On Fri, 28 Jan 2011 20:05:30 -0500, Steve Holden email@example.com wrote:
Antoine is quite correct. The strict process would be to include a note it the NEWS file pointing out that this deficiency was found too late for the necessary fixes to be applied in a controlled manner.
I think a warning in the documentation for the module would also be required (and in 3.1 as well).
body else has time. We can perhaps provide a patch in the sandbox for anyone who is curious about the kinds of problems that can be raised by Unicode issues and some of the progress that has been made towards
Well, the patch is in the tracker.
solving them. The same may be true of David's other partially-funded work on the email package.
I have ideas about that, but I'm waiting until after the release of 3.2 to put them forward.
I am OK with the patch going in, and OK with it not going in. I don't think it is quite parallel to Multiprocessing, though: it is a much smaller codebase, with good test coverage, and the fix itself, although non-trivial, is at its base simple (replace the ASCII-string-only calls to email by bytes-capable calls to email, and allow those bytes in and out via extensions to the appropriate APIs). In addition, the module is already *in* 3.2, it just doesn't work.
I agree with Antoine that the RC phase is not the right place to be doing this. However, it is rather disheartening to think that we need to wait for 3.3 to have this working, whereas if we put the patch in but there turn out to be edge case bugs, we can fix those bugs in 3.2.1...
If the module weren't already broken for most real-world purposes I'd no question want to wait. As it is I'm on the fence.
-- R. David Murray www.bitdance.com