bpo-37751: Document the change in What's New in Python 3.9 by vstinner · Pull Request #17997 · python/cpython

encodings.normalize_encoding() says "Note that encoding names should be ASCII only." You're correct: "КОИ-8" is normalized to "8" by codecs.lookup() because the C function _Py_normalize_encoding() ignores non-ASCII letters.

I don't know which behavior is correct. It sounds strange to me to have a non-ASCII encoding name. Which encoding is supposed to be used to encoding the encoding name?!? :-D Maybe encodings.normalize_encoding() should also ignore non-ASCII letters, be more strict.