Message2814

It makes more sense for this check to live in __findattr__(PyString) instead of __findattr__(String).  The first place would allow you to do an instanceof PyUnicode check instead of making a PyString out of every String that comes in to __findattr__.  __findattr__ is used from all over the place, so slowing it down is a bad idea.  If some Java code is calling __findattr__(String) with unicode Strings, this should be fixed there rather than in PyObject.  

Also, you're using tabs rather than spaces for indentation.