Here comes a new version of the patch.  I believe, I addressed all review comments made on issue12319_7.patch, with the exception of the EncodingError mentioned in the last post.

Most relevant changes compared to last patch:

 * Do not try to dertermine the content length for text files.
 * Change priority of object types for the body: try file like before byte like.
 * Convert byte like to bytes before feeding the body into the socket.  This fixes Issue 27340, at least as far as http.client is concerned.
 * Split test_http() in one separate test for each body type in test_urllib2.py.
 * Review documentation.


Another question is: what is the best method to test whether a file like is text or binary?  For the moment, I kept the original code that essentially does:

  try:
    isText = "b" not in readable.mode
  except AttributeError:
    isText = False

This logic fails for StringIO and possibly others.  Alternatives could be:

  isText = isinstance(readable, io.TextIOBase)

  isText = hasattr(readable, 'encoding')

or

  isText = isinstance(readable.read(0), str)

What do you guys suggest?