Okay, we've been going back and forth on this.
Functionality wise, 2.7.0 has json serialization in it (a good thing) , but something is wonky with some encoding..

I would prefer to use 2.7.0 vs 2.5.2 but I don't know all the encoding issues people have been suffering, or the details.  My own experience is that it's killing my unit tests, which is why we are back to 2.5.2 at the moment.

Does anyone know any details ?  Kwatters said he tried using jvm parameters to fix 2.7.0 encoding.  Do we know what is going wrong ?

Resistance is Futile !