In a discussion over at Sam Ruby‘s blog and on his “Distributed state machines” post, Robert Sayre made a random thought which I find intriguing.

In a service-oriented world, like that promoted by MEST, where integration between services happens through metadata-sharing about the services’ messaging behaviours (described in languages like SSDL), could such route-finding algorithms have a place? As interactions become more complex and simple protocol blocks become the building blocks of larger composites, could we employ such an algorithm in order to determine how applications are composed? Of course semantics will have to play a key role here.

The building blocks (the protocols) will have to be semantically described. A semantics-based categorisation will have to be created; this will represent the map. Semantic differences, usage costs, QoS requirements, etc. could become the obstacles in the map. The target will represent the semantics of the composite application. The route to the target will represent those protocols that can be composed in order to build the new composite behaviour. The choice of protocols would have only been based on semantic descriptions.

As I said… random thoughts.