> Thanks for the pointer about "patch -p0".  I *HAD* tried that, but it
> didn't seem to work either.  I'll double check that though...  "svn
> diff" is what I'd prefer, because then I can "svn commit" it when it's
> ready.

Ok, it seems the code inside crypt.py is duplicated in your patch.
Also, when you commit, it'll be better if you use "svn rename" for the C
file, so that history isn't broken.

> Any other review feedback?  I'll probably let this sit until 3.2 goes
> to maintenance and then check it into trunk, so there's some time
> yet...

Looks good mostly. Why do you need _MethodListClass()? Executing code at
module startup sounds fine to me. Or, at worse, use a global variable.

+   *salt* (either a random 2 or 16 character string, possibly prefixed with
+   ``$digit$`` to indicate the method) which will be used to perturb the
+   encryption algorithm.  The characters in *salt* must be in the set
+   ``[./a-zA-Z0-9]``, with the exception of Modular Crypt Format which
+   prefixes a ``$digit$``.

That paragraph is a bit confusing. Also, other uses of *salt* are
described separately two paragraphs above.