References: <5.1.0.14.0.20050417104145.0247e860@pop3.web.de>
            <yfiekd9oea0.fsf@triumf.ca>
User-Agent: Gnus/5.11 (Gnus v5.11) Emacs/22.0.50 (gnu/linux)
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 8bit
Message-ID:  <85hdi5ej28.fsf@gnu.org>
Date:         Sun, 17 Apr 2005 13:31:11 +0200
Reply-To: Mailing list for the LaTeX3 project
              <LATEX-L@listserv.uni-heidelberg.de>
Sender: Mailing list for the LaTeX3 project
              <LATEX-L@listserv.uni-heidelberg.de>
From: David Kastrup <dak@GNU.ORG>
Subject: Re: \output trick; Re: lineno.sty question
To: LATEX-L@listserv.uni-heidelberg.de
In-Reply-To:  <yfiekd9oea0.fsf@triumf.ca> (Donald Arseneau's message of "Sun,
              17 Apr 2005 04:04:23 -0700")
Precedence: list
Status: R

Donald Arseneau <asnd@TRIUMF.CA> writes:

> Uwe Lück <uwe.lueck@WEB.DE> writes:
>
>> This is due to a mistake in revtex's way of changing the output
>> routine.
>
> I agree.
>
>> After a \newtoks\output,
>
> And, for the output routine, I think that itself is the mistake!  I
> opine that for \output one should "bite the bullet" and do
> \output=\expandafter{\expandafter\mymacro\the\output}
>
> Hoping to silently hook in, without knowledge of what else has
> modified the output routine, and without making your own
> modification obvious, is insane.

The problem is that you sometimes need to call the old behavior, and
sometimes not.  In particular, if you have inserted a special penalty
yourself, then you will deal with it yourself, not passing anything on
and not actually calling the original output routine.  There really is
no sane reason to expose all this to completely unrelated packages and
formats and hope that they will not break in the process.

Using a clean separation between changed and called hook makes it
possible to give strictly coherent results which make changes strictly
confined to the package in question.

As you can see from the previous example I posted in the list, it does
not even seem that latex.ltx and amsmath.sty which are released
together get it right, as amsmath.sty fiddles inconsistently with both
\frozen@everymath and \frozen@everydisplay as well as \everymath and
\everydisplay.

So this clearly is already a mistake when just two packages in the
same code base are involved, and the complexity explodes exponentially
when more packages get pulled into the play.

--
David Kastrup, Kriemhildstr. 15, 44793 Bochum