[R6RS] draft Unicode SRFI

[Matthew Flatt]

At Fri, 8 Jul 2005 16:33:14 +0200, "Manuel Serrano" wrote:
> > Meanwhile, you might browse the MzScheme source to get ideas on using
> > the C library to get passable locale support.
> I don't feel comfortable with the idea exposed by Marc. Implementing
> in C is not always possible. Sometime you want to implement Scheme on
> a machine for which you don't have access to C (e.g. a JVM base
> platform). An implementation should be reasonable even without C.
> I don't think that (even for a prototype) we should rely on C.

I'm not sure how this relates to my reply, but in an effort to clarify:

 * I've removed all specification points related to locales, except
   to note that the specified operations are locale-independent.

 * I am currently not advocating any standard for locale-sensitive
   operations. Maybe the right thing is a library (like ICU for C and
   Java), instead of a standard.

 * Independent of any standard, an implementation might provide
   locale-dependent operations based on the C library. I agree with
   Marc that it's not obvious how to get good results with the C
   library. Since I've sorted through many possibilities for MzScheme,
   the MzScheme implementation might be a useful reference. Clearly,
   it's not useful to an implementor building on the JVM, and perhaps I
   should have taken this comment offline to avoid confusion.

Matthew