On Thu, Nov 18, 2004 at 01:27:12PM +0300, Valeriy E. Ushakov wrote:
> On Thu, Nov 18, 2004 at 12:57:21 +0900, MINOURA Makoto wrote:
> 
> 
> > Anyway, I do not think all of the above can be solved in a
> > single solution:), but it is worth discuss before adding a
> > mount option to specify encoding.
> 
> Not encoding - transcoding!  It's a cheap and effective way to dodge
> all the conceptual encoding problems that you mentioned.  I.e. instead
> of making a "declarative" statement "This filesystem has filenames in
> encoding FOO", we make a "procedural" request "Please apply this
> mangling procedure (transcoding) to the filenames in this filesystem".
> Thus we remain within the scope of the existing "file name is just a
> stream of bytes" paradigm, and it's up to user to ensure the mangling
> does the right thing.
> 
> Yes, this is not "generic enough" to support on a single system users
> that use different locales.  OTOH, they already have enough rope to
> shoot themselves in the foot - a single FFS filesystem, happily
> unaware of all things related to encodings, allow users that use
> different locales to create file names e.g. in Russian and Latvian.
> 
> There's existing practice (FreeBSD) and it worked out really well
> there.  I don't see why we can't adopt it.

Yes, this should be the most simple and convenient way. But we can use unicode
to represent file names internally. Then if we specify translation flag in
mount options we translate names from unicode to local charset for userspace.