoperability and hence is
> > not specified here.
> 
> Seems to be in conflict with:
> >
> > The Cookie AVP (AVP Code 3) is used for carrying a random value
> > generated by the PAA according to [RFC4086].

I think what we are trying to mention in the two part is that we don't
specify detailed cookie generation algorithm using a randam value
while generating a random value should follow RFC 4086.

> 
> 
> Seen elsewhere in the document.
> 
> First, there is obvious conflict in that the document says there is no 
> recommendation on algorithms in one place, but refers to RFC 4086 in 
> another (which clearly does have algorithms to recommend). But more 
> importantly, it seems that you are placing requirements on the 
> randomness of the cookie value by referencing RFC4086 (which is a good 
> thing) while at the same time mandating (with a MUST, no less) specific 
> implementation requirements with respect to state at the PAA. I'm not 
> sure thatit is even possible to state that the PAA MUST somehow be able 
> to verify that it generated a cookie, without state, and still be in 
> keeping with RFC 4086 requirements (this is where I would like Donald to 
> give advice).

You are absolutely right that it's not possible for PAA to do cookie
validation without any state, but it does not require *per-session*
state maintenance.  A per-request, small state (not
containing full PANA session attributes) should suffice for cookie
validation.

> 
> Also, if you are going to reference RFC 4086, I believe you need to 
> define what kind of random number you are looking for. To wit, I think 
> you are looking for a "Cryptographically Random" value, as you are 
> trying to protect against blind guessing attacks here.

Yes.   How about revising the latter part to:

"
  The Cookie AVP (AVP Code 3) is used for carrying a value that is
  generated by the PAA from a cryptographically random value.
  Although the exact algorithms and syntax used by the PAA to generating 
  a cookie value is not specified in this document, randomness 
  of the cryptographically random values should meet the requirements
  described in [RFC4086].
"

> 
> Finally, if keeping state at the PAA isn't a big deal for a specific 
> implementation, why include this kind of MUST? I think you are getting 
> too deep into implementation details with normative language.

I think keeping per-session state at the PAA is a big deal in some
environment and that is why stateless handshake is supported.  On the
other hand, I agree that this MUST may be too much.  We can revise the
former part something like:

"
  In order to do that, it is expected that the cookie be computed in
  such a way that ...
"

Yoshihiro Ohba

> 
> - Mark
> 
> 
> _______________________________________________
> Pana mailing list
> [email protected]
> https://www1.ietf.org/mailman/listinfo/pana

_______________________________________________
Pana mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/pana

Reply via email to