Minutes WS Choreography WG conference call 29 March 2005
Roll Call:
-------------
Steve Ross-Talbot (Co-Chair - Enigmatec), Abbie Barbir (Nortel)
Gary Brown (Enigmatec), Greg Ritzinger (Novell),
Martin Chapman (Co-Chair - Oracle), Nickolas Kavantzas (Oracle),
Monica Martin (Sun), Charlton Barreto (webMethods), Yves Lafon (W3C staff),
Jeff Mischkinsky (Oracle)
Regrets:
Tony Fletcher (Choreography)
Admin:
----------
irc log at http://www.w3.org/2005/03/29-ws-chor-irc
Scribe: Charlton agreed to scribe.
Agenga:
http://lists.w3.org/Archives/Public/public-ws-chor/2005Mar/att-0055/Agenda-03292005-0.txt
No changes, agenda approved
Minutes:
22nd March Minutes:
http://lists.w3.org/Archives/Member/member-ws-chor/2005Mar/att-0006/minutes_20050322_-_0.txt
No objections to 2005-mar-22 minutes
Section 6: Issues requiring discussion from Agenda
agreed as correct representation from minutes.
Action item review
--------------------------
Action 1 in progress
Action 2 in progress
Action 3 in progress
Action 4 now a different action
Action 4 still in progress; worth leaving here for chasing up
Action 5 in progress
Action 6 in progress
Action 7 standing item
Action 8 in progress
Action 9 in progress
Action 10 in progress (Yves, almost done, will work toward producing SGML toward
end of week)
Action 11 in progress (Yves still working on it)
Action 12 in progress
Action 13 in progress (not yet on agenda)
Action 14 done
Action 15 duplicate of 13
NEW ACTION: SRT to follow up on Action item 16 [recorded in
http://www.w3.org/2005/03/29-ws-chor-irc]
March 05 number 28:
http://lists.w3.org/Archives/Public/public-ws-chor/2005Mar/0028.html
March 05 number 29:
http://lists.w3.org/Archives/Public/public-ws-chor/2005Mar/0029.html
March 05 number 30:
http://lists.w3.org/Archives/Public/public-ws-chor/2005Mar/0030.html
March 05 number 33:
http://lists.w3.org/Archives/Public/public-ws-chor/2005Mar/0033.html
<SRT> NOTE: Action item 16 refers to: Finally started to discuss some of the
issues raised by the editing group.
<SRT> http://lists.w3.org/Archives/Public/public-ws-chor/2005Mar/0028.html
<SRT> http://lists.w3.org/Archives/Public/public-ws-chor/2005Mar/0029.html
<SRT> http://lists.w3.org/Archives/Public/public-ws-chor/2005Mar/0030.html
<SRT> http://lists.w3.org/Archives/Public/public-ws-chor/2005Mar/0033.html
Issues requiring clarification
----------------------------------------
Issue 1002
http://www.w3.org/Bugs/Public/show_bug.cgi?id=1002
Gary: Validating parser flag CDL implyign sequence between things, but no way to
enforce that
SRT: Nothing there to synchronize them
Issue desc: "There is no contraints indicating how relationships in enclosed and
enclosing choreographies are related - should there at least be a
constraint that the two choreographies must share atleast one common
relationship type?"
Note comment in 1002 is in the wrong issue, should be 1102
NEW ACTION: Remove comment 2 from 1002 and place it as a comment to 1102
[recorded in http://www.w3.org/2005/03/29-ws-chor-irc]
Nick: 2 parts to look at this
... Participant is one
... Lifecycle w.r.t. chor sessions is another
... even though participants c/b different, they can all be part of one chor
instance
... Can statically see role-association? But at runtime?
SRT: I think the point is simply clarifying what happens in the event that there
is no common participant
... With a disjoint chor - or with two performs in sequence in a chor between
two different relationship types
Martin: Unless there is a common particpant, you might not get a sequence
Gary: May still be multiple relationships
... Which m/b unrelated
... Need something about each of the elements in a sequence which c/b
determined/identified before the next action occurs
... Implied coordination
Nick: Fine if happens on one node
Martin: M/b be some shared participants here
Nick: 1) One participant: sync concept, 2) Multiple participants in the same
domain: incr. sync sophistication, 3) Multiple participants across domains: Need
to sync across such domains
... If nothing in common, should chor be legal?
Martin: Yes, but need some coord mechanism underneath
Nick: Roles a, b, c: relationship between a-b1, b2-c
... W/2 relationships here, why are they needed? To what do they add value?
SRT: I think "relationship" here is a red herring
Martin: Need sync point of some sort
... C/b participant, c/b coordination protocol, c/b whatever
... We need to clarify this, however
Nick: I think the issue of having a chor session in the language will help in
this case quite a bit
... All such interactions c/b managed in the same "instance"
... Same participant?
SRT: Can indeed do it - have a chor that just deals with price extraction from
market - diff from exercising that price - only thing that connects them is a
person
... Sync of sorts
... Not explicit in chor
... M/n/b visible
Martin: This c/b sync point
Nick: This c/b chor session id
Martin: Need a clarification....
SRT: One of the problems w/such a clarification is that it w/b a negative one
m2: Use "rather than" instead of "not"
SRT: S/go into spec
... e.g. "In case of no participant or role in a chor, there is no guarantee of
order"
... This is just stating what the operational semantics are in such a situ
Martin: Need to work on some wording
Nick: But you do care w/parallel
SRT: Right, under the covers there is a sync point
Nick: We're quite silent about [such under-the-covers sync points]
Martin: Seems to warrant its own section - e.g. "synchronization semantics"
SRT: Not a bad idea
Martin: Normative or primer?
