Minutes of the OGSI-WG Teleconference
07/Aug/2002 @ 15:00-16:30 GMT

Minute taker: David Snelling

Attendees:

Edward Boden, IBM
Karl Czajowski, USC/ISI
Ian Foster, ANL
Andrew Grimshaw, Avaki
Fred Maciel, Hitachi
Y. Masuoka, Hitachi
Andreas Savva, Fujitsu Labs
Frank Siebenlist, ANL
David Snelling, Fujisu Labs
Steve Tuecke, ANL
Peter Vanderbilt, NASA Ames
Von Welch, ANL/UC

1) Logistics for future teleconferences

    a) Teleconferences will be held on Wednesdays at 15:00 GMT.
    b) Call number will be sent prior to telecon, until a permanent
       one is setup
    c) For the next few meetings, we will also use IRC: 
       DAL network Room: #OGSI 
    d) Normally, the IRC transcript will also be published, but this
       week it was virtually empty.

    Actions:
       Steve Tuecke: With advice from others Steve will look into
       alternatives to IRC, e.g. PlaceWare, WebX, etc.

2) Logistics for interim meeting

    a) Meeting time: Sep. 4 @ 13:00 to Sep. 5 @ 15:00
    b) Location: Argonne National Laboratory
    c) There will be no teleconference on the Wednesday Sep. 4.

    Actions: 
       Steve Tuecke: Steve (or secretary) will post logistics details,
       maps, hotel, ...  Non-US Citizens: Contact Steve ASAP to
       arrange passes into Argonne.

3) New working group co-chair nomination

    a) Dave Snelling has been nominated as Co-Chair to the
       OGSI-WG. There have been several seconds and no other
       nominations.
 
    Actions:
       Steve Tuecke: Steve to take nomination forward to GGF formally.

4) Technical Discussion of "Sameness" Issue:

    a) The consensus is that "sameness" or service identity is a
       service property.
    b) A service MAY choose to provide mechanisms that allow a client
       to test for identity, see (5-l) below.
    c) Note that a client MAY choose how to interpret identity
       information provided by a service.
    d) Note that in the GSS, 'instance' is used freely and in many
       cases carries with it an assumption of identity.

    Actions:
       Open: Craft words to define sameness/identity as a service property.
       Open: Check 'instance' usage throughout GSS.

5) Technical Discussion of GSHs

    a) GSHs MUST have the syntax of URIs.
    b) The scheme of the URI determines the structure of the rest of
       the URI.
    c) The scheme is not required to define a protocol or process for
       resolving the GSH to a GSR.
    d) If a scheme puts requirements of the resolver (e.g. uniqueness
       of GSHs), these SHOULD be part of the scheme specification, and
       not mandated for all schemes.
    e) Various schemes are specified in separate documents.
    f) A service MAY have GSHs from several schemes.
    g) A service MAY have more than one GSH from a given scheme,
       although a given scheme MAY have more restrictions,
       i.e. exactly one.
    h) A service MUST have a SDE containing GSHs which refer to it.
    i) This SDE MAY be empty.
    j) A service MAY select, based on its policy, which GSHs are made
       available to which clients, e.g. for security or access
       control.
    k) A service is not required to know about all GSHs that refer to
       it.
    l) An "Is-this-also-you(GSH)" operator SHOULD be provided by a
       service.  The only possible answers are "Yes" and "I don't
       know".
    m) A GSR MAY contain a GSH for the service.

    Actions:
       Open: Craft words that reflect the above.
       Open: Extract the http and https scheme text separate documents.

6) Closed Issues Summary
      
      #26 Who controls handle namespaces? 
          Some are controlled by the instance itself, but not
          exclusively.

      #27 Multiple GSHs within one URI scheme to the same instance.
          Allowed, see above.

      #29 Operation to check if GSH (or two) refers to an instance
          Provided for, see above.

      #28 Restrictions on http GSH resolver protocol
          Specified in a separate document.

7) Open Issues

    Handles etc.

      #17 Recommend client use of HandleResolver over direct resolver
          In the teleconference, it was agreed that a client should
          use the HandleResolver portType to resolve a GSH. Whether
          the service providing this portType MAY/SHOULD/MUST be
          external or "over-the-wire" was still open. Comments to the
          mailing list please.

    Change Management
      #24 Loosen serviceType immutability statement
      #25 Version numbers vs immutable service description

    Registration
      #20 Is WS-Inspection document required in registry?

    Instance destruction
      #15 Combine SetTerminationTime and Destroy
      #14 SetTerminationTime of "forever"

    Time
      #16 Absolute or relative termination time
      #13 Use Measured TimeStamps
      #22 What if SDE lifetime attributes are not specified?
      #23 Expressing "forever" in SDE lifetime attributes

    Factory
      #05 Factory extensibility
      #19 Reduce Factory::CreateService output parameters

8) Next Meeting:

    Time/Place: Wednesday, August 14, 15:00 GMT, number/IRC will be
    sent before call.
    
    Draft Agenda:

    1) Identification of minute taker.
    2) Roll call by email to the minute taker.
    3) Approval of these minutes.
    4) Begin identifying goals/strategy for interim meeting at ANL.
        (This could wait until Aug 21)
    5) Complete Technical Discussion of Handles
        Issue #17
        SHOULD vs MUST for use of URI schemes in GSHs
        Should an http(s) GSH scheme be written? As separate or appendix?
        Assign actions for word craft.
    6) Technical Discussion Change Management