Considerations on current research issues in software safety

The current debate about software safety is the result of basic disagreements, among both researchers and practitioners, about the usefulness of proposed techniques and possibly the very goals to be pursued. The underlying issue is the problem of predicting the safety of a software or software-based system, or the effectiveness of a method for obtaining safety. An argument is made here in favour of quantitative, probabilistic evaluation of safety. Then, some technical problems and research results are reviewed. These include design methods for predictable timing behaviour, the roles of formal methods and of design diversity in achieving safety, and the limits to the levels of software safety that can be assessed or certified.