Dear All:
I'm very sorry about the slow delivery of the updated BES spec. I thought
that I'd be able to send it after my flight out to Vancouver prior to my
vacation, but failed to get email connectivity.
In any case, I enclose a new document with many changes. If you send
comments before Sunday morning, I can try to address them on the flight
back Sunday afternoon.
I've tried to document some of the primary changes below. However, I
really think it is important for people to read the document carefully,
as there may well be other things that seemed unimportant to me but are
in fact problematic. We should then discuss issues that seem problematic
on a telecon.
1) Things that have changed:
a) I've broken existing operations out into separate BES-Factory and
BES-Management port-types, and added a new BES-Activity
port-type.
b) I've removed the material on separate WSRF and RMF renderings, and the
material on notification, based on the assumption that we can do what
David Snelling proposes and define BES in such a way that other desired
behaviors (e.g. WSRF-specific or WS-Transfer-specific operations) can be
added if/when desired via composition. To this end, I've also introduced
the WS-BaseFaults faults. I realize that these changes haven't been
approved, so we may need to back them out. But at least this way you can
see what it looks like.
c) I removed the "laststate" component from the
"ActivityStatus" type, because this component seemed to be
there specifically to deal with suspended activities, which are not
longer part of the core specification. Presumably a BES that supports an
extended state model with suspension will need to introduce
"laststate" as an extension to the ActivityStatus
type.
2) Things that seem problematic, or that are not done:
a) We need a new state model section: I think someone is writing one? (I
removed what is there.)
b) I can't recall the new base states; these need to be
entered.
c) We need to update the BES-Factory attributes.
d) I assume there should be per-activity BES-Activity attributes. I
haven't entered any.
e) I've introduced "subscription" and "lifetime
management" elements as optional JSDL extensions. Certainly people
that compose WS-Notification/WS-Eventing and WS-ResourceLifetime,
respectively, with BES will want to use these. However, it may be that it
is not appropriate to mention them in the BES spec, because they are
specific to other WSRF/WS-Transfer specs. (I myself think that they
should not be there, but I want to raise the topic to get other people's
views.)
f) The normative stuff in the appendices hasn't been completed.
Regards -- Ian.