ADC> That leaves only the kernel API. The standard way of fixing an
ADC> API will do quite well: alternate system calls for raw 16-bit
ADC> Unicode.
16 bit is not enough to uniquely encode all characters in existing
multi-byte character sets, as has previously been stated.
Therefore you either have to use 16 bit with escapes (Would that be
UTF-16?) with all the disadvantages of UTF-8, or you have to go 32
bit. 32 bits per character is expensive, of course.
Benny