OGSA WG Teleconference call =========================== 30 November 2006 Participants ============ Hiro Kishimoto, Fujitsu Michel Drescher, Fujitsu Chris Kantarijev, Oracle Mark Morgan, UoV Wolfgang Emmerich, University College London Fred Brisard, CA Andreas Savva, Fujitsu Rich Kao, CA Jem Treadwell, HP Fabio Benedetti, IBM Mike Behrens, RR2C Steve McGough, Imperial College Andrew Grimshaw, UoV Jay Unger, IBM 1) Early discussion =================== - Minutes approval -- Monday 27 Nov minutes approved no change - Action Item review -- AI-11.27.a: Ongoing -- AI-11.27.b: Ongoing, updates not likely until mid-December -- AI-11.09.am-a: Postponed -- AI-11.09.pm-a: Skipped -- AI-11.09.pm-c: Ongoing -- AI-11.09.pm-d: Ongoing -- AI-11.10.am-a: Ongoing, updates not likely until mid-December -- AI-11.02.a: Closed -- AI-10.30.a: Skipped -- AI-09.15.d: Skipped -- AI-08.17.a: Skipped -- AI-07.F2F.a ["EGR-WG" AI from July F2F]: Skipped -- AI-07.F2F.b ["GFox-reply-mail" AI from July F2F]: Skipped -- AI-10.23.c: Ongoing -- AI-10.23.g: Ongoing - Agenda bashing -- Info model & Service Integration --- Discussion skipped --- XQuery live demo added, which was not available at F2F 2) Info Model ============= * XQuery presentation/live demo [Notes for this section kindly taken by A. Savva] Explanation of xquery docs/jsdl/resource - uses commercial version of saxon since it supports namespaces - ( context is extremely important ) - ogsa-xq : evaluate an xquery in the context of an xml document - uses saxon:evaluate-node - question on scalability of approach since must produce an xml document? - this is just prototyping to see feasibility and what kind of minimal set would be needed - Enforce use of xquery processor? - could do xquery in xquery as this example (but not require specific xquery engine, don't have to use saxon) - could parse and transform to a native query system ( - could query a db directly ) - minimal set; stated motivation to avoid DoS attack - that's a bigger target than identifying the minimal feature set that is expressible enough; sufficient expressibility for most uses should be the target for the minimal set instead - (...) - xquery expression affected by the xml document structure that will be searched - xml encryption can be used in for example password fields; not used here for simplicity - Approach looks good; good to have a proof of concept - Approach should be documented in the info model architecture doc. - AI-11.30a: Michel to talk with Tom and Ellen on how to go about it - But cannot release code at the moment; need to check internally whether and how it can be released 3) Workflow discussion ====================== - Mark runs the group through the slides "Workflow thoughts" provided by Andrew, Steve and Mark - OGSA could either pick a specific Workflow language, and alienate everybody else - Alternatively, OGSA could pick none and create a new one, and also alienate everybody else - Orchestration and workflow must be separated - OGF is best served with simplicity - The proposal is as follows -- stay away from language and API specification -- Tightly scope WFDL (Workflow Definition Language) -- Ensure reasonable mappings from existing systems into WFDL -- Toos support -- Reuse as much existing as possible - WFDL is XML based and describes directed graphs - Nodes are WFDL, ord JSDL - Nodes have labelled input and output - Mark walks the group along a sample WFDL instance document - Jay: This example is a dataflow graph, not a workflow graph -- Mark: Example 2 demonstrates other kinds of dependencies than data dependencies - Jay: Dependencies should be expressed as resources produced and resources consumed - The group discusses Jays suggestion - Jay points out that the Workflow effort should be very carefully evaluated against the current other efforts related to JSDL - Chris: Defining (yet another) WF language is probably a bad idea -- This would, among others, add delay -- Andrew: As pointed out on the slides earlier, picking a particular WF language or creating one is a matter of "choose your poison". -- Fabio: We probably should focus first on listing the requirements the languuage should meet -- Jay: I believe it is possible to combine this workflow effort with current efforts for JSDL and Information modelling - The group discusses possible further steps from here on - Andrew: After this first initial investigation, does OGSA think that the WF issue is important enough to set forth a design team now, or postpone for the moment - Andrew proposes to set up a separate phone connference with a specific WF agenda -- Jav and Michel support the idea - The group discusses the appointment for the next WF call - Jay proposes to use OGSA's Gridforge wiki to collect people's requirements, dependencies, favourite existing WF solutions, etc. together to get an overview on what's necessary and what's not. -- The group reaches rough consensus to picck up this idea Any other Business ================== - Date of next call: Thu 7 December 06 1 pm GMT Sumary of new Action Items ========================== AI-11.30a: Michel to talk with Tom and Ellen on how to go about documenting his current proof of concept in the Info model Arch doc.