WG211/M16Carette

I will present some ongoing work that seeks to generate all the artefacts involved in software (obviously code, but also specification documents, design documents, tests, user manual, Makefiles, etc). In the context of software which requires (re)certification, all of these artefacts are involved -- and they normally contain a huge amount of duplicate information. Our approach is to do very aggressive knowledge encapsulation, followed by relatively straightforward generation passes. For domains (such as scientific computation) where there is well-established theory, our preliminary experiments shows that this works quite well. Note that we do NOT expect this to work so well in domains without well-established theory.