Errors & software requirements tend to be more costly to correct the later they are detected. Use of automated requirements tools can help detect errors early in the software life-cycle. The paper describes a case study that demonstrated the capabilities of automated requirements tools in detecting errors in the requirements for a large management information system. Three different tools were applied to the same requirements document. Problem Statement Language/Problem Statement Analyzer ( PSL/PSA ), Input/Output Requirements Language ( IO RL ), and Software Requirements Engineering Methodology ( SREM). Each tool helped identify several hundred errors, many of which were associated with failures that occurred after system delivery.
specification, requirements, correctness, errors, failures, case study
Errors in software requirements are common and are often not discovered until after the system is delivered. The later errors are detected, the more costly they are to repair 1, 2. It is therefore important to identify as early as possible the most prevalent types of errors, namely, omissions, incorrect facts, inconsistencies, and ambiguities 3.
Automated tools (associated with structured languages) seem more appropriate than manual techniques (with natural language) for developing large system requirements, i.e., when it is crucial not to have the requirements misunderstood. Recent studies of structured requirements techniques, most of which are supported by automated tools, suggest that their use may help reduce requirements errors 4"5. Little research has been done, however, to verify the effectiveness of automated tools in detecting errors in software requirements.
The US Army Institute for Research in Management Information, Communications and Computer Sciences (AIRMICS) conducted a field study to investigate the usefulness of requirements languages and analysers in discovering errors in the requirements document of a large management information system. A secondary objective was to determine the extent to which these requirements errors were associated with failures that occur after system delivery. The notion adopted here is