unless it's possible to declare on the jar_library for hadoop the various ways it will be consumed