RFC4648 specifies two different alphabets for Base32: 1) the common one already used in RFC3548 and implemented in Python, and 2) an "Extended Hex" alphabet which is used by e.g. DNSSEC NSEC3 but not yet implemented in Python.

The attached patches add the "Extended Hex" alphabet to Python 2.7 and 3 via an optional parameter 'base32hex=False'. Default behavior is not changed.

patch Lib/base64.py < py27_base64.patch
patch Lib/test/test_base64.py < py27_test_base64.patch

Since this is a new feature it can go on 3.4 only.
It would also be better to include both the changes for base64.py and test_base64.py in a single patch and attach that, rather than uploading an archive that contains both.

FTR the links to the relevant RFCs are:
http://tools.ietf.org/html/rfc3548
http://tools.ietf.org/html/rfc4648

Added single patch as requested.

The patch can't be longer applied cleanly. It should be synchronized with sources. It is the last chance for this feature to get to 3.5 if somebody will update the patch and add the documentation.

The feature itself LGTM. The only my doubt is about the name of boolean parameter. "base32hex" is recommended as encoding name. Parameter can be named as "exthex" or "extended_hex". Or may be add separate functions b32hexencode/b32hexdecode. I'm not sure.

I've created a new patch that works against the current 3.5 sources. Should be fine for 3.6, I guess.

Separate functions b32hexencode and b32hexdecode are used now. There is no optional parameter "base32hex" anymore.

Matthäus, I think you uploaded the old patch again by accident. Also, see Berker’s old comments about documentation, and my new suggestions, on the Rietveld code review, if you haven’t already.

*facepalm* Yes, I uploaded the old patch twice. Sorry for that.

- Added doc update.
- Added test case that includes all 32 characters codes.

I'm reusing the existing Base32 table generation logic without changes. It has been changed (in 3.4 or so) since the first patch. If the logic is still to be changed as suggested by Martin in Rietveld review, I'd suggest to open a separate issue because it affects the existing Base32 implementation, too.

Added comments on Rietveld.

Is there interest in having this patch converted to a pull request?

New changeset 4ce6faa6c9591de6079347eccc9e61ae4e8d9e31 by Filipe Laíns in branch 'master':
bpo-16995: add support for base32 extended hex (base32hex) (GH-20441)
https://github.com/python/cpython/commit/4ce6faa6c9591de6079347eccc9e61ae4e8d9e31

Paul, the PR is now merged :). Can you close the bug?

Thanks Filipe!