[r6rs-discuss] #!fold-case and #!no-fold-case

From: William D Clinger <will>
Date: Thu, 21 Jun 2007 06:59:30 -0400

First a rant, then a couple of observations.

Historically, implementations of Scheme have used syntaxes such
as #!true and #!unspecified for miscellaneous self-quoting
constants. The draft R6RS abandoned this tradition by making
#!r6rs into a comment, which either complicates the formal
grammar(s) or requires special-case hand-coding in get-datum.
Then the non-binding appendix compounded the problem by
specifying a "region of its effect" for #!fold-case and
#!no-fold-case that is inconsistent with the region of
effect for #!r6rs.

Did these specifications arise out of scorn for tradition,
disregard of consequence, or mere carelessness?

Fortunately, implementations can just ignore the region of
effect specified for #!fold-case and #!no-fold-case, since
they are introduced by a non-binding appendix. They can
also ignore the suggestion that these directives act as
comments, leaving #!r6rs as the only wart of that kind.

That was the first observation. The second observation is
that I don't know quite what to do with #!no-fold-case when
presented with symbols such as

    Pi\x3bb;\x39b;\x3c3;

The problem is that the hexadecimal escapes could be playing
either of two roles:

    allow non-Ascii letters to be expressed in Ascii

    prevent the escaped character from being case-folded

The second role is problematic because the case-folding
of Unicode strings does not coincide with character-wise
case-folding. In general, therefore, it doesn't quite
make sense to perform case-folding on the un-escaped
characters but not to perform case-folding on the escaped
characters.

Even with #!no-fold-case, therefore, I am inclined not to
perform case-folding on symbols that were written using
hexadecimal escapes, and this is what I have implemented
in Larceny. This is certainly legal, since case folding
is forbidden by the draft R6RS proper, and I am just
interpreting the non-binding appendix.

But I'd like to hear from people who actually expect to
use #!no-fold-case in their programs, or have legacy
code to which they expect to add #!no-fold-case and
some hexadecimal escapes. What semantics would you
prefer?

Will
Received on Thu Jun 21 2007 - 06:59:30 UTC