Re: [ebxml-bp] RSD (Multiparty collaboration)

["David RR Webber" <david@drrw.info>]

DRRW@

JJ - I'm not thrilled by this idea - mainly because I'm seeing we need
more than CAF provides in the long run (V3.0) and I'd hate to point
ourselves down one direction for V2.0 that then does not give us
the full capability longterm.

Also - the scenario you describe is EXACTLY the one that I posted
from the IV&I use case.  With a disjointed flow control.

I was able to resolve this by using the
context mechanism combined with the delayed startBPSS mechanism
that John Yunker first showed a need for.  So the idea is that the
"startIf" mechanism can apply not only to when the BPSS itself starts,
but also to when an action inside the BPSS starts.

Essentially this acts as the coordinator - since you will need to use
the ebMS to bring some event along, and you will have to have the
BPSS engine periodically checking for input on ebMS queue.

I explained this much better in the IV&I use case - and how each
peice of the BPSS finishes - and then the next starts - to avoid
issues with timetoperform - and to draw together multi-instances
of this between multi-partners.

Anyway - the bottom line is we can build this within the framework
of what we have now for V2.0 - without involving the CAF stuff - so
I'd say - please review the IV&I use case I posted - and we can
then work on the details from there.

Thanks, DW.

JJD@
The WS-CAF specification provides a model to build context managers and
coordinators. 

I'll make a proposal at the end of the month for an architecture based
on WS-CAF.

I have not yet fully thought through the role of the coordinator, but
typically it needs to intervene when two parties execute a BTA and a
third party need to do something then, with respect to the completion of
this event.

Note that it must not be too common in business for this to happen,
right? When it is needed, the coordinator should use a signal rather
than asking the collaboration designer to create a specific message to
handle that particular transition.

I have a lot of respect for the current design and we should be very
careful in evolving it. It has an excellent execution model. The only
drawback is at design time we have to define all these binary
collaboration which hamper a global view.

@JJD

@DRRW