Lessons learned in a journey toward correct-by-construction model-based development

This paper discusses lessons learned in the attempt to apply the long-known principles of correct-by-construction (CbyC) first promoted by Dijkstra, to modern model-based development practices. We recall the intent and scrutinize the outcomes of a string of research projects that focused explicitly on the pursuit of CbyC by means of model-driven methods and technologies. The lessons learned show that when CbyC extends from the algorithmic and functional dimension to extra-functional concerns, some of the strength of original CbyC concept and its pull dilute. One of the possible causes of that phenomenon, is that – in some situation – the assertive style of algorithm refinement gives way to more tentative exploration of an unknown solution space where the known truths are insufficient to steer the development.