Configuration of Web services as parametric design

The configuration of Web services is particularly hard given the heterogeneous, unreliable and open nature of the Web. Furthermore, such composite Web services are likely to be complex services, that will require adaptation for each specific use. Current approaches to Web service configuration are often based on pre/post-condition-style reasoning, resulting in a planning-style approach to service configuration, configuring a composite web service "from scratch" every time. In this paper, we propose instead a knowledge-intensive brokering approach to the creation of composite Web services. In our approach, we describe a complex Web service as a fixed template, which must be configured for each specific use. Web service configuration can then be regarded as parametric design, in which the parameters of the fixed template have to be instantiated with appropriate component services. During the configuration process, we exploit detailed knowledge about the template and the components, to obtain the required composite web service. We illustrate our proposal by applying it to a specific family of Web services, namely "heuristic classification services". We have implemented a proto-type of our knowledge-intensive broker and describe its execution in a concrete scenario.