I don't really know why we were only decoding to unicode in Python 3 only. Just make sure that everything is in unicode. Should fix any remaining issues with docstring display.
ENH: Always use unicode.
Sounds correct to me, if we don't have only ascii or system encoding strings, then we have to take care also of unicode in python 2.
I assume since Vincent works on Linux, all the rdatasets will be utf-8 and we should be fine if we use it everywhere.
Yes, I think all of the docs are HTML in utf-8.
Working on a macbook nowadays :) But yes, pretty sure everything should be utf8