I think we could technically shade hadoop jars, re-publish them, and consume them in many cases