On Sun, 2007-12-16 at 13:18 +0100, Arnt Gulbrandsen wrote: > > Anyway having to specifically track a list of valid and non-valid > > strings and sort them separately would require adding a lot of new > > code, all just to comply with a specification that works worse (from > > user's POV) than what I could have implemented much more easily. > > As I see it, you're adding support for all codepoints and all character > sets, When it gets to collation, yes, but is this really allowed for implementing SORT? > so the exception we're discussing no longer applies to your code. Why would anyone do anything different then? I see only benefits doing it this way.
Description: This is a digitally signed message part