(I'm writing this in debconf, but the network is down and I'll start
travelling round from the day after tomorrow without access so I have no idea
when this will be send out.)
I talked here to i18n, i10n experts about the question "what encoding to use
for FFI and filenames" the answer is: you do not know in general.
The linux kernel talks utf-8 and accepts filenames as the program gives them.
MacOs X on the other hand also talks utf-8, _but_ does a 'normalisation'
before writing the data to disk, so if you use that disk with linux you'll
have to recreate that encoding. Windows does have 'wide character' UCS16
interfaces you can use.
For libraries 'it depends', gnome-1 talked inthe locale encoding to you (as I
thought), but gnome-2 always talks UTF-8. Database libraries in general allow
you to select the encoding.
So I propose to use UTF-8 when talking to the kernel, and make this the
default encoding for FFI, but allow the programmer to change the encoding
Oh, when I asked these Japanese guys how they handled the filename problems
they said 'we use latin-1 characters when we need to transfer files'. Bugger.
signature -at- pvaneynd.mailworks.org
"God, root, what is difference?" Pitr | "God is more forgiving." Dave Aronson|