Greg A. Woods wrote:
> I would *really* prefer if that were not the way they were implemented.
> 
> A dynamic interface explicitly for loading and *unloading* character
> codeset translation tables from the kernel would be fine, but the LKM
> scheme is *not* the appropriate interface for a production kernel on any
> system where security is important (especially since if LKMs are used in
> a production kernel they should only be loadable during boot and prior
> to multi-user startup and from then should not be un-loadable or
> re-loadable).

I'd need to pass not only data/tables, but also actual code -- the
implementation of encoding - i.e. "none" for iso-8859-[12], koi8-r;
"euc" for euc-jp; "utf-8" for utf8.  Some code sets may also
implement the conversion in software instead of using static tables.

LKM seems like natural choice here.

Jaromir
-- 
Jaromir Dolecek <jdolecek@NetBSD.org>      http://www.ics.muni.cz/~dolecek/
@@@@  Wanna a real operating system ? Go and get NetBSD, damn it!  @@@@