Measuring the Effect Of Formalization

We present an ongoing research activity concerned with measuring the effect of an increased level of formalization in software development. We summarize the experiences from a first experimental development. Based on these experiences, we discuss a number of technical issues; in particular, problems connected to metrics based on fault reports. First of all, what is a fault? Secondly, how should the fault counting be integrated in the development process? Thirdly, any reasonable definition of fault depends on a notion of satisfaction. Hence, we must address the question: What does it mean for a specification or an implementation to satisfy a requirement imposed by a more high-level specification?