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.
Sure, I might depend on Hadoop/S3 and Mahout in my project, but Mahout does not need to depend on Hadoop/S3 to accomplish that and should not in fact. Or is the issue around examples, which are like users? On Sun, Feb 14, 2010 at 11:00 PM, Drew Farris <drew.far...@gmail.com> wrote: > It becomes difficult when dealing with projects with large amounts of > transient dependencies, especially for things like hadoop. I guess > despite the desire to keep the number of dependencies small, it is > generally better to err on the safe side. > > 100% agree with your point that we can't rely on maven static analysis > or the junit tests to tell us when we got the dependencies wrong. >
