Introduction

Abstract

In our every day conversations we tend to use the terms ‘fault’, ‘error’ and ‘failure’ (often interchangeably) to indicate the fact that something is ‘wrong’ with a system. However, in any discussion on reliability and fault tolerance, more precision is called for to avoid confusion. The definitions for these terms presented here in the first paper by Anderson and Lee owe their origins to the unpublished work of Melliar-Smith and to his collaborative work with Randell (see their joint paper in Chap. 3). In true computer science fashion, Melliar-Smith and Randell defined a system recursively as composed out of ‘smaller’ systems and defined the occurrence of a failure to be the event when the behaviour of a system does not agree with that required by the specification. Why does a system fail? To answer this is it necessary to examine the internal state of the system, which then leads us to the notions of ‘errors’ and ‘faults’.