Minutes of the Seventh OGSI-WG Teleconference
25/Sep/2002 @ 15:00-16:30 GMT

Attendees:

Tim Banks <IBM>
Karl Czajkowski <ISI>
Shel Finkelstein <Sun>
Steve Graham <IBM>
Bill Horn <IBM>
John Karpavitch <Avaki>
Fred Maciel <Hitachi>
Andreas Savva <Fujitsu>
Frank Siebenlist <ANL>
David Snelling <Fujitsu>
Steve Teucke <ANL>
Peter Vanderbilt <NASA>

1) Approval of minutes form Setpember 18th. OK

2) Review of pending actions:

Action: Steve Graham to drive the creation of an XML Schema for the
specification, #35. - Pending

Action: Steve Graham to craft language for recommendations to
developers for use of resolvers. - Pending

Action: Dave Snelling and Andrew Grimshaw to look at requirements and
solutions for pull like semantics. - Pending

Action: Thomas Oinn to summarize the meaning of the presence/absence
of service data content. - Done

Action: Shel Finkelstein to write a working definition of Client and
some case studies where QoS properties become critical - Done

Action: A. Grimshaw to see if the dissertation referred to can be
pushed to the whole mailing list - Done

3) Review of plans for GGF

Tuesday, October 15
  07.00 - 08.00  GFSG meeting
  08.30 - 09.30  Opening Plenary
  10.00 - 11.30  OGSI WG Update
  12.00 - 13.30  OGSA-SEC WG
                   &  DAIS WG session 1
  14.00 - 15.30  OGSI WG Technical Discussion
  16.00 - 17.30  OGSA WG
  19.30 - 21.30  Social event

Wednesday, October 16
  07.00 - 08.00  GFSG meeting
  08.00 - 09.30  DAIS WG session 2
  10.00 - 11.30  OGSA WG
  12.00 - 13.30  OGSI WG Technical Discussion
  14.00 - 15.30  OGSI WG Technical Discussion
  16.00 - 17.30  Unscheduled (Globus people have R&D 100 award ceremony)

Thursday, October 17
  07.00 - 08.00  GFSG meeting
  08.00 - 09.30  OGSI WG Technical Discussion
  10.00 - 11.30  Area meetings
  12.00 - 13.30  Closing Plenary

4) Technical Agenda:

Discussion focused on two basic questions, although some effort was
required to separate these issues from each other.

a) Do we need a portType level mechanism for describing the QoS
expected by a GridService?

The key issue here was whether sufficient QoS information was
available as part of the binding options available at the grid service
or if some QoS information needed to be provided independent of the
binding. For example, Grid Service X expects "guaranteed delivery"
semantics on notification deliveries. The alternative argument was
that the GS should only be published with stated bindings that meet
these requirements. The problem remains that someone wishing to adapt
the portType to a new binding has no way of knowing what QoS
properties of the binding that portType relies on. The issue remains
unresolved.

Action: Shell will draft a SDE for consideration that captures some
QoS information and acts as a place holder for more detailed
development later.

Action: Steve T. will propose a binding layer project to OGSA WG.

Action: Proponents of specifying QoS at the binding only are asked to
answer the following: How does reliable mesaging work in the absence
of portType specification of QoS? How does the client/service
request/provide it?

b) Should we add a DeliverWithAck operation to the NotificationSink
portType?

Related to QoS, but with a more narrow focus, there was some concern
that the very relaxed model for notification delivery was not a
sufficient basis on which to build more sophisticated notification
frameworks. A DeliverWithAck operation was suggested (it had been
present in an earlier draft). This would provide a better basis for
building reliable messaging than the current DeliverNotification. With
out it, services requiring complex messaging services would need to
start from scratch rather than build on this standard. The issue was
not resolved.

Action: Dave will look into a DeliverWithAck option. One issue is
specifying which delivery operation should be used.

c) Open issues pertaining to Notification...