RE: [ogsa-wg] RE: GRIDtoday Edition: Tony Hey: 'Challenging Times for GGF & Standards'
Tony's email bounced. Tony: You can read whole mail threads triggered by your article at; http://www-unix.gridforum.org/mail_archive/ogsa-wg/2005/02/threads.html http://www-unix.gridforum.org/mail_archive/ogsa-wg/2005/03/threads.html There are more opinions than you've already read. ---- Hiro Kishimoto Subject: RE: [ogsa-wg] RE: GRIDtoday Edition: Tony Hey: 'Challenging Times for GGF & Standards' Date: Thu, 3 Mar 2005 11:37:06 -0000 From: "Tony Hey" <Tony.Hey@epsrc.ac.uk> Dear OGSA WG members and others, I am interested by this debate (obviously) but fear that I did not get my essential point across. Let me try once more - sinc e I think this is a crucial decision point for the whole GGF and Cyberinfr astructure effort. First, to reassure Dave Snelling and others, I am NOT trying to endorse 'proprietary' proposals such as WS-Transfer from Microsoft; nor am I trying to impede the progress of WSRF etc through OASIS. IBM and others have indeed acted as I had urged and submitted their p roposals to a recognized standards body. Second, the point is not about whether one can implement WSRF in .Net technologies. I respect the work of Marty Humphrey and h is team but this also is not the point at issue. The point is not about how well the WS-RF and WS-Trans fer stacks compare but rather whether it is always appropriate to use the "messages are directed at single resources" approach? Many people, includi ng people whose technical judgement I respect such as Tony Storey, Ian Foster, Dave Snelling and others, apparently believe that the answ er to this question is "everywhere: it is the foundation for all op erations on all services". It is therefore not surprising that this group do not see the need to worry about the question "is it a good idea to build architecture around the idea of sending messages to single resources?" In fact, although I am clearly not an implementor, I t hink that this is not a closed issue. People whose technical judgement I a lso respect - such as Jim Gray, Pat Helland and Andy Herbert (yes, a ll from Microsoft but wait for the supporting cast) - believe that this is not a good way to build robust wide area, inter enterprise distributed sy stems. I understand why opinions from 'Microsoft' may be viewed wit h suspicion but in this case I think their views are deeply held matters of principle. Andy Herbert, for example, is from Microsoft Re search and spent a long time trying to make distributed systems lik e CORBA work. He came to the conclusion that while CORBA was just fine for local area networks it encouraged a style of programming for wide a rea networks that led to such implementations being 'brittle'. Thus I believe that they think that while something like WSRF may be fine for use within corporate computing centers behind corporate firewalls, this approach to building distributed systems will not lead to robust wide area implementations. (I hope Andy and the others cited will either support or correct my interpretation of their views.) So I think this is not a small matter of WSRF versus WS-Transfer. This view is supported by others in the distributed computing community - not only people like Paul Watson, Savas Parastatidis and Jim Webber, but also professional database people like Dave Pearson from O racle who I believe has concerns that in an inter enterprise computing context, many companies will not wish to expose details of their datab ases to open scrutiny. What changed for me recently was the realisation after months of trying to persuade Microsoft to engage with GGF/OGSA/WSRF that this issue was actually a matter of principle and that there was no prospect of Microsoft being able to join the WSRF bandwagon or of their 'morphing' their ideas to be compatible. This being the case - and it was confirmed for me by the OGSA-UK meeting I referred to - and recognizing that the non-WSRF folks might, just might, be proved correct in the long run, at least for certain types of e-Scien ce applications, then for GGF to tie its 'flagship' OGSA to this approa ch to resources by mandating the WSRF route would be to take an unnecessary high risk with our user community. Such a route would also be highly likely to effectively exclude Microsoft from GGF and hinder their pl aying any positive role in GGF, Grids and Cyberinfrastructure. In addition, speaking as someone with responsibility for t he UK e-Science applications people, I can state categorically tha t they are not interested in religious debates about how the distribu ted systems they rely on have been implemented. They just want the underlying distributed middleware to work and for GGF to concentrate on delivering timely and high level standards that will assist them in constructing their interoperable Cyberinfrastructure. They absolutely do no t want to be forced to choose between an 'IBM' camp and a 'Micro soft' camp - they just want a standard on which they can build higher le vel standards and APIs. They want to be insulated from any low-level dispu tes about how to fix the plumbing. So that's it folks. I believe it really is a crucial time for GGF to show that it is really an open standards body dedicated to delivering useful standards for users - rather than backing one bra nch of a religious debate right down at the plumbing level that i ts user community does not want to know about. However, is it possible to define OGSA in such a way that it does not mandate the WSRF approach to resources? Well I have talked with Andrew Grimshaw and o thers and Andrew believes that this might indeed be possible. I would the refore ask you to be open-minded and assist such an investigation. Obviously the first implementations of OGSA will be WSRF based and I expect many UK projects will be very happy to adopt WS RF, GT4 and a WSRF version of OGSA-DAI. What about a non-WSRF implementation? All I know is that Paul, Savas and Jim investigated alternative, non-WSRF ways of providing OGSA-like functionality in a small UK project - but I suspect that the final word has not yet been said on how to provide such implementations. I do not want to de-rail WSRF - as I said it will certainly be adopted by many UK projects. I just want GGF/OGSA to leave the door open - so that if it does turn out that the WSRF way is not so good for inter enterprise applications - then GGF and OGSA are not made obsolete. Ian often asks me why the UK are so keen on Web Services . I used to reply that Web Services were significant only because Microsoft and IBM had agreed to agree - the Men In Black philosophy. Following this philosophy it is surely self-evident that GGF, Grids and the whole Cyberinfrastructure agenda would be much, much stronger if both Microsoft and IBM were engaged. This is why giving some space between OGSA and specific implementations using WSRF is so importa nt. I should end by noting that these are my own opinions. Furthermore, the people I have explicitly cited may or may not agree wi th the opinions I have attributed to them. I hope at least that my views and intentions are now clear. Tony -----Original Message----- From: Frank Siebenlist [mailto:franks@mcs.anl.gov] Sent: 01 March 2005 07:08 To: Ian Foster Cc: Dennis Gannon; Samuel Meder; ogsa-wg; Tony Hey Subject: Re: [ogsa-wg] RE: GRIDtoday Edition: Tony Hey: 'C hallenging Times for GGF & Standards' So... if ws-transfer and ws-properties are sets of convent ions that do essentially the same thing, then it's actually a great c ompliment to ogsa's choice of wsrf that MS decided that it needs a similar functionality! Do I understand that Tony essentially asks the OGSA-WG t o come up with an additional abstraction layer that allows one to model the usage patterns such that it can map to either wsrf or ws-tra nsfer implementations? ...but if we model the usage patterns with wsrf, and w srf and ws-transfer are doing conceptually the same thing, isn't t he ogsa-wg doing just that: it uses an abstraction that can be us ed to describe the relevant usage patterns without losing any generality. The idea of inventing "yet an other abstraction layer" d oesn't sound very productive. -Frank. Ian Foster wrote:
Dennis:
I'm not sure that the "we don't need WSRF" is th e heart of the debate.
