OGSA WG Phone Conference
========================

7 August 2008


Participants
------------
Steven Newhouse, Microsoft
Andrew Grimshaw, UVa
Krasimira Kapitanova, UVa
Andreas Savvas, Fujitsu 
Sergio Andreozzi, INFN
Michel Drescher, Fujitsu
Hiro Kishimoto, Fujitsu

New Action Items:
-----------------
- AI-0806a: Michel to prime OGSA WG on the CMDBf efforts for information management
- AI-0806b: Andrew to expand the port type straw man for active usage model (e.g. events, notifications)

Agenda
------
1) Early discussion
	Note taker assignment
		- https://forge.gridforum.org/sf/go/wiki1848
	Roll call
	Agenda bashing
2) Minutes approval and AI review
	Minutes approval:
		July 17 call: https://forge.gridforum.org/sf/go/doc15299
	Action Item review
		https://forge.gridforum.org/sf/tracker/do/listArtifacts/projects.ogsa-wg/tracker.action_list
3) Simple strawman portType
	Andrew provides a strawman port type definition
4) Generic query interface
	Sergio provides a reference to a work performed by colleagues in Padova about generic query interface
5) AoB

Minutes
=======

1) Minutes approval
-------------------
- Jul 17 minutes approved without changes

2) Simple Information Services strawman portType
------------------------------------------------
- Two camps favoring different architecture (e.g. WS/XML style or database style)
- Very simple strawman
- Information Services are considered a bag of things
- Hiro: How are attributes specified to select the desired information?
	-- Andrew: That would be part of the XML Query that is contained (as untyped string) in the request message
- The strawman is not supposed not to be pressed on anybody. It should be treated as a focal point to concentrate on aligning the mental pictures
- What is the relation to information dissemination services?
- What is the relation to event/notification mechanisms?
- Those use cases might be best suited being modeled in separate port types
- But what about information federation?
	-- Recursive queries?
- Federation should be implementation detail
- What metadata, e.g. scope does the provided information cover?
- Information guarantee? How “stale” is this information? 
	-- Services must advertise, “What is it?”
	-- Services must provide, “How accurate/recent is this information?”
- What should be the search criteria?
- How should the information being returned?
- How to handle the query result size? Potentially tens of Gigabytes clearly do not fit into an XML message
	-- Does this affect the query language? E.g. by allowing to specify a size threshold for the response?
- How about ordering languages?
	-- Maybe looking at the solutions of RSS for ordering criteria may help here
- HPCB Filters are defined as primitive queries on pre-defined attributes
- What types of queries do we want to support?
	-- If we find a common set of typical queries that can be modeled fairly easily, can this lead to not use a powerful generic query language?
- What are typical queries issued in EGEE information services?
	-- Anything that can be represented in GLUE is stored in EGEE IS and also queried
- Is what DMTF’s CMDBf WG doing worth investigating to solve our problem?
	-- Action Item: Michel to review and summarize the effort undertaken in CMDBf for the OGSA IS problem space
- Current strawman is a passive (i.e. pull) interface – it needs to be opened up to active (i.e. event/notification) interaction too.
	-- AI Andrew: Incorporate support for “active” interface

3) Generic query interface
--------------------------
- The interface features the notion of “dialects”
- The dialects depend on the availability of alternative renderings of the same information into the different dialects
- Example dialects are LDAP, OSG, ClassAd, etc.
- One interface is the same, but part of the query is the dialect, and the real query is given in that dialect
- The interface architecture features plug-ins for translators into the favorite dialect

4) Any other business
---------------------
- Date of next meeting
	-- 28 August 2008