On Sat, Nov 13, 1999 at 02:12:40PM +0100, Jaromir Dolecek wrote:
> Ignatios Souvatzis wrote:
> > I think you need two conversons:
> > 
> > kernel: filesystem-charset to utf-8
> > then
> > userland: utf-8 to LC_CHARSET.
> 
> I still think that doing the conversion in kernel is more efficient,
> at least memory-wise. Note that normally all programs run in a session
> have same locale. If the conversion would be done in userland,
> each and every program would have to load the whole translation
> table. This would be significant for e.g. eucjp, whose translation
> table Unicode --> eucjp is about 50KB.

This could well be shared, by being part of a read-only section of a shared
library... This way it stays pageable, and has less impact on the system than
in-kernel, where it would be 50 kB lost for all processes.

(50 kB once isn't that bad... but several different ones might be in use!)

	-is