A Case for Specification Validation

Abstract

Abstract. As we are moving from a traditional software development process to a new development paradigm where the process it largely driven by tools and automation, new challenges for verification and validation (V&V) emerge. Productivity improvements will in this new paradigm be achieved through reduced emphasis on testing of implementations, increased reliance on automated analysis tools applied in the specification domain, verifiability correct generation of source-code, and verifiably correct compilation. The V&V effort will now be largely focused on assuring that the formal specifications are correct and that the tools are trustworthy so we can rely on the results of the analysis and code generation without extensive additional testing of the resulting implementation. Most effort has traditionally been devoted to the verification problem. In this position paper we point out the importance of validation and argue that if we fail to adequately address the validation problem problem the impact of verifying code generation and compilation will be limited.

A four-way framework for validating a specification

“…Since a verification exercise is concerned with demonstrating that the implementation is correct with respect to requirements [17], this would consist of partitioning the user requirements to form a set and partitioning the implemented components to form another set in such a way that each element in one set is bijectively related to an element in the other set.…”

End-to-End Formal Specification, Validation, and Verification Process: A Case Study of Space Flight Software

“…The desire to raise productivity and reduce costs associated with software assurance has spurred on the adoption within the industry of computer-aided V&V tools and methods of automation. The V&V automation paradigm centers on the use of formal specification models and on the automated generation of source code, where the correctness of the models (or specifications) determines the correctness of the derived system [4]. A driving force on the cost side is to assure early on in the development lifecycle that the formal specifications represent the correct understanding of what the system should and should not do, and the verifiability of the correct sourcecode generation based on these specification models.…”

“…End-to-End Formal Specification, Validation, and Verification Process: A Case Study of Space Flight Software the requirements correctly, while avoiding the well-known garbage-in garbage-out principle-in this context, poor specification leads to poor software implementation [2]- [4].…”