Evaluation of the SR Language Design

摘要

SR is a language for writing distributed programs. It supports many forms of interprocess communication to obtain high expressiveness. It has also been designed to be simple to use; this is done by providing a few orthogonal concepts that can be combined in many ways. The paper evaluates these design goals of expressiveness and simplicity. It studies several programs for parallel applications, which have been implemented in SR. The results indicate that, even within the restricted domain of parallel distributed applications, nearly all facilities provided by SR are useful. It found uses for synchronous and asynchronous message invocation, explicit, implicit, conditional, and ordered message receipt, and multicast. The language lacks message passing through mailboxes, message forwarding, and globally shared variables. Despite its expressivity, SR has also been fairly simple to understand. A strong point in the design is the orthogonality of the message sending and receiving constructs. In addition, the language semantics and the integration of sequential and distributed constructs are in general well designed. Some points of criticism concern the multicast mechanism, the type system, and the type-insecurity of the language.