Some issues in the semantics of facile distributed programming

Abstract

Facile is an experimental programming language intended to support applications that require a combination of distribution and complex computation. The language originates from an integration of the typed call-by-value λ-calculus with a model of concurrency derived from Milner's CCS. At a theoretical level, an operational semantics has been developed in terms of labelled transition systems, and a notion of observability of programs has been defined by extending the notion of bisimulation. An experimental implementation currently supports distributed programming over networks of workstations. The implementation, obtained by extending the ML language, supports polymorphic types as well as mobility of functions, processes and communication channels across a distributed computing environment. A number of language constructs have been added or modified to handle certain issues that arise with real distribution. These include the need to control the locality of computation in a physically distributed environment, the potentially expensive implementation of certain operators and the need for a system to tolerate partial failures. In this paper we discuss a possible approach for the operational semantics of these constructs that follows the Facile philosophy and some recent results in concurrency theory.