OGSA-RSS Session #1 Feb 14, 2006
Notes by Mathias Dalheimer

Introduction by Mathias
#######################
Reminder: We have a new mailing list
There is a service description document, please read it!

Donals intro of his spec V.01
#############################
-> See his slides

Previous Decission
    - RSS Services do not commit
    - no AR in scope
    - not dealing with workflow
    - does not stay how to generate candidates (up to service implementors)
    - services should be composable
    - should allow for large responses (consider candidates from the whole EGEE)
    
Specification Core
    - CSG is an abstract interface
    - maps input document to "set of candidates
    - candidates should be ordered
    
-> Some questions regarding the ordering, just to make it clear

    - Binding to WS-*: WS-Enumeration fits very well
    
EPS is language Specialization of CSG
    - Input: decorated JSDL
    - Candidate is tuple (JSDL, BES-EPR, QoS)
    - CSG Concept usable elsewhere
    
Simple Candidate Ordering Language
    - Proposal at gridforge
    - can embed other ordering languages
    - language description, see slides
    
Q Stephen: How big can the candidate set be?
A Donal: Can be large since JSDL must be concretized.

Q Marek: Do we have a weighting of the candidate properties.
A Donal: This is just a math language where you can do everything.

Q Andrew: Should the input to the CSG not a list of resources to look at?
A Donal: No. CSG will discover those.

Q Stephen: What values will win the evaluation?
A Donal: Low wins. You could use the abs function to cut out values "too" bad.

Comment Andrew: CSG finds out where it is possible, EPS figures out which 
is best. But fine if we collapse it together to one.

Marek: Plan execution for workflows, then you might need it 
-> but out of scope for the moment.

Andrew: Focus down on cases. We know we have to handle workflows, keep that in mind. But keep focus on the initial case of use cases. Here: Single JSDL jobs, and how to chunk them through the system.

Overall feedback from the group:
- We are going into the right direction
- Keeping things abstract seems to be a good idea

Marek uses the remainder of the time for his presentation of ASKALON interfaces.