Please see inlined comments <AD/>. -----Original Message----- From: owner-ogsa-wg@ggf.org [mailto:owner-ogsa-wg@ggf.org]On Behalf Of Marty Humphrey Sent: Wednesday, June 22, 2005 4:57 PM To: 'Ogsa-Wg' Subject: RE: [ogsa-wg] OGSA-MWS-BOF at GGF14 on Tues June 28, noon-1:30 I want to clarify this ... Yes, composability is certainly important, both with regard to the specifications/standards as well as the profiles. With regard to composability of *profiles*, everyone agrees that WS-I BP 1.1 supports composability, and WS-I "Simple SOAP Binding Profile 1.0" and WS-I "Attachments Profile 1.0" support composability, but no one would attempt to layer WS-I "Simple SOAP Binding Profile 1.0" onto WS-I "Attachments Profile 1.0" (or the other way around). So they're not "composable" in this sense. <AD> Yes I agree the change from BP 1.0 to 1.1 was partly driven by a need to allow COMPOSITION of RELEVANT profiles. For the the example 3 profiles above: If your web service is not using attachement you would compose BP 1.1 with SSBP1.0, if it is using attachements you will need to compose BP1.1 and AP1.0. These can be seen as different bindings.
From a user point of view these are the only meaningful compositions (you either want to use SOAP with attachement or without.). When there is a tight coupling between profiles, the profile document itself should state how it is intended to be used or composed with other profiles. If other binding protocols become available (MTOM for example) you would need BP1.1 with MTOM 1.0 or whatever. But of course you web service could support many bindings .... But I take your point that some compositions might not make sense. </AD>
I see an analogous situation for some (all?) "OGSA Profiles" -- particularly the WSRF Profile and any potential profile coming out of this BOF. <AD> If by analogous situation you mean non composable, I disagree. A more sensible approach would be to allow all OGSA basic profiles to be implemented together, (although higher related profiles will only compose with their corresponding profile). This is starightforward when the profiles rely on different specifications. When they refer the same specification, care should be taken not to have contradictory statements that lead to non-composability problems. </AD> Note that you might say that a WSRF profile could layer on a "WS-I OGSA Profile", but then I would question the very existence/purpose of this "WS-I OGSA Profile" (as it would seem to add no functionality above WS-I Basic Profile 1.1 and WS-I Basic Security Profile 1.0). <AD> Yes I agree. I was thinking more along the lines of a new profile based on WS-Transfer and WS-Eventing. The new profile and WSRF profile should be made to live happily together. So people will have the choice of two OGSA stacks. But there shouldn't be anything stopping them from implementing both stacks. Abdeslem </AD> -- Marty Marty Humphrey Assistant Professor Department of Computer Science University of Virginia
-----Original Message----- From: Tom Maguire [mailto:tmaguire@us.ibm.com] Sent: Wednesday, June 22, 2005 8:01 AM To: Djaoui, A (Abdeslem) Cc: Ian Foster; Marty Humphrey; Ogsa-Wg; owner-ogsa-wg@ggf.org; Subramaniam, Ravi Subject: RE: [ogsa-wg] OGSA-MWS-BOF at GGF14 on Tues June 28, noon-1:30
+1 to the composabiltiy aspect. Certainly you should be able to compose any number of profiles together. Each one with their own comformance claim.
Tom
Frey's Law: "Every 5 years the number of architecture components double and the ability to comprehend them halves"
Perfection is achieved, not when there is nothing more to add, but when there is nothing left to take away. - Antoine de Saint-Exupery
T o m M a g u i r e
STSM, On Demand Architecture
Poughkeepsie, NY 12601
"Djaoui, A (Abdeslem)" <A.Djaoui@rl.ac.u To k> "Subramaniam, Ravi" Sent by: <ravi.subramaniam@intel.com>, "Ian owner-ogsa-wg@ggf Foster" <foster@mcs.anl.gov>, .org "Marty Humphrey" <humphrey@cs.virginia.edu>, "Ogsa-Wg" <ogsa-wg@gridforum.org> 06/22/2005 06:36 cc AM Subject RE: [ogsa-wg] OGSA-MWS-BOF at GGF14 on Tues June 28, noon-1:30
Hi
For me the main issues to address are: A) Do we start from the web services design principle of COMPOSABILITY of specifications and profiles? B) If the answer is YES then how are we going to make sure that varoius new GGF profiles and specifications are indeed composable with each other. Once we have composable profiles it really doesn't matter how many profiles we have. You can use one them or all of them, it is less tidy than having one basic profile but its workable. If we throw away the design principle of composability, then I am afraid it is back to pre-OGSA pre-Web services era.
Abdeslem
-----Original Message----- From: owner-ogsa-wg@ggf.org [mailto:owner-ogsa-wg@ggf.org]On Behalf Of Subramaniam, Ravi Sent: Wednesday, June 22, 2005 6:29 AM To: Ian Foster; Marty Humphrey; Ogsa-Wg Subject: RE: [ogsa-wg] OGSA-MWS-BOF at GGF14 on Tues June 28, noon-1:30
Hi Marty,
The BOF and the motivations you describe sound like a good idea. It would be great to hear your perspectives especially since you have experience with the described specs (WS-RF/WSN and WS-T, WS-E, WS-Enum). I think this is also a good test for the general policy that OGSA has adopted of admitting different profiles that are consistent with the general architecture.
