OGSA January 2006 Interim Meeting
=================================

  Location: Sunnyvale, CA
  Date:     16/1/2006, morning
  
* Participants

  Hiro Kishimoto
  Dave Snelling
  Fred Maciel
  Andreas Savva
  Fred Brisard
  Patricia Kovatch
  Chris Jordan
  Dave Berry
  Ravi Subramaniam
  Steven Newhouse
  Steve McGough
  Darren Pulsipher
  Takuya Mori
  Mark Morgan
  Andrew Grimshaw
  Jay Unger
  Tom Maguire
  Ellen Stokes

  Minutes: Andreas Savva

* Welcome and review of F2F agenda

** EMS breakdown proposal

  - Discussed proposal for a separate EMS track (2-3 people) to work
    on sequence diagrams.
  
  - Agreed to a short break out session in the afternoon to discuss
    what can be done: Andrew, Ravi, Darren, Steve M, etc
  - With the option for a longer breakout session on Tuesday

  [Note: In the end there was no parallel EMS session during this
  F2F.]

* BES-WG

** Overview of BES specification draft

   - EPRs are now a MUST and WS-Names are a SHOULD. The type used is
     the same (EPR).
   - The intention is for the WSRF rendering to be WSRF BP 1.0
     compliant. But it is not 100% yet.
   - Container is defined. Activity not defined. Influences what kind
     of operations are specified.

   - State diagrams is based on JSDL 1.0
     - Especially the stage-in/stage-out diagrams. But it is the BES
       decision to include those portions of JSDL in its
       specification.
       - Note that the staging state diagrams are not defined as
         sub-states. They are different state diagrams
     - Discussed and REJECTED: Modify the staging state diagram to
       simplify so that there are no separate
       staging-in/waiting/staging-out but a single staging state
       - Motivation is to keep a simpler (perceived) mapping to the
         JSDL spec 
       - It was also pointed out that the JSDL spec makes no statement
         on the state diagram for staging and that the authors did not
         intend a data staging state diagram as defined here.
       - People might still choose to comment during pc

   - Some states are optional. An activity does not need to (visibly)
     pass through them. (But see "Events, if supported, ..." further
     down.)
   - When an activity is suspended, the staging activites are also
     suspended.
   - What happens when a staging activity suspends?
     - The overall activity suspends, which in turn causes the rest of
       the activites to suspend. (Added clarification)
   - Suspend should be clarified further. Proposed text along the
     lines of "... stops the consumption of further resources, but
     does not necessarily release the resources."

   - start/stop container: Proposal to put them in a separate porttype
     to aid composition. (Rough consensus to do so.)

   - Added clarification: Events, if supported, are emitted per state
     transition. And even if the activity does not 'actually' go
     through a state an event should be triggered. E.g., if no data
     stage in occurs, an event is still emitted before going into
     execution.

   - Discussed rendering and notification types
     - Rendering: explicit that other renderings can be used
     - Notification: explicit 'not-supported' and also other types can
       be used.
       - Could a service support more than one? 

   - Discussed security considerations and how much to include in this
     document.
     - Proposal to add more detail on the issues (identity mapping,
       etc.) and state that they are out of scope. (Rough consensus)

*** Rendering
   - Fault definition: 
     - Proposal to do the definition as WSRF Base Faults, and then do
       a description for the WS-I rendering how they map to SOAP
       faults (used in ws-management).
     - An alternative is to do a new spec: OGSA Faults

     - Both REJECTED: Agreed to leave as is with parallel renderings
       from an 'abstract' description; and trying to make XML
       definition of {base faults, soap faults} as close to each other
       as possible

   - Resource advertisement: The WS-I rendering uses UDDI (mostly
     static information +'alpha' for the dynamic portion.)
  
*** Other issues
   - Resource discovery (resource information): It is out-of-scope of
     this specification. The choice of which BES Container to use is
     assumed to have already occurred. 
     - Another spec could provide the needed functionality.
   - Notification: It has been decided not to filter at the
     container. Filtering can be done further down.

*** Next steps

    Moving to a final call of the document on the list and
    teleconf. Then submit to the GGF Editor.

** Report on Interoperability meeting
  
  - Carried out at Imperial College between UVa (WSRF .NET
    implementation) and GridSAM (WS-I based)
    - Each project had its own test harness. Cross-checked test
      harnesses to test that both containers could run the tests of
      the other team.
    - Nothing specific to WSRF or WS-I rendering tested.
    - Tests worked first time (with the exception of a minor coding
      error)
    - Only correct functionality was tested. No error checking. Also
      no events.
    - It should be taken as a proof-of-concept, and working towards a
      formal interop. And towards an experience document.
  - Multi-grids interop to do both BES and JSDL and could lead to
    experience documents for both specs.