Conclusions and Further Work

Clearly, it is early and there is much work to be done. What is different about our approach is that the student teams are building real systems for real clients. Thus they face, immediately, the issues of communicating with their client and trying to understand the client's business context and their problem. This is vital. Normal student project experiments are rarely valid, because the whole exercise is something of a sham and everyone knows this. Nobody wants to be able to use the products in real life. The Software Hut approach also creates the desire among nearly all the students to do it properly because they realize that delivering software full of bugs or with an unusable interface just will not do. They have some professional pride and don't want to let the university down. We are convinced that this means that we can really carry out legitimate empirical experiments in controlled conditions and that the results will be meaningful.

This is only the start. We are bound to see, as XP evolves, the emergence of different ways of doing it, using different tools, methods, and notations. This will give us further opportunities to test the ideas in what we call our "Software Engineering Observatory": the Software Hut, for detailed comparative experiments, and Genesys, where we are investigating how new ideas and methods can be introduced into a working software company.