Model-driven development aims to enable the separation of concerns in software development, by allowing different people to work with different models which are automatically reconciled, eventually all contributing to running software. A key motivator for studying bidirectional transformation is the recognition that these models may not naturally fall into a strict hierarchy: rather, the information in them typically overlaps. Recently, the Bx community has started to pay more attention to situations involving more than two models that must be bidirectionally related. A megamodel can specify what models exist and the relationships between them. Among the MDD problems that are not yet fully solved is how to manage the process of reconciling the models, and ultimately building software, especially given that inconsistencies between models may need to be resolved by different means under different circumstances; for example, a bidirectional transformation may need to be applied in one direction or the other. A distinct but related problem is how to build software from a collection of artefacts such as source code files. Here a key challenge is to marry correctness (a successful build always incorporates the latest versions of relevant artefacts) with efficiency (no unnecessary rebuild steps or checks).

This talk describes ongoing work bringing recent advances in provably correct and optimal build systems to bear on the problem of building software from a collection of many bidirectionally-related models described by a megamodel.