Domain descriptions should be modular (preliminary report)

Abstract In this work we address the problem of what a good domain description for reasoning about actions should look like. We establish some postulates concerning this sore spot and point out the problems that arise when they are violated. Such problems can be overcome with the algorithms we propose