Graphs and graph rewriting techniques are well suited for describing structure and behaviour of systems. In order to be practical for large and complex systems these techniques have to be modular and scalable. In this paper we consider as a realistic example the problem of describing interaction in CORBA-based distributed systems. From the desire to formulate semantic properties we derive requirements for a modular graph rewrite approach. We use the interaction problem of distributed objects to plot out how such a modular approach to graph rewriting and how existing approaches can be used or integrated.