Coordination languages and their significance

Reviewer: Mark Brimhall Wells

In this paper, the authors view their Linda parallel programming system [1] as a coordination language orthogonal to classical computational languages such as FORTRAN and C. Coordination refers to the creation (but not specification) of computational activities and the support of communication among them. The authors argue that this approach is a better way to approach parallelism than integrating computation and coordination into a single language as in concurrent logic programming or relying solely on implicit parallelism as in functional programming. They maintain that portability, generality, conceptual economy, and flexibility favor this approach. The paper is informal and generally self-contained, although I found it useful to reread the above-cited paper and the subsequent technical correspondence [2]. The arguments are forceful but not overwhelming. I am particularly in tune with the defense of explicit parallelism. More generally, however, I believe that mechanisms already exist in most modern object-based languages that can allow the integration of coordination with computation and still preserve conceptual and practical orthogonality and their advantages. In any event, readers will find the paper thought-provoking, and most will agree with the authors that this is a fertile research area.