Re: Ideas for v2.1

Albert Cahalan (albert@ccs.neu.edu)
Sat, 15 Jun 1996 08:00:20 -0400 (EDT)

From: hpa@transmeta.com (H. Peter Anvin)
> Marcus Meissner <msmeissn@cip.informatik.uni-erlangen.de> wrote:
>>
>> UNICODE-able Filesystem layer? (As the win32 api is at least prepared for.)
>
> We already have that! Better yet, we don't need bulky conversions in
> the kernel. It has already been agreed that Linux will use UTF-8 for
> Unicode exchange, and ext2fs handles UTF-8 just peachy-dandy

We do need the layer. I'm quite sure I'd get a mess if I mounted
an ext2 partition with UTF-8 encoded filenames as /usr and an ext2
partition with Latin-1 encoded filenames as the root.

Not only would I get a characterset mismatch that I can't cure
with terminal settings, but I get characters that are invalid.
I either have 0x80 to 0x9F messing up my Latin-1 display or I
have 0xFE and 0xFF (maybe more) messing up my UTF-8 display.

File _contents_ are easy to fix from userspace. Filenames
are a bit more difficult, as I'd have to convert them all
at once. (with some crufty hack made from perl and recode?)

The kernel needs characterset translation anyway for the console
and the vfat filesystem.