quopri.py's functions encodestring and decodestring are documented to handle strings; and this is clearly suggested by their name. However, these functions accept and return bytes, not strings. This should be reflected in the documentation. 

Even better: deprecate these functions and introduce new ones with behaviour as suggested by their names: encode_string, encode_bytes etc.

The encode() and decode() functions also expect binary file objects.

Attached a patch that changes documentation and docstrings. Removed the mentions of readline(), because read() is called on input instead, if the binascii module is available. Also removed the repeated descriptions of quotetabs and header in encodestring()'s docs.

Can we have a review on this fairly small patch please.

Please do review it, Mark.

Just one small query on the patch (more for my own benefit than anything else).  In the rst file there's now no difference between the wording for the quotetabs positional argument to encode and the keyword argument to encodestring.  Is there a rule (of thumb) that covers this situation?  Other than that LGTM.

No, that's a good catch.  For a keyword argument, we generally do mention the default value in the text, and it looks like we still should in that instance, especially since it *is* different that it is a keyword argument and not positional like the method to which the text is giving a back-reference.

New changeset b4130b2f7748 by Senthil Kumaran in branch 'default':
merge from 3.4
http://hg.python.org/cpython/rev/b4130b2f7748

Thanks for the review, Mark. Addressed that and committed the changes in changeset 606a18938476 (3.4)
changeset b4130b2f7748 (3.5)