Itinerary-centred programming of mobile agent applications

Mobile agent programming involves building applications where programs move from host to host to perform tasks. There is an inherent distributed and dynamic nature of such applications which distinguishes it from traditional nonmobile software. We advocate itinerary abstraction as an approach to mobile agent programming. Itinerary abstraction is the process by which programmers define the itinerary of agents for an application, including which locations (resources) are visited, in what order locations are visited, and what tasks are performed at each location. We propose this as an initial step in developing mobile agent applications. The Itinerary encourages separation of concerns to simplify mobile agent programming. The mobility aspect of a group of agents are abstracted away from code details implementing the computations the agents are to perform on hosts. The itinerary also facilitates identification and reuse of patterns in mobility behaviour. We discuss an algebraic approach in developing an itinerary language allowing compositionality and disciplined extensibility via adding operators. The language allows economical explication of mobility behaviour and manipulation of itineraries independently of any programming language, i.e. we abstract away the clutter of the syntax of a full programming language. This also allows us to study itineraries in their own right and their translations into other representations. We examine two translations of itineraries, into workflow nets and controlled English, for their effective use in specific applications.