Re: File name completion with non-ascii filenames

[Richard Stallman]

    However, since working with multibyte strings means we need to decode 
    each file name returned by readdir, to address Richard's concerns, I'd 
    like to try to avoid decoding even if completion-ignore-case is non-nil.  

I pointed out there's a possibility that decoding will be slow--but we
don't know how slow it will be.  Perhaps it won't be slow enough to
matter.

I suggest trying it first, and seeing how much of a slowdown it causes
for a large directory.  Then you can tell which optimizations are
worth spending time on and which are not.