UTF-8 and local charsets in the VFS layer

[Joerg Sonnenberger]

On Fri, Mar 12, 2004 at 10:39:32AM -0800, Matthew Dillon wrote:
> :We are currently in a situation, where the kernel doesn't care
> :for the local filesystem, has/could have some conversion hooks for 
> :removable and remote filesystems. E.g. ISO-9660 with Joliet or RR extension
> :...
> 
>     Well, I guess the question is: what exactly would need to be done?  E.G.
>     our normal UFS filesystem can use any character from 0x01-0xFF, other then
>     '/', in filenames, so it should have no problem storing UTF-8.  It kinda sounds
>     like this is a userland issue rather then a kernel issue for the most part.

I whole agree on this. But to support this for userland, you have either to
add an encoding flag to statfs or provide a consistent encoding. I strongly
prefer the later since it makes things easier and safes apps from having
to lookup the encoding for each directory or even file name.

The current trend goes to using one Unicode Encoding or another and it would
make sense to use it too. Nothing prevents or forces us to enforce UTF-8,
but it has the handy advantage of being mostly backwards compatible.

Joerg

> 					Matthew Dillon 
> 					<dillon at xxxxxxxxxxxxx>