Minutes of SWSA Telecon of February 10, 2004 Participants Amit Sheth, Tim Finin, Katia Sycara, Frank McCabe, Chris Bussler, Michal Zarembe, Stuart Williams, Carole Goble, Mark Burstein MB: I asked Katia and Frank to join us to tell us about the recent WSA F2F (while they still remember it). FM: Last f2f we did a lot of work. Document that will be pub soon is fairly complete and consistent. Current draft available at: http://dev.w3.org/cvsweb/~checkout~/2002/ws/arch/wsa/wd-wsa-arch-review2.html W3C doesnt seem to know what to do with architecture. This document is a sound basis for going forward. At last telco I attended, I promised to send a brief about WSA. I will do that. The architecture has stablized to the framework we talked about. There are two main lessons for this group: 1. WS industry is already quite close to adopting semantics in a mainstream kind of way. MB: Please Explain. FM: WS-Chor is going in direction of modeling the semantics of the transactions. Whether they have the ear of mainstream is too hard to tell. Also, Katia has just joined UDDI. From a strictly personal point of view, the requirement for a better grasp of what goes on between two computers is more obvious to them now. Purely syntactic like WSDL, process like BPEL, choreography like WSCL will go further. Few people are saying we don't want this stuff. Will they adopt OWL? I dont know - may be mostly the academic community. But we are not there yet as far as semantics for business logic and transactions. AS: Isn't there something that needs to be there as far as annotating WSDL? Is there something in that space? FM: You are going to get transactions first. AS: In the semantic sense? or the traditional? FM: EBXML, Rosettanet are two areas that have domain semantics up front. They dont have a clear idea about how to go about it. In a domain independent way, Katia working UDDI is the most interesting. If you are going to have automatic discovery of services, you need semantics. AS: Katia approach and others like ours - we will releasing an eclipse tool very soon. FM: Most people still believe that you can get away with a process model, but automatic discovery is not widely accepted. Management wants it, know that they need it. Mostly for QoS. AS: Mixed initiative discovery may come first. CG: Is that actually discovery of different services (vs redundant/repeated services)? FM: That is part of the story, but automated configuration requires discovery from ground zero. There is trust involved. You need to be in a trust domain. They are dealing with thousands of services - e.g. Verizon. AS: Are there really that many ? Michael Brady said at ISWC that there were few real WS. FM: Well, he said that he had many to interoperate CB: I asked and he said that they were read only, don't have surviability support so can't go beyond that. FM: I was being a little abstract - rolling in deployed envs like CORBA, not just WS. For WS, there about 5 of them, and that's it. CB: and all producing stock ticks. FM: but the planning folk have experience dealing with thousands and they know what the game is. KS: I dont know what the lesson is for this group- whether this group should start looking at management. FM: Please dont. KS: I think his point is - dont know what the industrial standards groups are going to do, but realize that semantics is needed - they see the need for its incorporation into standards. The point is what should be the standard? How should we influence it? So agree with FM that there is more of a willingness - that we need semantics, but how to go about it? What is lesson for this group? We want to be making some kind of recommendation. Looking at WSDL, one strategy would be, concretely, look at particular standards and say how to increase the semantic content. W3C on a more glacial timeline, so influence thru more concrete proposal. That would be a strategy. FM: One of the hot topics is the controversial RESTifarians position they make the case that a constrained well agreed to semantics is powerful. They say that what we have is good enough. Not true. But collectively taking oup the challenge of what is enough - to get interoperability going. I'm suspicious of saying "can discover the meaning by looking at a website. AS: A lot more emphasis will be placed on guaranteed committments, Those things will be coming after you have addressed transactions... FM: I sympathize but dont agree KS: ditto - People started selling tennis rackets, etc on the existing web at the point where html just being created, security not addressed. FM: Big driver is in the enterprise. MB: That is one way to get additional shared semantics. CB: I think of the effort a few years ago to build corporate data models MB: Yes, but there the trouble was having a single shared ontology across the enterprise, which met no one's needs well. What we want is interoperable ontologies. One thing we need to be able to address is what community supports are necessary. As Mike Uschold said at our first F2F, semantic web is about communities with some higher level of shared interoperability at the level of semantics/terminology communicating