I'm neck deep in hadoop right now so I can't tackle this general refactor, though I strongly agree with you that this should be using the

compile_classpath

product