[tex-live] last weekend changes

[Thomas Esser]

> And I don't agree with the characterization of them as "broken", if
> that's what you're saying was broken.  Testing \ifx\pdfoutput\undefined
> *was* (and remains!) an absolutely standard practice.

pdftex, from the beginning, was able to produce dvi output. The pure
existence of the \pdfoutput primitive does not necessarily mean that
people want to produce pdf output. That's why I consider any code "broken"
that assumes pdf output if the primitive \pdfoutput exists.

People are actually already using pdftex features for producing dvi,
e.g. character protruding. So, the problem which we are discussing is
nothing theoretical and already existing. That's why we already have
good code which is aware of the fact that pdftex is running in dvi mode
(geometry, ifpdf, graphics.cfg, color.cfg, ...).

In fact, the whole teTeX distribution contains only two "broken" files:
changebar.sty and ltnews.cls. The authors are notified, so hope is that
these wil soon be fixed. User's code is a different matter, but we can
give them a simple one-liner solution to get their document compile again.
And, if they really want, they can still go back to good-old tex by a
simple change in theit fmtutil configuration.

> What is wrong with it?  It provides transparent compatibility.  I don't
> understand why you think it is bad.

Why are we making that move to pdfetex? To save a few bytes? I don't
think so. My impression was that the main idea was to be able to add
new primitives (e.g. \pdftexbanner). We can do this with pdftex, but we
can't do this with TeX.

It might be useful for someone to parse the expansion of \pdftexbanner
for something. If we remove/hide these primitives, this will be lost.

> I feel it is crucial that existing documents continue to work
> *unchanged*, one way or another.  There are zillions upon zillions of them.

I have made the switch to pdfetex 6 days ago when I released
teTeX-2.96.0.20040321 (a beta release announced at tetex-pretest list
and available for download from CTAN); so far, only a single user has
reported a problem with this.

Thomas