Discussion topics for MEP call

In thinking through the MEPs on operations issue, I've come up with a few
topics to discuss.  There are a few items that I think the group needs to
reach concensus on before we can make good progress.  Other items here are
clarification discussions.  This is probably more than we can get to, but
should provide a framework for today's discussions.

1) Should WSDL continue to define operations from the 'service' POV? 

2) Need for serviceRole on MEP? (Youenn's proposed addition)
   See: http://lists.w3.org/Archives/Public/www-ws-desc/2002Dec/0057.html

3) Should MEPs added at abstract portType layer, or binding layer?
  - I would agree we need the abstract layer -- some have argued this should
be at the binding layer.
  - Abstract could be bound to binding-specific MEPs (example SOAP
request/response and SOAP-response)

4) HTTP not the only binding used in Web services.  Example of other
possibilities:
   - SMTP/POP
   - NNTP
   - IP / IP-multicast
 - ARGUMENT: All MEPs need NOT be supported by ALL bindings.  That is OK.
 - It seems like many times, discussion of MEPs is held up with a 'but how
would you do this
   with the HTTP binding'?  'output' is a great example.

5) Better definition of abstract MEP - differentiate from choreography!
 - I think we sometime rat-hole on this issue (or is it trout pond?).
   Adding flexibility to the spec always runs the risk of people abusing
that flexibility.
 - MEPs - usually have 'goal' message to finalize transaction
        - usually represent a atomic transaction
        - NOTE 'usually' -- might have cases where these don't hold

6) Difference between 'WSDL-Operation-Type' as Prasad indicated and abstract
MEP?
See: http://lists.w3.org/Archives/Public/www-ws-desc/2002Dec/0061.html

7) What MEPs should be defined by the WG (if we can get this far ;-)?
 - input/output
     DonM's outstanding action item -- take SOAP MEPs and generate abstract
MEP
     How might SOAP MEP be rewritten to reference abstart MEP?
 - input
 - output (variation #1 and variation #2)
 - output/input (mirror or input/output)
 - output/input* (NOT mirror of input/output)
 - double-opt-in

8) output/input* raises issue of cardinality on messages -- allowed?
defined where?
   - MEP should be able to specify cardinality
   - should WSDL syntax allow it as well (evidently this was already voted
out early on)

9) Specific use-cases would be helpful...
   Use-case brain-storming session?
   Task specific people to write up top 'x' number of use cases?

10) Issue of sequence of messages within operation
   - should the sequence define the MEP?

Received on Tuesday, 17 December 2002 10:37:48 UTC