Milano is a Pig -> Hadoop -> ProtoBuf stack with goals to improve end users ability to write backwards compatible code and to use a binary format allowing a decrease in processing time.

Current Issues:
Maven doesn't run the tests.
The test files for the StoreFunc and LoadFunc tests no longer exist. They had propritary data in them which made them unsuitable for open sourcing.
The stack isn't 100% complete; there are missing pieces in the hadoop/pig layer, but core is solid.
There are bound to be tons of issues this code hasn't really been load tested nor profiled, buyer beware.

Future improvements:
Fix the issues above. :-P
Create a more hadoop friendly file format.
Support supplied descriptors instead of relying on the dynamic headers.