David, Andrew, Aleksander and All, Concerning the document 'GES Realization via Existing Specifications' : I do NOT know if Aleksandr is kidding, but I agree with Andrew : RNS is NOT a requirement. Requirements are : - Compatible AUTHN credentials, and compatible AUTHZ credentials (covered in another thread, but NOT to be forgotten), - GLUE 2 to reference grid entities, - A well defined States and Transitions model (State Machine) for jobs, with precise semantics, - Some BES specification to define user operations on jobs, job states and job transitions, with precise semantics, - Some JSDL specification permitting the user to describe job requirements, data staging and job execution. I think that HPC-BP and HPC-BP FSE are useful recommendations permitting implementation, as described by Andrew, but I suppose that they are NOT requirements. I am quite sure that WS-Addressing and WS-Naming are NOT requirements, and I do NOT know at all how useful these specifications are for us. As I already wrote in my mail dated 13 May 2009, I understand RNS as 'syntactic sugar' very useful for the end user, but belonging to a higher level (presentation) layer, and NOT required for PGI. Finally, we need a diagram of dependencies between the different specifications which Andrew proposes. So I would ask : - David to send to the PGI mailing list the slides 26, 27 and 28 of the presentation 'Standardisation: Recent progress review and best-practices sharing: Middleware Track' which he performed at the 6th e-Infrastructure Concertation Meeting in Lyon on 24 November 2008 (or improved versions). - Andrew to present a diagram of dependencies between the different specifications he proposes, indicating which ones are really required. He can take as example the above mentioned slides. Thanks in advance to David and Andrew for their help. Best regards. ----------------------------------------------------- Etienne URBAH LAL, Univ Paris-Sud, IN2P3/CNRS Bat 200 91898 ORSAY France Tel: +33 1 64 46 84 87 Skype: etienne.urbah Mob: +33 6 22 30 53 27 mailto:urbah@lal.in2p3.fr ----------------------------------------------------- On Sat, 16 May 2009, Aleksandr Konstantinov wrote:
On Friday 15 May 2009 16:54, Andrew Grimshaw wrote:
Aleksandr, RNS is not a requirement. It shows up in my "GES Realization via Existing Specifications" as a means to meet the requirements. I personally think it is a good idea and allows us to work with an existing code base and access layer. Listing a directory of things is pretty common.
Looks like a good idea for me (personally). I think You should make one more logical step and suggest to drop BES. No kidding. As You explained usage of RNS diring last telecon it can do anything BES does. One simply has to provide some data transmission capability to trasfer bigger chunks of data to/from nodes (not sure about term) presented by RNS. And that could be same ByteIO proposed by You or even simpler - HTTP(S) which is already used as underlying protocol of SOAP anyway.
Actally we are using similar approach in ARC (production version) except that it uses GridFTP (and hence TVFS) instead of RNS.
Concerning "work with an existing code" I'm not sure. I doubt many of participating projects have an implementation of RNS. And those developed probably won't be very reusable in different environment. Do You have an implementation for libxml2? On another hand AFAIR RNS is not a complex interface and wouldn't take much effort to implement.
A.K.
A
-----Original Message----- From: pgi-wg-bounces@ogf.org [mailto:pgi-wg-bounces@ogf.org] On Behalf Of Aleksandr Konstantinov Sent: Friday, May 15, 2009 9:32 AM To: pgi-wg@ogf.org Subject: Re: [Pgi-wg] Promised document
On Friday 15 May 2009 13:10, David Wallom wrote:
Hi All,
Can we first agree (on the list, possibly with a doodle vote) that the requirements described in Andrews document were accurate. At this first stage please ignore the implementation, just are the requirements correct and if not what changes are required. Is RNS requirement or implementation?
A.K.
David