Critical Description
If you use the Atmos storage driver, and you attempt to stream the upload of an object, and either your container name or your object name is a unicode string, the presence of these unicode strings will cause the HTTP message body to be converted into a unicode string.
However, file content is provided as a byte string; if the file content contains binary data, httplib will try to convert this file content into unicode, yielding encoding errors.
For example, if you try to stream upload a PDF whose name is stored as u'foo.pdf', you'll get a message something like:
UnicodeDecodeError: 'ascii' codec can't decode byte 0xc4 in position 10: ordinal not in range(128)
(Position 10 is where the binary content in a PDF starts, after the "%PDF-1.3\n%" header)
The behaviour of httplib in the presence of unicode content is a known issue (http://bugs.python.org/issue12398); All path tokens should be encoded as ascii before being passed to httplib to prevent this problem occurring.
As far as I can make out, this problem only exists under Python 2.7 – I've observed it on Python 2.7.1 and Python 2.7.3. Python 2.6.7 is unaffected.