Though that's a good start. The trick lies in identifying when apparantly different behaviour is in fact only different because of some related data - which can thus be abstracted away.

Wade.