Choreographing Complex Services

One of the key challenges of the CancerGrid project will be modelling of the dynamic aspects of clinical trial protocols: service composition, document workflow, exceptional behaviours, and so on. Just like the static aspects concerning data and metadata formats and ontologies, these dynamic aspects should not be hardwired directly into program code, because then the only thing that can be done with them is to execute them. Rather, the dynamic aspects should be recorded declaratively. This still allows execution, by interpreting the declarative descriptions; but it allows other uses too: browsing of the protocol, modification without having to write code, meta-processing by other applications, and so on. And of course, none of these concerns are specific to CancerGrid — the same considerations apply to any complex service-oriented architecture. This project is to develop a theory (and, if appropriate, a supporting, prototypical tool suite) for the modelling, interpretation, manipulation and processing of declarative descriptions of dynamic behaviour in service-oriented architectures. This will build on work from the coordination language (using notations such as Darwin) and web service choreography (WSCI, BPEL) communities, but it will not be limited to designing yet another programming language and its interpreter; it will also explore mechanisms for manipulating such programs, meta-processing, model-checking, and their relationship with data ontologies.