[tex-live] last weekend changes

[Karl Berry]

I know for a fact there are dozens of TUGboat articles which use
\ifx\pdfoutput\undefined -- that is what I recommended in my TUB
templates!  Not that TUGboat articles are a big deal per se, we're not
likely to be rerunning them, but it's just a sample.  Virtually every
user who ever tried to adapt their document to coexist peacefully with
PDF and DVI simultaneously will lose here.

People (including me!) were/are trying to do the right thing.  Sure, of
course in theory pdftex could be used for DVI output, but in practice it
never was, and in fact most people probably did not even know it was
possible.  To my mind, It's not fair to use an unused and little-known
feature as an excuse for breaking a standard idiom in longstanding TeX
documents that have worked ever since pdftex has existed.  I still do
not think it is right to call it "broken".  It is the cleanest, simplest
test and I think we are gratuitiously causing grief to many people if we
break it.

As for keeping the new primitives accessible, the clean answer is Hans'
\pdfstate.  The ugly answer is \let\saveprimitivepdfoutput=\pdfoutput
\let\pdfoutput=\undefined, with another command/package to switch
\pdffoo back to \saveprimitivepdffoo.  I am certain that the number of
documents using pdftex primitives in DVI mode is absolutely minuscule
compared to the number of documents assuming \pdftex=undefined for DVI
output.

I can see that, five years from now, maybe even two years from now, if
the pdftex-in-dvi-mode primitives explode in popularity, it could be
pretty annoying to not have them available by default.  So let's revisit
it then.  It's not the case here in 2004.  Meanwhile, we can at least
warn people it may change, use ifpdf.sty, blah blah blah.

I think it would be a terrible, terrible mistake to make a release that
we know will break documents in this way.