On Thu, Jun 02, 2005 at 08:29:24PM +0100, Roger Leigh wrote: > Forcing the Linux system call interface to require valid UTF-8 would > be a fantastic extension to POSIX. (Generic, not per-filesystem.)
I do agree, however, it'd introduce several nontrivial problems that'd still needed to be solved somehow: - zip extracting :-), web/ftp/rsync mirroring etc. could cause files not to be extracted/mirrored. - What to do with non-UTF8 files that already exists on the disk? Should e2fsck, reiserfsck etc. also force filenames to be UTF-8? Could they still be accessible somehow? Would readdir() return them? - Is it then sane to limit filename length to a fixed number of bytes instead of a fixed number of characters then? - Would one of NFC and NFD be forced (� la MacOS)? Perhaps there are other problems as well... Once I also heard a fantastic idea from someone: this could (at least temporarily till everyone switches to UTF-8) be a per-filesystem mount option instead of a global kernel variable or hardcoded to the kernel. This way when you mount a filesystem, you could specify whether or not UTF-8 is forced there. -- Egmont -- Linux-UTF8: i18n of Linux on all levels Archive: http://mail.nl.linux.org/linux-utf8/
