A software accident is an unforseen outcome that arises from a failure of a software project or software product. Death, severe injury or severe financial loss, could arise from such a failure. This paper asserts that there is not yet a good accident investigation model with which to investigate software accidents. Accident models from other fields are examined to determine their suitability for use in the field of software development. Although many existing models are very useful, all assume that the environment of the accident was operational. That is, there is a steady state of operations during which events or failures produce an accident. Software development and software operations are very different circumstances. Additionally, none of the existing models makes use of the considerable body of knowledge about software development. For these reasons a new system theoretic model is proposed, extending Leveson 's model, STAMP. The model is briefly described and discussed, and areas or research arising from the model are briefly described.
展开▼
