Component-based interaction design

It has been recognized that there exist dialog structures which re-occur in several applications and even in different application domains [11, 15, 3]. Examples are navigation and search support structures, help facilities, human error handling, and guided tours. Even though applications generally provide an implicit composition of such structures, it is often possible to identify and separate them from one another. So-called interaction patterns (see, for example, [11, 3]) provide an approach to reuse these structures. They define a framework to informally describe common, system-independent interactive structures, their relations, and usage. Besides their usefulness according to reuse, they lack support for verification and composition. In this paper, we use interaction nets (a slight extension of coloured petri nets) to model, verify, and compose interactive system behavior. It is suited for the domain of information services, since it permits to specify structures of human-computer dialogs as well as interaction in general which comprises interaction with databases in particular. Besides the known dynamic semantics of nets, we introduce a component semantics which considers nets as black-boxes with an interface. The component semantics is well-founded since the behavior of a net composition coincides with the composition of the black-box behavior. It opens the opportunity to combine the advantages of a net-based specification with those of component approaches. More precisely, it permits (i) to specify and verify interaction by a net model, (ii) to consider interaction patterns as implementation-independent black-box specifications called components, and (iii) to infer properties of compositions from properties of elementary specifications.