Minutes of the Data Design Team telcon, 15th September 2004.

0) Summary of actions
- Dave to arrange telcon that includes GFS-WG & EGEE.
- Everybody to review the naming requirements documents
- U. Virginia team to put their implementation on the web.
- Dave to chase Brian and Malcolm re data movement.
- Mark to produce a use-case scenario for EMS use of metadata.
- Other assignees to produce use-case scenarios for use of metadata.
- Andrew, Allen and Dave to continue with a strawman architecture.


1) Early discussion (10 min)
      Roll call 

Dave Berry, NeSC (chair and note taker)
Mario Antonioletti, EPCC
Simon Laws, IBM
John Easton, IBM
Andrew Grimshaw, U. Virginia
Allen Luniewski, IBM

The minutes of previous meetings were accepted without comment.

We added an agenda item to discuss the U.Virginia implementation of Contexts and BasicFiles.
        

2) Update on actions arising from FTF (10 min)
	
- Andrew has talked further to GFS-WG (at GGF).  He reports a possible disconnection w.r.t. level of naming.  Andrew decided to implement the simple approach to show what it looked like.  Follow-up action: Dave to arrange telcon that includes GFS-WG & EGEE.

- Allen's work federation is on hold awaiting (a) feedback or (b) the straw man data architecture.

- Malcolm has done some work on data movement under the aegis of NextGrid, but has no specific report.  Follow-up action: Dave to keep pursuing.

- The Metadata group has grouped the use-case scenarios into categories and assigned to people to work on.  These will provide concrete examples and answer general questions.

At this point there was some discussion of whether we should have a general metadata architecture?  Or define a general format?  Clearly we need to integrate with EMS.
Andrew noted that Legion, Condor, Globus, etc. are more focussed on metadata about resources. They use (Name,value) pairs; then have operations to gather metadata, subset it, and search it.  Action: Mark to produce a use-case scenario for EMS use of metadata.

- Dave has sent an e-mail to Andrew and Allen to trigger a discussion of the strawman architecture.  Follow-on action: to continue with a strawman architecture.


3) U.Virginia implementation of Contexts and BasicFiles.
- WSRF.net (not specific to this implementation of WSRF).
- Uses WSRF EPRs until OGSA naming resolved.
- Contexts implemented as WSRF Service Groups.
- Client side: like a port of "ls",
- BasicFiles: read, write, append, truncAppend.

Q. can we document key architecture decisions:
- offset held on client and passed as parameter
- entries in contexts have a little metadata (name, type (file, directory, ...))
- type stored in context because we can't (in WSRF) guarantee each EPR has appropriate resource properties
- "stat" properties on files as resource properties of each file
- Client side handles r/w status of files
- How to handle atomicity - effectively guarantees - policies that we can define.
- Position: has to assume binary file.

Action: U. Virginia team to put their implementation on the web.


4) Update on naming
- GGF agreed that it's important.
- Handle.net a possibility
- Various feedback: including use cases.  In current version of document on GridForge.
- Next steps for the Naming team: Hiro to arrange weekly telecon.
- Action: everybody to review the naming requirements documents.


5) Towards a strawman architecture (30 min)

We briefly discussed how to unify POSIX access and query access.  Andrew said that Avaki does this.

Other discussion was postponed until people have had a chance to respond to Dave's e-mail.


6) Wrap up

DONM - Possibly Oct 13th, if GFS-WG and EGEE can be engaged.