[r6rs-discuss] [Formal] Scheme should not be changed to be case sensitive.

From: Aubrey Jaffer <agj>
Date: Wed Nov 22 17:06:28 2006

 | From: Michael Sperber <sperber_at_informatik.uni-tuebingen.de>
 | Date: Wed, 22 Nov 2006 21:38:50 +0100
 |
 | The problem with that whole mess is that there's not really a
 | formulated rationale for the change. Basically every rational
 | argument for one side or another has an equivalent on the other
 | side ...
 | Anyway, trying to get out of my rant, I see two possible clear-cut
 | rationales for one side or another:
 |
 | - A majority of the community would like to see case sensitivity.
 | There's anecdotal, but no empirical evidence to that effect.
 |
 | - The change breaks backwards compatibility, and would break a lot of
 | code. There's anecdotal, but no empirical evidence how bad that
 | problem is.

In December 2005, I converted SLIB (Scheme library) and JACAL (a
symbolic algebra application) to run in case-sensitive Scheme
implementations.

For SLIB, this involved changes to about 6 .scm files out of 150+, and
most of the 18 .init files. I changed configuration (.init) files for
all implementations, not just the ones currently case-sensitive, to
save myself the trouble in the future. Only Scheme code was
downcased, not comments. It was about 2.5.hr of work.

Changing JACAL required more effort. Most of the source files were
involved; and more than a few identifiers and symbols had to be
dowcased. The "English.scm" file having templates for 2-dimensional
output have (what were uppercase) symbols mixed with strings, some of
which are case-sensitive. I estimate it was more than 4.hr of work.

Had I left uppercase and mixed-case symbols in the code, then new case
mismatches introduced into the code would only be detected by a
case-sensitive implementation. Although I do regression testing with
several implementations (including case-sensitive ones), those tests
do not have 100% code coverage. Better coverage would require using a
case-sensitive implementation for my day-to-day work -- no, thank you!
Received on Wed Nov 22 2006 - 17:03:56 UTC