I have cleaned up the changes and ensure the behavior remains the same, however there are still a few points I need input from@malemburg
(And as Benedikt said, should be their own issue)

• This function is documented as taking strings, but during the 2->3 conversion and undocumented, and untested change was made which allowed it to accept bytes. I have kept it this way (in Python, to make removal simpler), though I think this should either be documented and tested, or removed.
• The function has been documented asascii only, and forbytes, it is. However, for strings, it has not been enforced with an error. What should we do?

Benchmarks:

import timefrom encodings import normalize_encodingimport pyperfdef bench(loops):    range_it = range(loops)    t0 = time.perf_counter()    for _ in range_it:        normalize_encoding('utf_8')        normalize_encoding('utf\xE9\u20AC\U0010ffff-8')        normalize_encoding('utf   8')        normalize_encoding('%%%~')        normalize_encoding('UTF...8')    return time.perf_counter() - t0runner = pyperf.Runner()runner.bench_time_func('normalize_encoding', bench, inner_loops=30)

Main branch:

normalize_encoding: Mean +- std dev: 173 ns +- 7 ns

This PR:

normalize_encoding: Mean +- std dev: 42.9 ns +- 1.1 ns

Sorry for the lack of response. I'm currently at EuroPython and pretty busy with other things. I'll have a look on Saturday during the sprints.