From: FLAIRS-02 Proceedings. Copyright © 2002, AAAI (www.aaai.org). All rights reserved. Semantic Web Service Architecture — Evolving Web Service Standards toward the Semantic Web ½ ½ ¾ ½ Tanja Sollazzo½ ,Siegfried Handschuh ,Steffen Staab ,MartinFrank ,NenadStojanovic ½ Institute AIFB, University of Karlsruhe, 76128 Karlsruhe, Germany [email protected], sha, sst, nst @aifb.uni-karlsruhe.de ¾ Information Sciences Institute of the University of Southern California, Marina del Rey, California 90292 [email protected] Abstract and layering of languages in the Semantic Web. To the detri- ment of the latter community the adoption of their schemes Theimportance of Web services has been recognized and into industry (quasi-)standards for Web services is far from widely accepted by industry and academic research. How- trivial, because there is no coherent architecture of immedi- ever, the two worlds have proposed solutions that progress ate practical benefit. along different dimensions. Academic research has been, mostly concerned with expressiveness of service descrip- The core idea of this paper is to present an architecture tions, while industry has focused on modularization of ser- that combines the two worlds and their potential benefits. vice layers — mostly for usability in the short term. This pa- The benefits of the integration include increased visibility per is concerned with merging these two streams of progress. of Web services because open ontology frameworks allow Our point of departure is the current proposal by IBM. Its pro- for semantically expressive advertising on the Web that may posal is extended by Semantic Web technologies such that be found by Web crawlers. They include better usability asmooth evolution from Web services in the current Web because of more expressive Web service descriptions. They to Web services in theSemantic Web appears possible and include a smooth evolution from Web services for human —infact — highly desirable. As a showcase we describe users such as targeted by current industry (quasi-)standards SWOBIS,anontology-compatible registry for software tools, toward Web services for personalized machine agents that that represents a first step towards developing a search engine forWeb services based on Semantic Web technologies. assist the user. The structure of this paper is as follows. First, we sketch the general model of the Web service setting, of the archi- Introduction tecture of IBM in particular. We consider the latter the most elaborate and the best described industry quasi-standard for Webservices are “self-contained, self-describing modu- WebServices so far. Second, we analyse assumptions of lar applications” (Martin 2001). They constitute software this (and related) architecture(s), describing several parame- modules that “describe a collection of operations that are ters which may be varied to turn the “traditional” view into network-accessible through standardized XML messaging” aSemantic Web view. Third, we outline the current descrip- (Kreger 2001, p. 6). With OAP, UDDI, WSDL, and .NET in- tion of DAML-S (Ankolenkar et al. 2001). Fourth, we crit- dustry has made a bold move and started initiatives that tar- ically evaluate the achievements of DAML-S. Fifth, we de- get the potential benefits of Web services. The focus of the scribe the integrated architecture that we propose. Finally, initiatives was an evolutionary step from current Web tech- we sketch SWOBIS, an ontology-based registry for software nology toward a technology for Web services. Key concerns tools that we are currently developing into the direction of of the initiatives were, e.g., short-term applicability or scal- our proposed integrated architecture. ability. This implies that the corresponding Web service ar- chitectures build on little really new technology inside, e.g. IBM Web Service Architecture they use standardized taxonomies and vocabularies that ex- hibit little flexibility and expressiveness and that restrict the The typical procedure in a Web service setting is the fol- usability of Web services mostly to human users rather than lowing: A Web serviceprovider offers services on the Web. machine agents. For the latter one would need, e.g., Web He may choose to register her service at an online registry service description languages that support semi-structured of a service broker (Trastour & Bartolini 2001, p.2/3). The data, constraints, types and inheritance. registry publishes and locates services. To allow for service In contrast to the industry point of view, academic re- discovery, the registry also provides standardized descrip- search has investigated languages that fulfill exactly these tion facilities, e.g. taxonomies that allow the description of, needs (Horrocks et al. 2001; Ankolenkar et al. 2001; (i),the functionality of a service, (ii), its service provider, Fensel et al. 1999) offering extensible ontology frameworks (iii),how to access and interact with the service. The corre- sponding information about a particular service is registered Copyright ­c 2002, American Associationfor Artificial Intelli- by the provider at a broker. gence (www.aaai.org). All rights reserved. The (human) requestor searches for a service at the reg- FLAIRS 2002 425 istry. It finds one by browsing or querying the registry. Then ture, because issues like security, management and quality she uses the service description to create a binding for her of service span all other layers. application to be able to invoke or interact with the Web ser- vice implementation. In the IBM Web service architecture Discussion of the IBM (Quasi-)Standard this procedure boils down to the architecture layers depicted Architecture in Figure 1 (cf. (Kreger 2001)). This description, the current industry quasi-standard, how- Network Protocols and SOAP. The lower layers are based ever, is too “flat” to be comprehensive. Rather, there are a on general network protocols, like HTTP or FTP, and SOAP number of parameters that may be varied: (simple object access protocol1), which is a standardized en- veloping mechanism for communicating document-centric Table 1: Dimensions of Web service features messages and remote procedure calls using XML. DimensionChoices WSFL Service Flow “Traditional” Semantic Web Service Simple Composed Requestor Human Machine Static à UDDI Service Discovery Quality o Mana Provider Registration No registration Direct à UDDI Service Publication Security Broker Key Player Facilitator gement fSer Service description Taxonomy Ontology vice Descriptive elements Closed world Open world WSDL Service Description Data exchange Syntactic-based Semantics-based SOAP XML-Based Messaging In particular, the Semantic Webwill allow richer descrip- tions of Web services (e.g., semi-structured data, types, in- HTTP, FTP, Network Protocols email etc. heritance, semantic constraints). The key role of the broker maydisappear, it may still be viable as a kind of search en- gine for Web services (with meta search engines on top), but Figure 1: IBM Web Service Architecture it will lose its central role as aregistry, because everyone WSDL. The interface of a Web service and its service in- may publish semantic descriptions and crawlers may find teractions are described in WebServices Description Lan- them. Personalized machine agents will take over the role guage (WSDL2). WSDL allows the composition of XML of a service requestor from the human user. And, they may documents to describe technical details of a Web service. also do the composition for the human user. When we distinguish between “traditional” and Semantic UDDI. The service publication is regarded as a business- Web features, we do, however, not require that only the lat- related description of a service; e.g. answering questions ter be valid for two reasons. First, we do want to allow for such as: What products are associated with this service? an evolutionary development ofaWebService architecture. Which organization is offeringthisservice? Universal De- Second, also “traditional” features offer several advantages. scription Discovery and Integration (UDDI), initiated by Mi- For instance, a service broker as *intermediary* may take crosoft, IBM, and Ariba3, does not just consist of defining a care of concerns like validation of information or even se- data structure standard for all business-related descriptions lection of service providers. of services (i.e. service publication), but it also contains all mechanisms that allow the service requestor to gain access to the service publication and service description, and thus Service Description with DAML-S is also in charge of the service discovery layer. For the lat- DAML-S is a DAML+OIL-based Web service ontology, ter purpose, a UDDI registry contains data in a combination which supplies Web service providers with a core set of of white pages, yellow pages, and green pages. The white markup language constructs for describing the properties pages include items such as a company’s name, its address, and capabilities of their Webservices in unambiguous, and other contact information. The yellow pages help to cat- computer-interpretable form. DAML-S markup of Web ser- egorize a business; they define its business type, as well as vices facilitates the automation of Web service tasks includ- theindustry in which it operates. Finally, the green pages ing automated Web service discovery, execution, composi- define what kind of services a business offers, and how to tionand interoperation. In particular, it provides language electronically communicate