In the discussion on issue 18814, Antoine pointed out that in a Python 3 world, using backslashescape during decoding actually makes sense - it lets you accurately report arbitrary bytes in the sequence, without needing surrogateescape or surrogatepass to be used when encoding later.