[java] Mina sensitivity

[Steve Vinoski]

I'm not sure how or why we decided to move to a snapshot build of the  
Mina stuff, but I think it's potentially going to cause big problems.  
In the test/maven work I'm doing this weekend, I let maven pick up  
the 1.1.0-SNAPSHOT of the Mina stuff, assuming it was equivalent to  
whatever snapshot we've committed to the local lib directories. That  
was a bad assumption -- I finally figured out that when we run  
against the Mina snapshot jars from the maven repository, the  
PoolingFilter is not initialized or destroyed correctly, and many  
tests fail miserably.

This is bad, obviously, as it greatly decreases Qpid stability. Can  
anyone explain why we moved to what seems to be a very particular  
interim snapshot of the Mina stuff? I'll continue to poke around and  
see if I can fix Qpid to work with the maven snapshot, but in  
general, do we really want to expose ourselves to what seems like  
unstable code? Should we instead go back to 0.9.5?

--steve