Re: [pcp] PANA implementatinos to consider

[Margaret Wasserman <[e-mail]>]

Hi Yoshi,

On Sep 14, 2012, at 9:22 AM, Yoshihiro Ohba wrote:
> On the other hand, if my understanding is correct, the needed set of
> interactions between an AKM function and the base PCP functions are
> not much different regardless of whether the AKM function is PANA or
> PCP-specific, and therefore I think the interactions do not have to be
> considered as a disadvantage of PANA-based approaches.

I agree with you -- the need for integration of the authentication and PCP functions is effectively the same in all of the approaches we've considered.  I was responding to Alper's assertion that this sort of integration was not needed in the side-by-side (or demux) approach.

There are some things (like how we handle authentication requests to PCP Servers that don't implement PANA) that will work better in either encapsulation approach than in the demux approach.  There are other things (like the need to check that the fields in the PCP header match corresponding fields in the PANA packet that are specific to the PANA encapsulation approach).  And there are other issues (like the need to respecify and re-implement some functions already included in PANA) that will be specific to a PCP-Specific approach.  These are the things that the WG needs to understand and decide between, IMO.  To understand these things, we need to understand each of the PANA-based approaches in more detail than is currently documented.

Margaret