The issue can be summarized with this code:

>>> urlparse.urlparse(u'http://www.python.org/doc')
(u'http', u'www.python.org', u'/doc', '', '', '')
>>> urlparse.urlparse('http://www.python.org/doc')
(u'http', u'www.python.org', u'/doc', '', '', '')

Once the urlparse library has "cached" a URL, it stores
the resulting value of that cache regardless of
datatype. Notice that in the second use of urlparse, I
passed it a STRING and got back a UNICODE object.

This can be quite confusing when, as a developer, you
think you've already encoded all your objects, you use
urlparse, and all of a sudden you have unicode objects
again, when you expected to have strings.