Semi-Automatic Integration of Little Languages.

This paper is about a kind of preview model integration that can be done for complex software managed systems, that maintains the flexibility of well-trusted software and hardware component models (“islands of good behavior”), and makes explicit all of the influences expected for each component. In the systems that have been built using these methods (medium sized mathematical and natural language analysis systems), almost all of the software errors were caught at this stage, long before deployment. The purpose of this paper is to focus on the languages used to describe these mutual influences, making them explicit with sufficiently well-defined semantics, and automating these calculations as much as possible. We do not have a finished solution, but we think that this direction will be very fruitful, as it extends others that we have already found very useful.