Michael Sperber wrote:
> You want to argue this
> technically, see how far to the right my screen will display this
> thread? Be my guest.
OK, thanks.
> While we're at it: Asking that Scheme symbols be changed to accomodate
> XML identifiers is like asking that Scheme numbers be changed to
> accomodate XML identifiers. Symbols are the wrong type. There
> already is a Scheme type for arbitrary "cases-sensitive" sequences of
> characters, called string. (If we accomodate XML, why don't we
> accomodate ML or Haskell? Or mathematical identifiers with arbitrary
> accents, superscripts and subscripts?)
XML is just one example, and ML and Haskell certainly aren't precluded.
I believe Shiro Kawai raised the example of case-sensitive X11
identifiers. Mathematical identifiers? Great idea, probably out of
scope for R6RS.
I personally like programming in a case-insensitive language. However:
A case-sensitive Scheme accomodates embedded case-insensitive languages
(or just embedded symbols & identifiers) very well, since it's almost
always straightforward to avoid depending on case-insensitivity in
embedded code.
By contrast, a case-insensitive Scheme cannot conveniently accomodate
embedded case-sensitive languages.
For the record, though, I agree[*] with Chris Hanson, who wrote:
> None of the above. As I have been saying, I prefer selectable case
> sensitivity with an implementation-defined default.
However, if standardizing on selectable case sensitivity is seen as too
complex (or whatever), then the most flexible fallback position is case
sensitivity, because of the inclusivity issue mentioned above.
The argument about breaking backward compatibility seems less strong in
the R6RS context than it might otherwise be, because few significant
R5RS programs will be valid R6RS programs anyway, and practical
implementations will continue to support old, non-R6RS code.
Anton
[*] I suggested something similar on the editor's list:
http://www.r6rs.org/r6rs-editors/2006-February/000929.html
Received on Thu Nov 23 2006 - 15:10:32 UTC