HOW RELIABLE ARE THE REQUIREMENTS FOR LARGE DIGITAL SYSTEMS?

摘要

Requirements for large information systems are not reliable if we interpret that term to mean complete, consistent and current. A General Accounting Office examination of 10 major systems shows deficiencies in generating complete requirements and in keeping them consistent and current. Requirements analysis tools, formal methods, and recursive development formats are beneficial but do not address organizational and contractual issues. As an alternative to the current methodology we discuss a figure of merit as the key requirement and giving the developer more responsibility for trade-offs during development.