If it was, then I think things are fairly clear:=2 0W SRF is just some conventions for the messages that you send to do c er tain things (e.g.,
getResourceProperty to get state, Terminate to destroy=2 0s omething, or whatever the names are) in a WS context. If you do n't have those conventions, then everyone ends up defining their own,
so that e.g. a
job management interface might have "getJobStatus" and=2 0" destroyJob", a
file transfer interface might have "getTransferStatus" a nd
"destroyTransfer". This lack of consistency just makes=2 0l ife difficult,
without providing any benefits.
The debate with MS, as I understand it, seems to=2 0r ather relate to the fact that they are promoting a *different* set of co nventions for doing similar things, e.g., WS-Transfer instead of WS-ResourceProperties.
Ian.
At 10:23 PM 2/28/2005 -0500, Dennis Gannon wrote:
hi Sam, i don't think MS has any orchestrated view on WSR F at all (but i may be wrong.) I think it is more the case that th ere are people working
on grid standards (outside of microsoft) that feel th at what exists in the ws-spec world is sufficient. hence if there 20 is any onus, it is
on those folks to show us that this is true. =2 0w hat tony is saying is that users, i.e. application builders, should not hav e to deal with these details. The should see clearly defined OGSA se rvices and they should have an easy to understand set of interactio n patterns to use these services to build thier applications. the OG SA point of view is that to be precise in the definition of these 20 behavior patterns requires a framework like wsrf.
i actually feel that these things can all coexist. 20 but from the politics of "what is simple", we seem to live in 20 interesting times.
dennis
On Mon, 28 Feb 2005, Samuel Meder wrote:
On Mon, 2005-02-28 at 14:38 -0800, Frank Siebenli st wrote:
Could anyone summarize MS' WS-view, and how i t differs from WSRF?
So far I have not seen any substantial differen ce between the two approaches and I definitely believe the onus is=2 0o n MS to show why people should adopt their proprietary specifications vs. a do pting something that is being developed in a open sta nd ards body, is getting very close to a 1.0 version and has m ul tiple implementations behind it.
/Sam
Thanks, Frank.
Hiro Kishimoto wrote:
Hi all,
> >Absorbing article by Tony Hey.
<http://news.tgc.com/nview.jsp?appid=3D360&print=3D1#342708>
---- Hiro Kishimoto
GRIDtoday
NEWS AND INFOR
MA TION FOR THE GLOBAL GRID COMMUNITY
--- February 28, 2005: Vol. 4, No. 8 ---
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ SPECIAL FEATURES =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
=3D =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
[ ] M342708 ) WSRF? WS-*? Where is GGF's
20 OGSA Headed?
By Tony Hey
, Contributing Editor
Tony Hey, director of e-Science for the
Engineering and
Physical Science Research Council, continues to
elaborate the
need for open standards in the realm of W eb services-based Grid
computing. He discusses the great debate of=2 0W SRF vs. WS-*, and lays out what the GGF must do with OGSA in order to give e-Scie nc e application developers something to rally around.
-- Sam Meder <meder@mcs.anl.gov> The Globus Alliance - University of Chicago 630-252-1752
_______________________________________________________________ Ian Foster =2 0 www.mcs.anl.gov/~foster <http://www.mcs.anl.gov/%7Efoster> Math & Computer Science Div. Dept of Computer Sci en ce Argonne National Laboratory The University of Chic ag o Argonne, IL 60439, U.S.A. Chicago, IL 60637,
U.S.A.
Tel: 630 252 4619 Fax:=2 06 30 252 1997 Globus Alliance, www.globus.org <http: // www.globus.org/>
-- Frank Siebenlist franks@mcs. anl.gov The Globus Alliance - Argonne National Laboratory ********************************************************************** Internet communications are not secure and therefore EPSRC does not accept legal responsibility for the contents of this message. Any views or opinions presented are solel y those of the author and do not necessarily represent those of the EPSRC unless specifically stated. All EPSRC staff can be contacted using Email addresses w ith the following format: firstname.lastname@epsrc.ac.uk **********************************************************************
participants (1)
-
Hiro Kishimoto