Scenarios: A model of non-determinate computation

Abstract

Many researchers have sought a semantic theory that extends the history function model of determinate networks to a history relation model of non-determinate networks. By demonstrating that history relations are a fundamentally incomplete characterization of non-determinate computation, we have shown that no such extension exists. This demonstration was accomplished by constructing two networks which, though indistinguishable by history relations, cannot be substituted for each other in a larger network without changing the history relation of the larger network.

A few alternative theories for non-determinate networks have been proposed. Unfortunately, these theories offer complicated semantic models revealing many details of network computation, far more than is often relevant for an interface specification. We have introduced a relatively simple network model, the set of scenarios. This model is a straightforward extension of history relations, for a scenario is merely an element of a history relation whose constituent values are partially ordered by a relation representing causality; but it can reveal subtle, though important, distinctions unseen in history relations.

This research was supported by the National Science Foundation under contract 7915255-MCS.