SRT: Endpoint generation is one aspect of this
... Endpoint monitoring - supports but is not part of our exit criteria
Nick: Moment intro concept of projection, there is really nothing there until
reach a sync point - we are currently silent on this - should we be?
SRT: When do projection, if have chor with A & B which may sync, only guarantee
we can make if A & B perform bits of contract is that there is virtual synchrony
between then - the items A should do w/b in correct order w.r.t. A, likewise w/B
- if there is any sync point between A & B, guarantee that they will reach that
point and synchronize appropriately, otherwise not
Nick: Globally enforce sync at projection?
Gary: [as issue raiser] preference is not to do anything at this point
Nick: We need to somehow address this issue, whether in primer or not
SRT: Suggest having text for this high up in primer, more as preface
... Not doing anything about it in spec, but to do in primer
Agreed to CLOSE wontfix, and raise new issue against the primer.
NEW ACTION: SRT to close issue 1002 and take action to add 1002-addressing
text to Primer [recorded in http://www.w3.org/2005/03/29-ws-chor-irc]
Issue 998
http://www.w3.org/Bugs/Public/show_bug.cgi?id=998
SRT: Since no good example, leave it open for time being
see http://lists.w3.org/Archives/Public/public-ws-chor/2005Mar/0024.html
http://lists.w3.org/Archives/Public/public-ws-chor/2005Mar/0023.html
Martin: Nothing broken, nothing needs clarifying here
... We have something that works
... Not keen on Tony's clarification on this one
Nick: We should vote on it and close it
Martin: Motion to close, no action
Seconded
agreed to close subject to checking with Tony.
NEW ACTION: Put this on the agenda to verify as WG agreed to close
[recorded in http://www.w3.org/2005/03/29-ws-chor-irc]
Issue 1003
http://www.w3.org/Bugs/Public/show_bug.cgi?id=1003
http://lists.w3.org/Archives/Public/public-ws-chor/2005Mar/0005.html
Gary: In Kohei's email (see Gary's pasted URL)....
... Don't reliquish channels particularly
... Prefer having an option that can specify whether or not to relinquish
channels in particular, vs. simply relinquishing the output channel when it is
passed
Nick: If have linear channel, not an issue, right?
SRT: In Kohei's example, if have usage once, then can absolutely enforce
causality
Gary: in reality, what business protocol defined in CDL would sensibly adhere to
the idea of using a channel once?
Nick: In order for Kohei to support these concepts of termination, etc., we must
have linearity
Gary: Best way to proceed - Nick send email to group outlining to Kohei what
he's seeking to clarify
Nick: Linear or once (using the channel once); Concept of localized-? - channels
are always "new" when passing
... Two intentions w.r.t. channel pipe
... Also - number of usages for a channel
Gary: Kohei's concepts - not # of usage
NEW ACTION: Nick to clarify in an email to the group that Kohei understands
particular aspects on a channel which enforce linearity per Kohei's concept -
and that we'd be delighted to see him at the next conf call to discuss further
[recorded in http://www.w3.org/2005/03/29-ws-chor-irc]
Issue1008
http://www.w3.org/Bugs/Public/show_bug.cgi?id=1008
Last email: http://lists.w3.org/Archives/Public/public-ws-chor/2005Mar/0047.html
<Gary> http://lists.w3.org/Archives/Public/public-ws-chor/2005Feb/0035.html
Gary: My proposal w/b to ensure a consistent approach between WSDL 1.x and 2.0
... The info types they point to should only indicate type
Nick: Use different names to distinguish between multiple faults of the same
type
Nick: What exactly we cannot do with CDL that requires this faultName?
Gary: Need to diff between two faults of the same xsd type in WSDL 1.1
Nick: You can use InformationType to uniquely id faults
NEW ACTION: Charlton, Gary and Nick to discuss 1008 off line to resolve
(leaving 1008 on the agenda) [recorded in
http://www.w3.org/2005/03/29-ws-chor-irc]
AOB
-------
SRT: F2F 13, 14 and morning of 15
... F2F booking done
Summary of NEW Action Items
--------------------------------------------
[NEW] ACTION: Charlton, Gary and Nick to discuss 1008 off line to resolve
(leaving 1008 on the agenda) [recorded in
http://www.w3.org/2005/03/29-ws-chor-irc]
[NEW] ACTION: Nick to clarify in an email to the group that Kohei understands
particular aspects on a channel which enforce linearity per Kohei's concept -
and that we'd be delighted to see him at the next conf call to discuss further
[recorded in http://www.w3.org/2005/03/29-ws-chor-irc]
[NEW] ACTION: Put this on the agenda to verify as WG agreed to close [recorded
in http://www.w3.org/2005/03/29-ws-chor-irc]
[NEW] ACTION: Remove comment 2 from 1002 and place it as a comment to 1102
[recorded in http://www.w3.org/2005/03/29-ws-chor-irc]
[NEW] ACTION: SRT to close issue 1002 and take action to add 1002-addressing
text to Primer [recorded in http://www.w3.org/2005/03/29-ws-chor-irc]
[NEW] ACTION: SRT to follow up on Action item 16 [recorded in
http://www.w3.org/2005/03/29-ws-chor-irc]