Concurrency has always been a fact of life in the areas of system and network programming. Networks now being standard, the interest for concurrency has long shifted from specialists to ordinary programmers. Moreover, the ever growing availability of multi-core machines augments the population of those that wish to have several computing “agents” to cooperate. However, as everyone who has tried knows, writing a distributed application, one that runs on several machines, is not an easy task.
The join-calculus [5] is a process calculus in the tradition of the π-calculus [15]. The main purpose of such calculi is to describe concurrent and distributed systems; programming such systems is a different, although related, issue. The issues are related because a good model offers suitable abstractions that help programmers. To shorten the distance between model and program, the join-calculus has been designed with implementation in mind. We will not discuss this point on theoretical grounds, but rather on practical grounds. Indeed, we are developing a programming language based upon the join-calculus.
Our language, JoCaml, is an extension of Objective Caml (OCaml) [11], a popular dialect of ML. By choosing to extend an existing language, and not to design one of our own, we first intend to minimize our work. We also intend to benefit of functional programming, of pre-existing code base, and of a population of programmers open to innovation.
In this work, we more illustrate the JoCaml language than the JoCaml system (however, see Section 2), by showing how a middle-sized OCaml program can be made into a JoCaml program (Sections 3–7) suited for execution on a network of computers. We then conclude after a few performance figures.
The complete source for the program is available at http://jocaml.inria.fr/pub/joex/JoCamlsRUs.tar.gz.