After decades of research, and despite significant advancement, formal methods are still not widely used in industrial software development. Industry practitioners believe that the methods and tools coming out of academia are, to a large extent, irrelevant and ineffective in tackling real-life projects. They are difficult to use, esoteric, and do not scale (at all). This paper reflects more than twenty years spent in first experiencing the problems in industry, and then struggling to do something about it in academia. We present some examples of formal method madness/blindness, as well as a few successes. We believe the problem is deep. To start to address it and make progress in producing methods that are truly usable in industry, and rigorous enough to make them effective engineering methods, we need to reconsider the role of computer scientists, software engineers, software developers, as well as the hurdles to promotion for academics. Along the way, the paper will present a few fundamental principles that we think spell the difference between success and failure in producing usable formal methods, and convincing software professionals in industry to adopt them.
展开▼
