Thanks for the review.  Attached is a new version of the patch.

1) Done.

2) Good point, I didn't think of that.  I've changed it into a class that stores the methods list, and made the module "methods" point to that method on an instance of that class.

3) This entirely depends on the underlying C library implementation of crypt.  It won't cause mksalt() to blow up, it's just that the crypt(3) call won't know how to deal with it.  On my Linux system using glibc, it simply uses the first two characters as the salt, which isn't entirely surprising except that "$" is not a valid salt character according to the standards.

4) I was being lazy and not looking up the locale implications of doing that.  They look fine, so I've changed it to use that.  Good suggestion.

5) I almost did that, but I figured that, generating a random salt, it was obvious that the return value would be different for the same result.  However, since you mentioned it as well, I've added a note.

6) I don't know, I thought everything in Python 3 was a new style class?

7) I don't see that.  Perhaps you mis-read "Lib/test/test_crypt.py" as being another copy of "Lib/crypt.py"?  In any case, I don't see it in v3 or v4 (the one addressing your questions).