OGSA F2F Session – 10 19 2007 – DMI (Data Movement Interface)
=========================================================

Participants

Hiro
Steven N
Duane
Fred M
Michel
Donal
Steve McG
Joel
Andreas
Chris Sosa (UVA)
Fred B 

Minutes – Chris Sosa

Summary of Actions:  Look at BES terms (from the BES spec) and see if we can get states so we can create uniformity across specs.  Also a call for action to Review Current Draft, Participate on weekly telcons, Work on imp and Review submitted draft.

How deeply do we want to define the interface, etc.  How much do we want to restrict implementations?

Consensus – Use EPR’s 

Problem:  Many ways to define interface.

-Already have EPR’s, etc.  
-No need for separate notion of another abstract data handle
-Use Data EPR = DEPR but not specified by OGSA-DMI (EPR + misc metadata)


Problem:  How to deal with 3rd parties (not just src-sink)
- Movement fractures the EPR.

Bad to reuse names between different related standards (but we should if it’s exactly the same) –might want to borrow terms from BES to apply here (unified state model so that manager (workflow

Problem:  State model is very similar to BES state model
Action – Look at BES terms and see if we can get states so we can create uniformness.

We need more detailed information about the failed state (DMI) needs more information about the clean-up.  Diff strategies work for diff protocols.  What do we do with tainted state?

BES model has sub-states (maybe borrow that notion as well)

Open Issues:
- Where oh where do we get the EPR’s?
- Support for diff protocols (part of implementation probably not of the specification)