On Sat, 30 Mar 2002, Martin Husemann wrote:

> > AFAIK, amiga and msdos never supported unicode or kanji.  I don't think Amiga
> > had other charsets than 8859-1.  Msdos had only one charset, too (but it
> > was different from the one used by windoze, of course).
>
> No, msdosfs (the fat32 variant) is used on japanese and other windows
> versions. They do store file names in multibyte character sets, I've
> used a japanese windows with a DBCS locale (I think that was code page 931)

Also, there is all of the unicode stuff we have hiding in ntfs right now.

> > IMHO, it would be better to implement simple case-insensitivity support
> > than do nothing.
>
> I aggree.

Well, I'm concerned about adding this as it stands. We have a limited
supply of mount flags, and once we allocate one for this, we are stuck
with it. To be able to re-use it, we have to version the mount system
call.

I also find it a bit ironic that an American is advocating more indepth
multi-language support than the Europeans. :-)

> If/when we come up with a more general solution (something like the "struct
> emul*" pointing from each process to it's active ABI description), we can
> make the simple routines use that.
>
> But I'd suggest we start with the simple case now. Of course, if our friends
> from the multi-byte locales join in and we come up with a general solution
> soon, we can do it right at first try.

I'm not so much championing we get the full-case correct right from the
start, but that we have a good idea of the full-case _design_ from the
start. Then implement the "simple" case as a first step.

If we have a plan that should give us full multi-lingual localization,
then doing it just for say 8859-1 for now would be fine.

I don't think this will be hard to do. Corner some of the multi-byte
locale folks and get their input. It shouldn't take more than a week on
the mailing list. :-)

I think this is important. Until now, the kernel has had a policy of
preserving case. Yes, some file systems ignore that, but the kernel on a
whole has done it. This change will be the first time that the kernel
overall starts ignoring case. So we should do it at least mostly right.
:-)

Take care,

Bill