A technique for distributed systems specification

Abstract

In this paper we show how an object-oriented specification language is useful for the specification of distributed systems. The main constructors in this language are the objects. An object consists of a state, a behaviour and a set of transition rules between states. The specification is composed of three sections: definition of algebraic data types to represent the domain of object attributes, definition of classes that group objects with common features, and definition of relationships among classes. We show two possible styles for defining the behaviour of objects, on one hand we use a transition system (state oriented) and on the other hand we use an algebraic model of processes description (constraint oriented). We illustrate the paper with the specification of the dining philosophers problem, a typical example in distributed programming. >