RE: Communication means in draft-ietf-impp-cpim-pidf-05

> The real question here is how tuples should be constructed and
> interpreted.
> Unfortunately, RFC2779 defers this question to the model in RFC2778, and
> RFC2778 offers absolutely no motivation for tuples or explanation for how
> they should be used. We might assume that if you have multiple
> presence user
> agents that are publishing presence information to the presentity, the
> presentity composes each chunk of presence information into its own tuple
> before publishing the composed presence information to a presence service
> for distribution to watchers (I would be interested to hear the
> perpsective
> of any RFC2778/RFC2779 authors on this matter if they feel there is some
> other motivation for tuples).

Tuples were simply intended to be general enough to cover all the cases of
presence systems that had been identified. Remember that the original goal
of 2778 was to be able to map consistent terms onto each presence system,
not to serve as a basis for interoperability. If I have a Honda and you have
a Ford, we could probably successfully agree on which parts of each other's
engine are a "spark plug", "oil filter", and so on without expecting that we
will then be able to swap parts.  At the point when 2778 was written,
discussions tended to fall apart because one person would use "spark plug"
for what someone else called "oil filter." Part of reaching a point where
everyone could nod and say, yes, 2778 covered *their* favorite system as
well, was to allow for "bundling" of information in tuples.

We periodically restart a very old discussion about whether presentities
have multiple presences or whether principals have multiple presentities. As
far as I can tell, there's no reason at the protocol level to choose one
over the other, but there are a bunch of other reasons why people wind up
being heavily committed to one approach and very reluctant to consider the
other.  So 2778 was written to accommodate both styles.

I think the relevant point for interoperability is that implementations
should avoid assuming that another party splits presentities/tuples in the
same way. For example, it's perfectly reasonable for the user agents of one
system to display presentities as "people" and tuples as "devices" while the
user agents of another system displays presentities as "people's devices"
and suppresses tuples (since maybe each presentity only ever sends one). But
if those systems are interoperating, it would be bad if they applied those
same display techniques to the information received from the other system.

--Mark




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