Re: File name completion with non-ascii filenames

[Richard Stallman]

      - A user might have enabled an input method that produces
        mule-unicode-* characters, but her file-name-coding-system is
        something like latin-2, which produces latin-iso8859-2 characters
        in decoded file names.

Why would a user use this combination of input method and
file-name-coding-system?  Isn't it an anomaly already?

      - The completion functions could have been called by a Lisp program,
        such as somethink ffap-like, that gleans the partial file names
        from the buffer text.  If the buffer text has mule-unicode-*
        characters, but file-name-coding-system is latin-2, we have the
        same situation again.

Is it really a serious problem to fail in that case?

If we wanted to handle that case, how could we do it?