Towards constraint-based composition with incomplete service descriptions

We apply our work on Web service discovery with initially incomplete information to the problem of service composition. Rich semantic descriptions of Goals and Web services allow the unambiguous specification of constraints on what services are required and offered respectively. However, the data provided in service descriptions may be incomplete. We describe how missing instance data may be fetched dynamically and used for richer service discovery. We show how the resultant knowledge base can be used for simple compositions of services having mutual constraints.