Re: [ogsa-wg] Profile documents
Dave's comment captures a major concern I have about how the profile notion seems to be evolving. A profile, in the sense of the WS-I profiles that formed at least the starting point for our discussion, documents a way of using well-established and widely used specifications. I believe we have fairly broad consensus for that definition. However, we are now introducing the idea of a "draft" profile that does not need to be grounded in adoption. In so doing, I fear that we abandon the discipline that will allow us to ensure that profiles represent authoritative statements on how to build Grid systems. GGFers will conclude that a profile is a way to respectability. We will start to see many proposals for profiles, many with no connection to implementation experience or adoption, and I don't see how we will be able to say yes to some and no to others, as we will have no clear rules for doing so. Working groups will start to produce their own profiles, as well. We will end up with a set of profiles that look as diffuse and ill defined as the current set of GGF working groups. Of course, some will be "draft" and some "recommended", but I think that distinction will be lost on the community. DAIS for me represents an excellent test case for what a profile should be. It's a nice piece of work and has at least one academic implementation. However, it hasn't seen any adoption by database vendors. That to me means that it doesn't belong in a profile. This is not to say at all that DAIS is not valuable, or that the DAIS team should not be working to get DAIS adopted by vendors. It's simply saying that a profile isn't the way to do it. Ian. At 09:22 AM 12/21/2004 +0000, Dave Berry wrote:
I definitely want to allow the DAIS spec in the first data profile. This is partly the chicken and egg question; if a spec isn't in an OGSA profile, people will be less inclined to implement it. This is particularly applicable to the DAIS spec, because we want people to implement it as part of their DBMS systems, not as a standalone executable. (This is to avoid unnecessary copying of data in the implementations).
_______________________________________________________________ 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
Yes, I would like to echo Ian's comments. That is, I have only followed OGSA WG off-and-on - only to see how it might impact our project (WSRF.NET). But I was *very* surprised to hear the word "profile" come up, as it seems premature (certainly in the sense of WS-I) for many of the reasons that Ian mentions. I think the community will be confused by the use of this word. I think Ian's points (and Dave's points before him) are very important and should not be taken lightly. -- Marty Marty Humphrey Assistant Professor Department of Computer Science University of Virginia _____ From: owner-ogsa-wg@ggf.org [mailto:owner-ogsa-wg@ggf.org] On Behalf Of Ian Foster Sent: Tuesday, December 21, 2004 8:04 AM To: Dave Berry; d.snelling@fle.fujitsu.com Cc: ogsa-wg Subject: Re: [ogsa-wg] Profile documents Dave's comment captures a major concern I have about how the profile notion seems to be evolving. A profile, in the sense of the WS-I profiles that formed at least the starting point for our discussion, documents a way of using well-established and widely used specifications. I believe we have fairly broad consensus for that definition. However, we are now introducing the idea of a "draft" profile that does not need to be grounded in adoption. In so doing, I fear that we abandon the discipline that will allow us to ensure that profiles represent authoritative statements on how to build Grid systems. GGFers will conclude that a profile is a way to respectability. We will start to see many proposals for profiles, many with no connection to implementation experience or adoption, and I don't see how we will be able to say yes to some and no to others, as we will have no clear rules for doing so. Working groups will start to produce their own profiles, as well. We will end up with a set of profiles that look as diffuse and ill defined as the current set of GGF working groups. Of course, some will be "draft" and some "recommended", but I think that distinction will be lost on the community. DAIS for me represents an excellent test case for what a profile should be. It's a nice piece of work and has at least one academic implementation. However, it hasn't seen any adoption by database vendors. That to me means that it doesn't belong in a profile. This is not to say at all that DAIS is not valuable, or that the DAIS team should not be working to get DAIS adopted by vendors. It's simply saying that a profile isn't the way to do it. Ian. At 09:22 AM 12/21/2004 +0000, Dave Berry wrote: I definitely want to allow the DAIS spec in the first data profile. This is partly the chicken and egg question; if a spec isn't in an OGSA profile, people will be less inclined to implement it. This is particularly applicable to the DAIS spec, because we want people to implement it as part of their DBMS systems, not as a standalone executable. (This is to avoid unnecessary copying of data in the implementations). _______________________________________________________________ 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 <http://www.globus.org/>
As a DAIS co-chair, it seems reasonable for me to pitch in a view here - a personal one rather than a group one for now, as there are no scheduled calls for the group until the new year. It seems reasonable: 1. That profiles should only have dependencies on formally accepted standards (any dependencies on less mature or stable specifications means building the house on sand). 2. That once a specification has been accepted as a standard by the GGF, the GGF should seek facilitate the wide adoption of that standard. The DAIS specs are not formally adopted standards (but are currently rather stalled waiting for stability in the WSRF space), and thus should not be candidates for inclusion in a profile in the short term. However, a feature of the DAIS specs is that they have deliberately constrained ambition. That is, they seek to provide specific pieces in a jigsaw that as a whole provides comprehensive data access, delivery and management capabilities in a service-oriented environment. As such, although the DAIS specs seek to provide useful functionality on their own, they will be even more useful (!) when used in conjunction with other specifications. As such, the development of profiles that demonstrate how DAIS specs should be used in conjunction with other specifications as part of a collection of data-oriented services seems likely to be important to uptake. I would expect a similar case to apply to other specifications (i.e. whole > sum of parts). As such, profiles should be able to be seen as a way of encouraging wider adoption of GGF standards. Regards, Norman p.s. A more minor comment on two words in Ian's email - one shouldn't see "database vendors" as the only candidate commercial developers of DAIS specifications. Clearly such vendors have been major players in the development of the DAIS specifications, but one could certainly see other forms of commercial organisation (e.g. middleware vendors) as viable candidates.
DAIS for me represents an excellent test case for what a profile should be. It's a nice piece of work and has at least one academic implementation. However, it hasn't seen any adoption by database vendors. That to me means that it doesn't belong in a profile. This is not to say at all that DAIS is not valuable, or that the DAIS team should not be working to get DAIS adopted by vendors. It's simply saying that a profile isn't the way to do it.
Ian.
At 09:22 AM 12/21/2004 +0000, Dave Berry wrote:
I definitely want to allow the DAIS spec in the first data profile. This is partly the chicken and egg question; if a spec isn't in an OGSA profile, people will be less inclined to implement it. This is particularly applicable to the DAIS spec, because we want people to implement it as part of their DBMS systems, not as a standalone executable. (This is to avoid unnecessary copying of data in the implementations).
_______________________________________________________________ Ian Foster www.mcs.anl.gov/~foster <http://www.mcs.anl.gov/%7Efoster> 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 <http://www.globus.org/>
participants (3)
-
Ian Foster -
Marty Humphrey -
Norman Paton