- From: Shi, Xuan <xshi@GEO.WVU.edu>
- Date: Thu, 1 Sep 2005 22:43:00 -0400
- To: "'David Martin '" <martin@AI.SRI.COM>, "Shi, Xuan" <xshi@GEO.WVU.edu>
- Cc: "''Carine Bournez ' '" <carine@w3.org>, "''Battle, Steven Andrew ' '" <steve.battle@hp.com>, "''public-sws-ig@w3.org ' '" <public-sws-ig@w3.org>
Hi, David, Thanks very much for your kind attention and advice. I hope my straightforward comments will not make people in this group unhappy but rather will promote more discussion about the semantic Web service research and development. Even the saying like semantic Web services = semantic Web + Web services is questionable. In the paper "Potential Modeling And Simulation Applications Of The Web Ontology Language - OWL" which can be found at: http://www.informs-sim.org/wsc04papers/031.pdf the author discussed that " ... OWL was developed to support the Semantic Web. OWL provides web-ready object-oriented information representations with an open vendor-neutral language. OWL is best used for representing object oriented descriptions of items in a well-defined domain. The information representation constructs in OWL supports object-oriented descriptions. OWL supports the definition of classes, individual instances, and property relationships between classes, individuals, and properties. ..." It seems that such statements demonstrated that the current technologies (OWL, etc.) can only define the "nouns" other than the "verbs". Thus OWL is good at defining the object relatioship used in OOP in class-subclass definition, but how can we define the meaning, behavior, and action of the functions and services? Thus we need to consider how to expand such methodology to define "verbs" (actions, behaviors, etc., i.e. the meaning of services and funtions). Given the following example, how can we define the semantics (the meaning) of such a Web service and function: select geospatial features from feature layer 1 that contain, or are contained by, the features in feature layer 2. The function interface can be described as: Function functionName (String polygonLayer1, String polygonLayer2) : String selectedPolygonFeatures This can be a typical interface to process geospatial data or perform spatial query. Moreover, the exactly same function: Function intersect (String polygon1, String polygon2): String polygon3 can perform different tasks (spatial query vs. geoprocessing) and get different results. Then how can we use OWL to define such different purpose and bahavior of such functions with the same name and interface? Thanks to Naveen's kind attention and assistance, I can use CMU's WSDL2OWLS tool now for testing again on the WSDL file - http://arcweb.esri.com/services/v2/AddressFinder.wsdl The problem, I am not sure if anyone tried this sample, is that the users need to first invoke another Web service to get a dynamically generated authentication token and then they can use the AddressFinder Web services. The problem is here again, how OWL-S and WSMO can describe such requirement or meaning inside the Web services? That's to say, user name and password are used to invoke another Web service which is the precondition to invoke AddressFinder Web services? Actually the user can have a global account then they can use such user name and password to invoke any paid Web services. Such design complies with OOP requirement but leaves the problem for creating semantic Web services. In this example, "token" has the same meaning to the user but when we use WSDL2OWLS tool to get the OWL file, we can find "token" is defined three times because it is used in three functions. Any suggestions to such examples? -----Original Message----- From: David Martin To: Shi, Xuan Cc: 'Carine Bournez '; 'Battle, Steven Andrew '; 'public-sws-ig@w3.org ' Sent: 9/1/05 2:44 PM Subject: Re: FWD [Work in Progress on Semantics for Web Services (Advance Notice)] Hi Xuan - Thanks for your comments. You make some good points about OWL-S and other SWS approaches such as WSMO. But it appears to me that you are giving an inaccurate picture of what OWL-S is about, so I would like to provide a clarification. Your comments seem to be based on the assumption that OWL-S is only meant to be used in a "backward-engineering" style; that is, only meant to be used to annotate pre-existing WSDL services, which were designed independently from Semantic Web approaches. Here I only want to point out that OWL-S was not primarily conceived for use in that "backward-engineering" style. As Alois Reitbauer has also pointed out in a previous message, it is possible to employ Semantic Web approaches in a forward-engineering methodology; that is, incorporating ontology-based semantics from the beginning in the design process. A number of researchers have been working on this way of doing things, and tying it in with widely used methods such as those based on UML. Furthermore, please note that WSDL-specified Web services are not mandated to use XML Schema to describe their inputs and outputs. WSDL has extensibility mechnisms that allow for other "typing" mechanisms. Thus it is possible to imagine Web services that are "native speakers" of OWL (their inputs and outputs are exchanged using one of OWL's serialization syntaxes), which would largely address the gap that you describe between a WSDL and an OWL-S specification. I hasten to add that no such services exist today, and furthermore there are some unresolved difficulties. (These have to do with the fact that OWL instances can be serialized in so many different ways, and there can be ambiguity regarding how much content needs to be carried to adequately characterize an OWL instance.) However, there has been a good deal of discussion about these difficulties, and I believe they can be addressed. Presumably this might be a topic of concern if any SWS standardization or pre-standardization activities materialize at W3C. Finally I would like to add that there is no basis for your characterization that SWS is "big business". There is a great deal of research interest in these technologies because the promise is so great, but that's not at all the same as saying that it is "big business". (For instance, I do not know of any companies that are trying to ensure profits by trying to control the direction of SWS standardization activities.) I am certain that Carine is only trying to make progress possible by considering some activities that would promote greater collaboration between those who are working on WS and those who are working on SWS, so that (among other things) we can make further progress towards addressing issues such as those that you raise. Regards, David Martin
Received on Friday, 2 September 2005 02:42:49 UTC