On Sun, Feb 14, 2010 at 6:03 PM, Sean Owen <sro...@gmail.com> wrote: > Could be a dumb question, but aren't these sorts of dependencies > necessarily things that particular users of Mahout/Hadoop depend on, > not Mahout? even at runtime? Kind of like depending on a particular > SLF4J binding to a logger.
It depends, mvn will warn that jets3t is unused, but the code in MAHOUT-249 depends on it because it advertises the ability to write to S3 in its command-line options. Do we want to add the caveat that if you want to use s3 that you (as a user) have to provide the jar? That's the one example I'm aware of, are there others, I don't know. One could say that because there are utilities/examples that don't run using the hadoop driver but do use the Hadoop FS abstraction that we must include all of the jars for filesystems supported by that abstraction, but I'm not certain that I agree with that statement or not. I believe the Hadoop FS is probably the only case where we'd need to consider this but I'm not 100% certain. The logger binding is a somewhat similar class of problem. Drew
