New directions in implementing the pi calculus

Do you know what the pi calculus is? It is a language invented ten years ago for describing concurrent and distributed systems. It has come to dominate theoretical research into concurrency and distribution, and now its time has come to be used in practice. The author, along with Laneve and Gardner, has recently developed a distributed virtual machine [10, 12, 2] for the pi calculus. This is new territory – ours is actually the calculus’ first true distributed implementation. We depart in several ways from mainstream ideas in the research community. Indeed, our implementation has more in common with the commercial product Microsoft Biztalk [5] (a recent tool used to integrate business systems and which itself is based partly on the pi calculus). In our future plans we have been partly inspired by the practical concerns faced by Biztalk; in turn, the designers of Biztalk are taking some of our ideas for their next version. The goal of this paper is to present the practical lessons learned from our ongoing implementation experience and also from Biztalk. We hope to challenge some existing ideas, and to draw attention to fresh areas needing investigation. The basic motivation is that the pi calculus seems an easier way to write a wide range of concurrent and distributed programs.