[Python-Dev] PEP 277 (unicode filenames): please review

[Martin v. Loewis]

Jack Jansen <Jack.Jansen@oratrix.com> writes:

> After a few more experiments I did manage to confuse the filesystem
> APIs: it turns out ligatures are not correctly decomposed. I.e. if you
> create a file "\uFB03" you cannot open it as "ffi".

LATIN SMALL LIGATURE FFI is a compatibility character. Those are not
normalized under NFD, only under NFKD (in which case it would decay to
ffi). Since NFKD loses information (of typographical nature in this
case), NFKD is only recommended for restricted domains (identifiers
being an explicit example).

Regards,
Martin