[cddlm] GGF17 Report

[Hiro Kishimoto]

Hi Jun,

My slide presentation at CDDLM-WG session attached.
----
Hiro Kishimoto

Jun Tatemura wrote:
> Dear CDDLM team,
> 
> In the session, I have presented
> - current document status + CDL public comments
> - interoperability status
> - interoperability demo (NEC client to access NEC/UFCG/HP portals)
> followed by discussion for OGSA/EMS
> - basic POSIX components for EMS
> lead by Hiro.
> Please find the slides and note (by keisuke fukui) attached.
> 
> Best Regards,
> Jun Tatemura
> 
> 
> ------------------------------------------------------------------------
> 
> 
> /// CDDLM: 
> /// Jun Tatemura+ about 10 attendees
>     - CDL
>     - Deployment API
>     - Component Model
>     Document Status 
>     - Two specs published
>     - One spec in public comment period
>     -- CDDLM CDL (- May 26)
>     CDL Document Clarifications
>     [1] Import namespace
>     [2Typical use case of @cdl:lazy
>     Question
>     - Then what will happenwhen the user assigns a value to a property with 
>       @cdl:lazy?
>     - Is this allowed?
>     Answer
>     - You (the user ) can assign a value
>     - The reference to the property is resolved at deployment time, anyway
>     - How this is interpreted by the component depends on the component
>       (application specific)
>     - Clarifying statements will be added to the spec
>     Hiro: These two comments are answered already on the tracker?
>     Jun: The first one is done. Not yet for the second. will do after the   
>           session.
> 9:15
>     Interoperability: Three Specs to Test
>     CDL Interoperability Definition
>     CDL Test Environment
>     Toolkit for CDL Component Tests
>     Test Document
>     CDL Test Status
>     Hiro: Do you think standard XML normalization doesn't solve the 
>           problem?
>     Jun: An Example. When doing reference the same prefix can mean 
>          different namespace depending on the 
>     Hiro: This is a common problem to XML document.
>     Jun: Yes. That's why this is not mentioned in CDL spec itself. The 
>           implementation needs to care to do correct namespace 
>           interpretation.
>     Hiro: Is to compare many implementation can generate equivalent
>           document?
>     Jun: Yes. 
>     Hiro: Is there any test planned for deployment API?
>     Jun:  It addresses to component model, while CDL is more general.
>           The test doesn't cover deployment API usage of CDDLM 
>           implementation.
>     CDDLM Interoperability
>     Our Test Approach
>     Deployment API test
> 9:35
>     Deployment API Test Demo
>     - NEC (Muse)
>     - UFCG (Muse)
>     - HP
>     (There are less problems between NEC and UFCG because of this)
>     Component Model Test
>     HP Reference Implementation
>     HP Testing Status
>     UFCG Reference Implementation.
>     Testing Issues
>     - Security: need to restrict compoenents to deploy.
>     - Need well known compoenenets & <cdl:import> URLs.
>     - Without standardised failure codes for certain faults (e.g. unresolved 
>       reference) failure tests just test for failure
>     Underlying spec issues
>     - WS-A is complex
>     - Faulting: SOAP Faults vs BaseFaults. SOAP
> 9:45   (So this is the status. Questions?)
>     Hiro: What about the implementation from Stuart.
>     Jun:  It will come later. It might present new underlying problems since 
>           it is the Windows implementation.
>     Hiro: Understood that CDL test doesn't use real application. 
>     Jun: Applications used for to verify the component model don't do real 
>       task. Document named "Component Test Model" is explained.
>     (Question about application and component continues.)
>     Keisuke: Is that box in the diagram a compoent resource or XML fragment?
>     Jun: It is a compoenent resource. It can be Java code, but it is 
>         implementation dependent. 
> 
> 9:55  // Basic POSIX component for EMS
>       // Hiro Kishimoto
>       The Basic POSIX component is a generic component usable to deploy POSIX 
>       applications.
>       Deployment Infrastructure
>       BLAST CDL Document
>       Deploy Component
>       Jay: I don't want the use to specify all parameters when writing 
>            scripts to do transfer, copy, set attribute, and so on.
>       Jun: Script needs to be somehow parameterized to use the value from 
>            CDL.
>       Basic POISIX component
>       - Identified "gap" between family of CDDLM specs and EMS services
>       - Provide services
>       -- Application binary installation
>       -- File system mount
>       -- Database setup (in case of BLAST)
>       - Provide output (will be incorpoated in JSDL document)
>       -- Application path
>       -- Environment variables
>       CDL document includes
>       - 1. Application binary installation
>       - 2. File system mount
>       Hiro: I think Filesystem mount will be of value for many. Is there 
>                 any plan to privde it?
>       Jun: Not yet, I think.
>       Basic POSIX template
>       - Companion template to Basic posix component
>       Next steps
>       - Jointly develop BPC and BPT spec
>       -- OGSA EMS & CDDLM-WG
>       - More example than BLAST
>       -- OGSA EMS & JSDL-WG
>       - Put Job Manageer in this deployment picture
>       - Put ACS in this picture
>       - Advanced Common Component
>       Hiro: How the components can be stored in ACS?
>       Keisuke: One of the goal of ACS is to bundle a set of files for a task 
>             as a unit. When retrieving contents, one can do that at the 
>             granularity of the contents inside an archive. I can imagine 
>             components can be archived and later retrieve each. So I think
>             it is possible with the current ACS at this moment.
> 10:20 End of session

-------------- next part --------------
A non-text attachment was scrubbed...
Name: Basic POSIX component 20060508.ppt
Type: application/ppt
Size: 74240 bytes
Desc: not available
Url : http://www.ogf.org/pipermail/cddlm-wg/attachments/20060519/e83e0733/attachment.bin