The interface between information, terminology, and inference (problem solving) models is now a central issue for the development of standards and of principled information systems. There is potential overlap amongst the models, and principled means are required to choose how to allocate information items amongst them. Any given set of choices implies mutual constraints and obligations amongst the models which together form a well defined interface allowing work on each to proceed independently of the other. This paper outlines the criteria for these choices and a possible notation for recording these choices and their consequences.