Minutes from todays interop call

[Daniel Millwood <MILLWOOD@uk.ibm.com>]

Attendees Paul Fremantle, Ondrej Hrebicek, Dan Millwood, Matt Lovett, Doug
Davis

Ondrej - Problem in request-reply scenarions.  Acks are piggy-backed but
AcksTo EPR is different to ReplyTo EPR so should they be?

Doug - Propose updating the scenario doc

---------------------

Paul to check on main TC call if any non-voting members are planning on
coming to the F2F, so Doug can get proper numbers

Doug clarified that everyone saw his note on the interop status webpage  (
http://wsi.alphaworks.ibm.com/interop/RXInteropStatus.jsp )

Paul F - will MSFT have endpoints up before the F2F?

Ondrej - Hopes to have something up by the end of the week, not up yet.

Paul F - Proposal for scenario 2.4.  Scenario 2.4 is not upto the CD 3
level because of the lack of TerminateSequenceResponses.  Given the time
left between now and the F2F, implementations may not be able to get any
fixes done prior to the F2F.  Propose to move scenario 2.4 into a separate
document.  Leave main document fairly static, but other doc for 2.4 can be
changed and worked on at the interop.  Put some text in it saying its
experimental and tested on a best can do basis.

Ondrej - Was going to update scenario doc once Marc puts proposal on offer
to the main TC.

Paul - Scenario 2.4 cannot work based on the level of the spec - CD 3.
Changes to the wire protocol would probably be needed to reflect Marcs
proposal.  Paul is hoping some of the interop time can be spent testing
around scenario 2.4.

Ondrej - Key point I wasnt aware of was that F2F was based on CD3.  Given
that, the proposal makes sense.  Wants to run by Marc.

Paul - We need to agree this now.  Can we get Marc on the call.

Doug - Willing to take scenario 2.4 out and put into separate doc.

Ondrej - Cant get hold of Marc.  Agrees that proposal makes sense.

-------------------------------------------
Back to the piggy-backing problem.

Matt - Cant require an implementation to piggy-back Acks.  Safe thing to do
is two EPRs and separate messages

Doug - Could say as an optional variant, make them the same to test
piggy-backing

Matt - Some implementations may not be able to put a new variant of the
scenario in at the 11th hour

Ondrej - Can say implementations should be able to accept piggybacked Acks,
even if they dont send them.

Doug - Lets change both ReplyTo and AcksTo EPRs to be the same, and make a
note that its upto the RMS to decide if it wants to piggyback or not.

Ondrej - Agrees

Paul F - Wants to see the proposal in writing

Ondrej - Should we add in more tests for CloseSequence?

Matt - Only need close in a situation where you want to resolve ambiguity.
The scenarios in the interop doc dont in general have ambiguity to resolve.

Paul F - Its recommended that RMD closes a sequence.  As a client if you
know all messages have been Acked you dont really need to send close.  If
the scenarios were more complete, it would be nice to have a scenario where
the RMD unilaterally closes the sequence and the RMS reacts to it.  Maybe
for next time.

Doug - Will try and pull out scenario 2.4 today.

Paul - Will then need to be iterations on it as Marcs proposal on offer
becomes clearer.



Thanks,  Dan

WS-Reliable Messaging Architecture/Development
IBM WebSphere Messaging Design and Development
MP 211
Hursley
Tel. Internal 248617
Tel. External +44 1962 818617
Email. millwood@uk.ibm.com