John's email bounced. ---- Hiro Kishimoto From: John Shalf <jshalf@lbl.gov> Subject: Re: [saga-rg] Re: OGSA EMC BoF on Simple job submission Date: Tue, 8 Feb 2005 09:37:14 -0800 To: Bill Nitzberg <bill@computer.org>, ogsa-wg@ggf.org, Hrabri Rajic <hrabri.rajic@intel.com>, Simple API for Grid Applications WG <saga-rg@ggf.org>, Stephen Pickles <stephen.pickles@man.ac.uk>, Tom Goodale <goodale@cct.lsu.edu> On Feb 8, 2005, at 8:46 AM, Tom Goodale wrote:
Would SAGA-RG people like to engage OGSA EMC and possibly co-sponsor a Simple job submission WG? This could be a nice opportunity to tie in the classic API and the WSDL approach.
I'm not sure how relevent this would be, unless you are thinking of a direct mapping of the SAGA API into WSDL, which may not make sense, and would probably constrain the new group too much, however we should certainly make sure we have some people at the BoF.
Well, this should be cause for further discussion. The application programmer point of view on this will be "I don't really care what's underneath as long as it does what I ask of it when I call the right subroutine". So the WSDL mapping is more on the implementation side of things. That being said, having a working implementation is quite important. So it may be very interesting to pursue as a path to implementation (albeit not the only path). I think the concern on the SAGA side is that the WSDL interface will be adjusted to accommodate the semantics defined by the application use cases rather than the other way around. But this sounds pretty reasonable to me.
One thing to consider: Do we really need APIs in SRM, APME, and now in ARCH areas?
I'm not sure I understand this question ? We have APIs for the functional areas we identified from our use cases.
Just to echo this part of Tom's note, the API's are driven by the use cases. The use cases thus far didn't describe any good SRM applications (or at least SRM-like requirements were not described in enough detail for us to act upon). It could just be that the apps people are so bogged down in remote file access issues that they don't realize they need good Storage Resource Management systems yet. If we get some well described SRM, APME, or ARCH use cases, that will motivate development of APIs in those areas. (its not that those areas are not important) -john