On the Separation of Concerns in Distributed Programming: Application to Distribution Structure and Fault Tolerance in Mozart

Writing distributed applications is diicult because the programmer has to explicitly juggle many quite diierent concerns, including application functionality, distribution structure, fault tolerance, security, open computing, and others. An important goal is to separate the application functionality from the other concerns. This article presents one step towards that goal. We show how to integrate mutable pointers into a design that separates functionality, distribution structure, and fault tolerance. Mutable pointers, as a realization of explicit state, are an important data type that forms the basis for object-oriented programming. We start by deening mutable pointers in a centralized fault-free system. We then reene this deenition by successively adding a distribution model and a failure model. The resulting semantics can be implemented eeciently and is a suucient base to build nontrivial abstractions for fault tolerance. The design presented here is fully implemented as part of the Mozart Programming System (see http://www.mozart-oz.org).