Separating PIDF from tuple/presentity semantics (was Communication means in draft-ietf-impp-cpim-pidf-05)

> We need some kind of default, fallback
> understanding of segmented presence information, with an
> understanding that
> particular presence protocols and/or applications may be able to
> have a more
> sophisticated understanding of the same presence information. I think this
> matches both the letter and spirit of RFC2779 3.1.4.

I had a conversation with Derek this morning, and I think it's fair to say
that we as chairs are generally in favor of pursuing this goal but not at
the risk of derailing progress on CPIM or PIDF, which appear to be quite
close to being finished soon.

I think the crucial first goal of IMPP has to be getting to the point where
there is an agreed-upon basis for moving information between different
systems, and it is a matter for future refinement to achieve better
alignment among the different kinds of information those systems choose to
expose to each other.

If we happened to get a rare convergence of unlikely events and quickly came
to a rough consensus on what tuples/presentities "Really Mean", it would be
great to fold that into CPIM and/or PIDF as appropriate.  But I think we
have to allow for the possibility of such a definition being itself a long
and contentious process.

I think there is already a "default" understanding of segmented presence
information, however weak those semantics are: tuples are somehow related to
a presentity and somehow subordinate to the presentity. I don't think
there's much more that can be said without provoking argument (and even this
vague characterization may attract dissent.)  I'm happy for people to work
on more-detailed semantics, but the current tuple/presentity arrangement is
the best that we were able to do in previous attempts to resolve this kind
of question.

So concretely, I'd ask for any effort to define additional shared semantics
of presence information to be a different document from PIDF. PIDF should
stand on its own without that document.

A more speculative suggestion: We might think in terms of simply registering
how different IMPP-compliant systems use presentities and tuples, rather
than trying to define a single segmentation structure.  This would then
function as an aid for the developers of user agents trying to decide how
best to present received information.

--Mark




  [reminder: [email protected] for non-technical discussions, please]