Re: Ebcdic rule

[Paul Eggert]

> From: Harlan Stenn <address@hidden>
> Date: Wed, 03 Oct 2001 15:50:43 -0400
> 
> Somebody can write an AC_CHARSET_{ASCII,EBCDIC} macro that embodies that
> test.
> 
> It gets carried around inside autoconf so improvements/changes can be easily
> used, and then it gets dropped in to config.h

I think this is an overly optimistic view of Autoconf's goal.
Autoconf is not a tool for centralizing and organizing all information
about porting programs.  That would be far too ambitious, as it would
include subjects like internationalization, localization, and
arithmetic that have never really been part of Autoconf's bailiwick.
These subjects would take a lot of effort for Autoconf to support
properly, and they are already addressed adequately enough by other
tools and techniques.

Autoconf is merely a tool for configuring source code for UNIX-like
systems.  That problem is hard enough.

> (does '-DCHARSET_EBCDIC=('A'==0xXY) work as a CFLAG item if config.h
> isn't being used?).

I wouldn't write it that way.  I would simply put something like this:

#define CHARSET_LOOKS_LIKE_EBCDIC ('A' == 0xXY)

into one of my application's C headers, as it's better to keep CFLAGS
short.  And once I've written it that way, I don't need Autoconf at
all.  So it's not a problem that really needs Autoconf's attention.

If we really wanted to add EBCDIC support, we should add support for
all the EBCDIC variants out there, as well as the ASCII variants, the
ISO 646 variants, the ISO 10646 versions, the ISO 2022 variants, etc.,
etc.  (For example, have you heard about UTF-EBCDIC?  No, it's not a
joke.  See <http://www.unicode.org/unicode/reports/tr16/>.)

All this stuff would be nice to have, but libiconv is a better place
for it than Autoconf is.