Re: Proposal for 2 Byte Unicode implementation in gcc and glibc

[Ping Yeh <pyeh@xxxxxxxxxxxxx>]

Florian Weimer wrote:
> 
>   Ulrich Drepper <drepper@xxxxxxxxxx> writes:
> 
> > "Too late"?  If *never* would happen.  Every sane system has wchar_t a
> > 32bit type.  Even the AIX people realized this and will switch to a
> > 32bit type once they can figure out how to handle the compatiblity
> > problem.
> 
> Hmm.  Why is UTF-16 so crazy? ;-)
> -

But UTF-16 is not fixed width [surrogates], thus not in the same
ball park when we are talking about wchar_t.

The wchar_t should be defined such that it is wide enough for
all characters to be used by a program.  That means 20 bits for
unicode, which leads to 32 bits type.

I know it is odd, but using 24 bits is enough and saves 25% space... 8)
Does anyone care about memory alignment for strings?

Ping
-
Linux-UTF8:   i18n of Linux on all levels
Archive:      http://mail.nl.linux.org/lists/