OGSA Teleconference March 04, 2004
==================================

* Participants

  Abdeslem Djaoui (RL)
  Bill Horn (IBM)
  Hiro Kishimoto (Fujitsu)
  Fred Maciel (Hitachi)
  Roger Reich (Veritas)
  Andreas Savva (Fujitsu)
  Frank Siebenlist (ANL)
  Dave Snelling (Fujitsu)
  Latha Srinivasan (HP)
  Ravi Subramaniam (Intel)
  Jem Treadwell (HP)
  Jay Unger (IBM)
  Jeffrin J. Von Reich (HP)

  Minutes: Andreas Savva

* Early discussion

** Approved minutes of Mar. 1.

** Document update for GGF10

   - Propose to prepare a new version of the document for GGF10 to put
     together the latest stuff the group has been working on.
   - Due to time constraints focus on major (one section or more)
     updates.
   - Some people that probably have (or were planning) to revise/add
     one section or more: Andrew Grimshaw, Bill Horn, Abdeslem Djaoui,
     Fred Maciel, Andreas Savva.
   - Instead of passing the pen around, send updates to Hiro & Andreas
     by *Monday (Mar. 8) noon GMT*. They will do the merge and upload
     the draft in time for GGF.

   - Agreed that the glossary should be kept separate for now.

   Action: Andreas to send email to list inviting people (some by
           name) to submit section revisions. (Done)      

* Use case review

  - Jeffrin has divided Use Case document in two: 'complete' and 'tier2'
    - Complete: Reviewed
    - Tier2: Not in the right format or not properly reviewed yet.
  - Proposed to add reference from complete to tier2 and explain
    the document process.
    - Initial submissions go into tier2 document
    - After they get reviewed, checked, etc, they are moved to complete.
  - (There might be a small window for submitting complete ones for
    GGF10 -- deadline Mar.4.)
  - Some Contributors are missing
    - Add Fred Maciel
  - CDC p4
    - End user is a *direct* customer of the service not of the
      CDC. (Definition of customer used by the CDC UC.)
  - Fusion
    - p25 'hardware service' > experimental hardware services (apparatus)
    - p26 delete provisioning; the group is still not in agreement on
      the definition yet.
  - Interactive grids
    - p56: Is 'interactive grid' a correct term or should it be
      'interactive applications on grid'? The idea seems to be that
      the application has more stringent requirements.
    - Individual user interactions -> unclear, rewrite.
  - Due to lack of time continue detailed review offline.

* GGF10 session plan review
   
  - Session 3: 
    - Data Services presentation
      - Tentative. Not sure if there will be an update.
      - (Jay mentioned that Allen Luniewski (IBM) has been working on a
        WSRF update of the data services.)
    - Have presenters for Information Services 
    - Session owner Hiro. To talk offline with Bill & Abdel.

  - No security presentation in OGSA session.
    - Frank/Takuya will give presentation at AuthZ session. Focus is
      on the gap between WS-Security and OGSA-Security. Closely based
      on the VO work they have been doing in this OGSA.

  - Hiro to upload latest session schedule. (Done)

  - (Side-discussion on Grid Filesystem WG. It was approved. It is
    related to hierarchical namespace, but the entity at the end is
    always a file. They do not claim to be OGSA but would be
    interesting to pursue connection to the naming proposal.)

* Information Services

  Ref:
  http://www-unix.gridforum.org/mail_archive/ogsa-wg/2004/03/msg00014.html

  - Difference between low level and high level interface (for user)
    - low level i/f is internal to this architecture
    - high level i/f is visible from the outside

  - Not clear whether putting everything behind the same interface is
    appropriate:
    - Semantics/nuances may be different even if the 'interface' looks
      similar at the 'verb' level. E.g., there may be a high level
      'write' (or publish) but the details may differ substantially.
      - write (fire & forget)
      - write (& ack)
      - write (& callback)
    - Doing UML may help to clear up things.
  
  - Interfaces+properties (of producer)
    - E.g., a property may be QoS etc
  - At high level semantics are not fixed only common/high level
    properties.
  - This work looks correct from the taxonomy (pattern) point of view
    but might break down when going down to details (e.g., wsdl).
  
  - High level i/f on p.3 is an extension of low level i/f on p.2
  - No clear resolution. Continue discussion at GGF.

  - Dave mentioned that there was some work on breaking down
    notification functionality into a hierarchy (probably as part of
    the wsrf work). Find email/reference and forward it to the group,
    if public.