There has been a discussion around "lack of interoperability" with multiple profiles but without visiting the profiles that are compatible with the architecture one cannot decide which is the most appropriate (from one or many of the criteria like implementation easy, expressiveness, composability etc). Prematurely deciding that only one profile is the right way to go may not be beneficial in the long run. Given the way the industry has evolved, there is a process of "natural selection" and the most sophisticated or theoretical best solution has not won out in most cases. I agree with Mark's observations. Theoretically the degrees of freedom in the number of profiles seem infinite but I don't think in reality it will pan out into the more than a few since the "affinity" of these specs to be in a profile would constrain the number.
Having said this, it would still make sense to be concrete in your discussion of what would be the primary motivation of the BOF and subsequent WG. Your points in response to Ian's question could be added to the formal description of the intent of the BOF.
Thanks for spawning this discussion.
Ravi
PS: I just read Fred's comments in this thread. I think his points make sense. You may want to state that the need for alternate profiles as a primary assumption (not to be debated) and move on to the salient aspects of what the alternate profiles to be debated in the BOF are.
From: owner-ogsa-wg@ggf.org [mailto:owner-ogsa-wg@ggf.org] On Behalf Of Ian Foster Sent: Tuesday, June 21, 2005 7:43 AM To: Marty Humphrey; 'Ogsa-Wg' Subject: RE: [ogsa-wg] OGSA-MWS-BOF at GGF14 on Tues June 28, noon-1:30
Marty:
Thanks, that makes a lot of sense. And Mark puts it nicely: poetically, even (-:
Ian.
At 10:28 AM 6/21/2005 -0400, Marty Humphrey wrote:
Hi Ian et. al.,
A very good question! ("can you define the goals of the BOF more precisely other than "not WSRF"?) I sent this email in part to hear from the community how "broad" they might like it to be. That is, we're being flexible. If someone wants to talk about one of these specific topics, we will certainly try our best to accommodate.
Note that by default I believe the discussions will center around WS-Transfer, WS-Enumeration, WS-Eventing, et. al. I will talk about my team's use of WS-Transfer, WS-Enumeration, and WS-Eventing. I think we're in a very good position to discuss the pros and cons of these specs as compared with WSRF (as our project has implemented/used both).
Let me answer the question in a different way. GGF Chair Mark Linesch said recently: "Our approach with the OGSA architecture along with our collaborative work on OGSA profiles is to: (1) describe 'the most traveled paths through the forest' rather than to dictate that there is only one path; and (2) to continue to highlight that multiple, overlapping paths may not be in the interests of the industry over time" (
http://news.taborcommunications.com/msgget.jsp?mid=403500&xsl=story.xsl ) Simply, we believe that there has been sufficient "hallway discussions" on BOTH (1) and (2) that it makes good sense to gather people to discuss BOTH of these topics in a realistic and productive way.
- Marty
Marty Humphrey Assistant Professor Department of Computer Science University of Virginia
________________________________________ From: Ian Foster [mailto:foster@mcs.anl.gov] Sent: Tuesday, June 21, 2005 9:54 AM To: Marty Humphrey; 'Ogsa-Wg' Cc: 'Marty Humphrey' Subject: Re: [ogsa-wg] OGSA-MWS-BOF at GGF14 on Tues June 28, noon-1:30
Marty, Steven:
It is of course feasible, in principle, to define many different profiles for OGSA. E.g., one could build one on WS-Transfer and friends, one could build one that uses a different construct than the WS-Addressing EPR to address things, one could define one that uses JINI mechanisms, one could define one that uses CORBA, one could define one that renames all of the current WSRF and WS-Notification calls to be slightly different (oh wait, that's the first in the list (-: ), etc.
Given the wide variety of possible alternative profiles, it would be helpful for those like me who are considering attending the BOF to know what more specifically what the goal of this work is going to be. The name doesn't provide any information, other than to imply, perhaps, that the interfaces on which the WSRF profile builds are not in some manner "minimal" and/or "simple." I.e., can you define the goals of the BOF more precisely other than "not WSRF"?
Regards -- Ian.
At 09:11 AM 6/21/2005 -0400, Marty Humphrey wrote:
Folks,
There have been a number of informal conversations lately about the feasibility/value/implications of a possible non-WSRF-based profile for OGSA.
To bring all interested parties together at the same time, a BOF has been scheduled for Tuesday June 28 noon-1:30 (unfortunately at the same time as the EGA session, but there were no good times available). The agenda is still being finalized, but we expect to broadly discuss the pros/cons of such an effort, and, if the BOF attendees decide that such an effort would be valuable, produce a concrete plan for the formation of a Working Group.
Here is the information that Steven Newhouse provided for the GGF organizers:
"A BOF meeting to discuss the creation of a WG to define an OGSA Basic Profile that builds upon a minimal set of simple web services. The output of the group would be a document, similar in nature to the OGSA WSRF Basic Profile that would allow OGSA services to be rendered using an alternative set of WS specifications."
I hope you can attend this (hopefully) productive and constructive session!
-- Marty and Steven
Marty Humphrey Assistant Professor Department of Computer Science University of Virginia _______________________________________________________________ Ian Foster www.mcs.anl.gov/~foster Math & Computer Science Div. Dept of Computer Science Argonne National Laboratory The University of Chicago Argonne, IL 60439, U.S.A. Chicago, IL 60637, U.S.A. Tel: 630 252 4619 Fax: 630 252 1997 Globus Alliance, www.globus.org
_______________________________________________________________ Ian Foster www.mcs.anl.gov/~foster Math & Computer Science Div. Dept of Computer Science Argonne National Laboratory The University of Chicago Argonne, IL 60439, U.S.A. Chicago, IL 60637, U.S.A. Tel: 630 252 4619 Fax: 630 252 1997 Globus Alliance, www.globus.org