Synthesizing Semantic Web Service Compositions with jMosel and Golog

In this paper we investigate different technologies to attack the automatic solution of orchestration problems based on synthesis from declarative specifications, a semantically enriched description of the services, and a collection of services available on a testbed. In addition to our previously presented tableaux-based synthesis technology, we consider two structurally rather different approaches here: using jMosel, our tool for Monadic Second-Order Logic on Strings and the high-level programming language Golog, that internally makes use of planning techniques. As a common case study we consider the Mediation Scenario of the Semantic Web Service Challenge, which is a benchmark for process orchestration. All three synthesis solutions have been embedded in the jABC/jETI modeling framework, and used to synthesize the abstract mediator processes as well as their concrete, running (Web) service counterpart. Using the jABC as a common frame helps highlighting the essential differences and similarities. It turns out, at least at the level of complication of the considered case study, all approaches behave quite similarly, both considering the performance as well as the modeling. We believe that turning the jABC framework into experimentation platform along the lines presented here, will help understanding the application profiles of the individual synthesis solutions and technologies, answering questing like when the overhead to achieve compositionality pays of and where (heuristic) search is the technology of choice.