Re: UCS-2 in gunicode.h

[Robert Brady <robert susu org uk>]

On Fri, 7 Jul 2000, Derek Simkowiak wrote:

> 	ran acrossed a character which required 32 bits?  Would you just
> drop that character alltogether?  Treat the 32-bit character as two 16-bit
> characters (which would break your character-count code and possibly not
> be renderable)?  What is the standard behaviour for apps that take UTF-8,
> but only support 16-bit Unicode (i.e., what does Java do?)

The thing to do here is to use UTF-16, which is like UCS-2, but can use
two "Surrogates" to represent characters outside the BMP, thus combining
the worst characteristics of UTF-8 and UCS-4. There are enough bits to
represent characters up to plane 16 - and the committee responsible for
such has promiseed that no characters above plane 16 will ever be
allocated.

-- 
Robert