Re: [sipcore] composition or just indirection (was: location-conveyance-03 just submitted)

["Thomson, Martin" <[e-mail]>]

Hi Hannes,

Inline...

> Still trying to catch up with the mail thread but a few basic
> observations:
> 
> 1) There is the 'entity' attribute in the <presence> element that
> points to the Target.
> 
> 2) There is RFC 5491 http://datatracker.ietf.org/doc/rfc5491/ that
> describes how different location chunks (including "compound location")
> are interpreted. It also describes the meaning of different location
> objects in a PIDF-LO.

I'd like to rely on 5491.  But we need to get to the point where a recipient only receives a single document.  That's the snarl we're trying to tease out.
 
> 3) We know that we want to transmit multiple location chunks, and
> multiple location objects of the same Target. We do not seem to have
> use cases for transmitting location of >>different<< Targets in the
> same SIP message.

I don't think that this is entirely correct.  Yes, it's trivial to convey multiple pieces, but we're looking for a way to have a single PIDF but reconcile that with by-reference conveyance.

> So, this should allow us to carry one of multiple MIME bodies with
> PIDF-LOs in a SIP message (referring to the same Target) in a single
> SIP message.


> The following items are left for discussions, I believe:
> 
> - How to encode the LbyR (in addition to the LbyV)

How that cat is skinned depends on the logical model that we agree upon.  It's easy enough to just throw in another body (or header even), but that doesn't address the concern that a recipient/UAS might be unable to make sense of the information that it receives.

> - Procedures for how an intermediary attaches a reference to a PIDF-LO

I think that this has already been solved.  The intermediary sends a 424 with the reference in it, the UAC attaches the reference if it agrees.  (In an emergency, there are no such niceties: the intermediary steps into a B2BUA role.)  

An intermediary has to comply with any constraints we place on the header: one value or one header, one value and one reference, or whatever conclusion we come to.

> - Whether there needs to be a relationship between the 'entity'
> attribute in the <presence> element and the From header of a SIP
> message.

That comes down to header semantics.  Equality of the entity attribute and From header is both unnecessary and difficult to achieve.  A weaker relationship might be necessary, see my response to Paul.

It's a damned shame that we don't have a good understanding of the semantics we want from the header.  How many years have we been working on this?

> I am less interested in the used-for-routing optimizations etc.

Agreed.

> Did I correctly capture the challenge?
> 
> Ciao
> Hannes