Compositional reasoning about (possibly large, complex) distributed systems requires tools for reasoning about the flow of information between their components. I describe some first ideas for modelling and reasoning about interfaces between system models, illustrating with examples from security policy modelling.