Re: changequote

[Ralf Wildenhues]

Hello Bruno,

* Bruno Haible wrote on Tue, Feb 13, 2007 at 01:57:11PM CET:
> > 
> > Sigh.  I wish that if you had to keep your own ways of doing things this
> > way, rather than using [[ ]] or quadrigraphs, then at least you'd be
> > consequent enough to also fight for a change of the Autoconf manual to
> > reflect it, rather than keep diverging from it.

> My arguments were not strong enough, so I applied the patch of the [[ ]]
> proponents. The result was that a little more than one year later, a
> quoting bug appeared in gettext.m4:
>   http://lists.gnu.org/archive/html/bug-gnulib/2006-11/msg00260.html

So now I posted a patch for another bug in a gettext macro file, which
is more than 5 years old, and thus predates your "lesson learned".  Can
I now use this bug you did not prevent with your `changequote' strategy
as a valid counter argument?  I don't think so.  Both were merely bugs,
and bugs happen with both styles.

> But I doubt that
> 1 bug that I introduced will be enough to convince people.

Yes, because one bug is statistically insignificant.  But I digress:
it's not statistics that counts, but whether users play well with
Autoconf (and where Autoconf may eventually go) or not.  And whether
you're going to complain if Autoconf eventually goes out and actually
changes in this direction.

FWIW, I think the other styles that are possible (the maximal quoting
one) fit your necessary criteria for bug prevention you outlined as
well.

> > <http://www.gnu.org/software/autoconf/manual/html_node/Changequote-is-Evil.html>
> 
> I think this section in the autoconf manual was written because Akim wanted
> to make deep changes to the way autoconf macros are processed, and wanted to
> warn people ahead of time. It's a different motivation than the worry
> "what leads to less bugs in the long run".

Yes.  The question is now: which part of the warning was hard to
understand?  Just because Akim isn't hacking on Autoconf so much
these days, doesn't mean not going by it is safe in the long run.

Don't get me wrong.  I do understand your motivations, and I would
like your concerns to be addressed.  But also I think consistency
among packages viewed from the outside as "belonging together" is
a big benefit.

Cheers,
Ralf