with another community that has different models. How to support this wider interoperability is one of the issues, and it relates to both automated discovery and transaction support. We need to be able to make some consistent assumptions, for example, about how semantics are shared/located on the semantic web. OWL is the main model at the moment, but this is a small community. RDF/URI model is wider. FM: You mention the term infrstructure - lesson of the web is that infrastructure is very light. Youve got TCP, BNF. not much else AS: That is, for light for WS, Ecomerce is not so light. When you start trying to support transactions, security, policies, ... then it is not light. but hopefully layered and compartmentalized. When you start adding it is not light. TF: Corba tried to do too much in one step. then it evolved. earliest web sites were simple, then it evolved. Difference from old EDS approach, be bottom up and support evolution. KS: Definitely the infrastructure of the web exists. we are not recreating. One could take as given some version of current semantic web infrastructure. MB: We could assume OWL or really RDF - URIs for coreference. KS: best to say what is the minimal functionality that semantics can provide on top of existing standards. One could look at it as the WSA doc bounding from above particular concepts, particular standards bounding from below. What would be a good tradeoff between there. If you look at WSDL and so on, they are tring to be a constrained interface. FM: the idea of constrained interface is important - This group defining a constrained interface analogous to REST but richer would be interesting and potentially helpful. KS: WSDL and SOAP could be looked at as the currently proposed interface. What can we add/modify, etc, to allow constrained interface that allows sufficient interoperability. SW: I'm intrigued by the focus on REST, - had conversations with Mark - spoken to both he and Roy. Like a memory bus for strangely shaped memory locations. Go thru sequences of resource states (this part is above REST). Set of constraints on sequence of steps. FM: Interesting thing abou REST - introduces state as a resource. what Katia and I are groping toward, there may be other wide ranging concepts that would engender different committments. Developing a language and model for committments, gives you a framework for bus apps. Operate at higher level than web browser. CB: REST doesnt talk about state FM: It talks about modifying resources - assignment. It is powerful but doesnt do what we need. What we need will not be an extension, but by analogy will have a global model. Agreement is very expensive. a model that revolves around that raises the bar for interop, reduces the cost of entry. Dont know what it will look like. Has to do with speech acts, committments, those kinds of ideas. CB: Everybody requires committments, but it is very complex. Others come with SOAP - that is basically the hacker community. There is a big gap. Is there a way to find something in the middle to bridge that gap? MB: This brings to mind shared intention/shared plan models and BDI, but I think most people find BDI difficult to udnerstand (as a programming model). KS: Start with FIPA? FM: FIPA went down wrong route with strongly mentalistic model. In the end, we couldnt rescue it. (Speaking as a former FIPA chair) SW: What do you mean by mentalistic? FM - Depended on the internal state, not on the societal interaction. TF: That may be a red herring, most of the community was not concerned with that part of the semantics. KS: May be, but it is not a great basis for grounding the semantics fo web service. FM: One problem was that after Dave -?- left, few left who would have that conversation. It was mostly sofware devleopers. Most agreed that the model was not right, didn't go far enuough. Committment, social semantics was needed. But that is history. Basic idea of having a model at level of communicative acts is still sound. KS - I wish we could convince W3C to continue WSA or create a new working group on interoperability. Then we could talk about these issues there. They don't seem to know what to do with WSA or web services. Maybe we should couch this as semantic interoperabiltiy - OWL may not provide all that is needed. This could be an opportunity for SWSI to enter into that vacuum. SW: For W3C, you need membership to go thru and review. "Walking the hill" You need to know that your ducks are lined up. FM: Hurdle Is High, But Payoff Is higher still. Need community agreement. Agreement is expensive. I think the real problem in W3C is lack of leadership in the area of Web Services. SW: Big players are taking care of their interest ouside the W3C. This is a concious decision. FM: This is a group that is outside the w3c as well. SW: But not much work with IBM, MSFT among us. AS: We do some with IBM in the area of semantics, policy - a joint project. Watson mostly. With Francisco Cubera, others KS: There are individuals who have some sympathy to semantics there. AS: We have to encourage them to take ideas to standards. SW: Have to work out how to make it real. Need the key ideas.