The terminology of the field of dependable computing is used non-uniformly by many authors and standards. This paper presents a survey of definitions drawn from a variety of sources in an attempt to abstract away from the terminology and focus instead on the fundamental concepts underlying the definitions. Two results are reported: first, a characterization of 'dependability' as a subjective portmanteau of 'entities' and 'attributes;' second, a simple, yet rigorous, model for 'errors, 'failures' and 'faults.' The latter model leads to a careful examination of the mechanisms by which undesired situations occur and propagate in complex integrated computer systems. The significance of these results is discussed.
展开▼
