I think there are only a couple instances where we'd not know of a problem at compile time. I'd hate to never be able to purge defunct dependencies due to a lurking suspicion that 'something' won't work -- and there were clearly plenty. Probably such subtle dependencies deserve an XML comment.
But why not put back whatever dependency you have in mind, if there's reasonable doubt? rather than rollback and throw the baby out with the bathwater. Sean On Sun, Feb 14, 2010 at 10:24 PM, Grant Ingersoll <gsing...@apache.org> wrote: > I committed it. In general, I think we should probably revert the dependency > removal changes. I highly doubt Maven static analysis is going to be correct > here. Dependencies are usually added because otherwise the thing won't run. > For instance, some of the Hadoop dependencies that I think were removed are > optional for some file systems and won't work otherwise.
