I see.

On Tue, 2009-02-24 at 15:11 -0500, Andrew Pochinsky wrote:
>  However, once we have Unicode, the  
> complexity of staying CI grows tremendously and threatens to result in  
> a ball-of-hair design. 


The call to which I'm responding was the 
suggestion by a few that It Would Be Nice
if Scheme defined mechanisms by means of 
which both CI and CS environments could be
hosted and could share code.

If you unpack that request it turns out
to touch on a fairly deep, essential question:
the concept of "identity" among symbols in 
mathematical / programmatic notations.

The desire, among serious programmers, with real
needs for both CI and CS helps to give an example
in support of a hypothesis...  The hypothesis 
is that symbol identity doesn't have a single, 
right answer.

"Exact integer 2, self-multiplied, yields exact
integer 4."

There is no question about that.  It follows from
the ordinary meanings of "integer" and "multiply".
The standard doesn't really arbitrarily exclude 
anyone - doesn't give us an unnecessary restriction 
- if we all agree that, in Scheme, two times two is
four.

"Symbol names are CS" and "Symbol names are CI" is
demonstrated, in this thread, not be like 2 * 2.
Each is a valid choice, depending on context, and
the *consequences* of the choice matters non-trivially
in some contexts.

So a good question for the standard to pose is
"What is the simplest, most general, least restricted
thing we can offer that let's people choose CS and CI?
or that at least let's implementers choose to give
users the option in a way portable among implementations
that offer such a choice?"

I'll write something quick but separate about the
problem with how people regard the standard.



> > I suggest that that is more like an EQ? / EQV?
> > distinction.
> 
> If you mean that to make some symbols eqv? but not eq?, I find it a  
> very unappealing change in semantics.


But done in the way I suggest it would break no 
existing code.

-t



_______________________________________________
r6rs-discuss mailing list
[email protected]
http://lists.r6rs.org/cgi-bin/mailman/listinfo/r6rs-discuss

Reply via email to