Hi, the code should use `unicode_literals` throughout to avoid using `u'…'` which is one the reasons why Python 3 versions before 3.3 aren't supported. @jayvdb said somewhere (determine), that he'd prefer to do that in one go. I suggest to atleast do it in all new code, so that changing `u'…'` into `'…'` won't break git blame for those files when it's introduced later. It is possible to use `u'…'` on conjunction with `unicode_literals`, but that could confuse newcomers into thinking that `u'…'` is necessary.
One primary advantage is that the usage of `ur'…'` has been disallowed in Python 3 so unicode_literals is a way around it, as that does allow to use `r'…'` with unicode chars inside.
In case we want to revert the `unicode_literals` patch 1e54a7d6 the following patches are solely fixing the original patch and could be reverted with that too:
* https://gerrit.wikimedia.org/r/203652
* a0a194d5ca54
* 0c4197272821
* 853e6b0bdce3