DVI, PDF and TAC

DVI, PDF and TAC

[Paul Taylor]

Mike Barr reported
	that Adobe Acrobat 8 tacitly suppresses all ligature glyphs of the
	fi, fl, ff, ffi, and ffl sort and displays blanks in their place
and then that
	I have to admit that I never tested it, just copied the complaint
	from texhax
but nevertheless this presumably gives us some idea why,
	at TAC, we still consider the dvi to be the official format.

However, as I shall demonstrate, the rest of the world nowadays regards
PDF as the standard format in which to publish technical documents.

DVI (the normal output from LaTeX) was based on the 1950s Monotype
typesetting system, and puts characters from various fonts at given
positions on the page, but cannot rotate them, and has no graphics
capability.  The fonts also have to be supplied separately.   On the
other hand, it has the virtues of being a compact and simple format
that future digital archeologists would have no difficulty in
deciphering.

Adobe's PDF and PostScript have general graphics capability.

By insisting on DVI, "Theory and Applications of Categories" severely
limits the ways in which authors can express their mathematical ideas.
But its restrictions go further than this:  the use of ANY macro package
other than those by Mike Barr and Kris Rose is forbidden!



Does anyone know of another journal that publishes primarily in DVI?

One candidate might be the journal of the TeX Users' Group,
	tug.org/TUGboat
but even that uses PDF.   When I enquired about this, Karl Berry (who
also wrote the Web2C Unix implementation of TeX) replied that
	DVI files are not self-contained, so they simply don't work
	for online archives.

Turning to other respositories, arXiv.org generates papers in various
formats on-the-fly.  On its page for each paper, eg
	arxiv.org/abs/math/0512110
it offers PostScript and PDF, with other options (including DVI) only
being available via another link.

On my own web site,
	www.PaulTaylor.EU
I also offer my papers in various formats, albeit statically, with DVI
listed first.  However, the downloads in September and October were
	DVI: 210     PDF: 1704     PS.GZ: 75     BKLT.GZ: 52
This includes 504 PDF and 69 other downloads of "Proofs and Types", but
excludes the lecture slides and scanned manuscripts that I only offer
in PDF.  So, 85% of my readers choose PDF,  even though I also offer 
DVI.

Looking at my colleagues' web pages, my impression is that most of them
ONLY offer their papers in PDF nowadays.   I can't give authoritative
figures for this, as Hypatia has been dead for seven years.  Papers that
people send me as attachments to read or referee generally come in PDF 
too.

Even TAC has published a number of papers that it only offers in PS or
PDF. Maybe Bob Rosebrugh could tell us how many downloads there have
been in the various formats from the main TAC web site at MTA.



Turning to software, although PDF legally belongs to Adobe, I really
don't care about bugs in their programs, as I never use them.   For me
and anyone else who writes in LaTeX,  PDF is de facto the language that
is output by pdflatex and input by xpdf and ghostview (gv).   It is a
well documented open format, as Peter Selinger and Andrej Bauer have
pointed out, citing
	www.adobe.com/devnet/pdf/pdf_reference.html
and	en.wikipedia.org/wiki/Portable_Document_Format
See	www.PaulTaylor.EU/technote.html
for brief explanations and links to further information.

When using my Mac OSX laptop, I use the program (Preview) that came with
it to read PDF files.   This also understands several other file 
formats,
but not DVI.

When I run "latex" in the current tetex distribution of the TeX system,
it actually invokes pdftex (the version of TeX that was modified by
Jirı Zlatuska to generate PDF instead of DVI) with \pdfoutput=0.

Google indexes various other formats of web pages besides HTML,
	www.google.com/help/faq_filetypes.html
including PDF and several versions of MicroS**t W*rd, but not DVI.



Returning to TAC,  I quote from its  "Author Information" at
	tac.mta.ca/tac/authinfo.html

	An article must be submitted as a single source file.
	All macros must be included at the beginning of the file.
	Any macro that is not actually used should be deleted from
	the source file.

	The only exception is diagram macro packages. The currently
	acceptable diagram macro packages are those authored by Barr
	and Rose and Moore(Xy-pic). Recall however that authors are
	expected to provide source code which produces usable .dvi
	from these packages (see Note 4 above). Do not rely on .ps
	options. The author is responsible to ensure that the current
	version of a macro package has been used.

Leaving the DVI/PS/PDF issue aside, why is it necessary for articles
to be submitted as single source files?   According to Mike Barr, in
an email to me dated 10 July 2002,
	our rule is, no inclusions, with the following exceptions:
	diagram packages, including yours, and packages that are part of
	standard distributions.  This is so that we can store each paper
	in a single file, without a growing (and essentially unidentifiable)
	directory of inclusions.

Now it has been customary for as long as I have had anything to do with
computers that software (by which I mean both programs and papers in 
this
context) is developed in modular parts, divided into several files but
collected in a directory or folder.  In particular,  programming with
macros should be separate from writing text about mathematics.  Indeed,
the design of LaTeX2e presupposes this in its facility for passing
options to macro packages via
	\usepackage[options]{package}
Has Mike Barr not heard of sub-directories, or of tar-archives?
Why does he create so much inconvenience for TAC authors for such
a trivial benefit to himself?

Notice, in particular, that my diagrams package was approved for use
in TAC in 2002, but is now forbidden.   Is this perhaps because Mike
disgrees with me over DVI?

The design decision to use PostScript inclusions to rotate diagonal
arrows in my package was made in 1992,  in consultation with users.
Neither Mike nor Bob nor anyone else at the time argued against that
decision, whilst several people said it was a good idea.  The package
web page,
	www.PaulTaylor.EU/diagrams
explains the background to this decision,  and also how to make half-
decent diagrams using the UglyObsolete pre-1992 code if you REALLY
need to use pure DVI.    I would strongly request that anyone who uses
my package, or  who wishes to reply to my comments here, should READ
this web page first.	

TAC began two or three years AFTER this decision was made,  but Mike
and Bob did not discuss their pure DVI policy with me.   If they had
done so, we might have been able to lobby the maintainers of DVI
programs (such as Paul Vojta of XDVI) to add support for rotation,
or they might have persuaded me to improve the old DVI code in my
package.  The time for doing either of these things has now long past.

In his email of 23 January 1990 to me and 24 other people that resulted
in the establishment of the "categories" forum and subsequently the TAC
journal,  Bob Rosebrugh said
	it seems clear (maybe only to me) that a TeX-based journal is
	a starting point.  My guess is that LaTeX together with some
	version of Mike's macros should be the starting standard.
It would appear that they are still trying to impose this standard.

There are plenty of people who regard MY package as the standard.  For
example, I recently heard from a blind mathematician who "draws" 
diagrams
by dictating the input language of my package to his wife,
	lalitalarking.blogspot.com/2007/09/great-dictator.html


In conclusion,  I call upon Mike, Bob and the Editorial and Advisory
Committees of TAC to come out of the 1980s, and support the production 
of
papers using the modern typographical software that other journals use.

Paul Taylor


PS.  Whichever /"Mr" Paul taylor/ it was who wrote a joint paper with
Phil Scott on locally cartesian closed categories is no doubt greatly
honored to find that "Monsieur" Jean benabou regards him as a special
case.

However, M. Benabou will perhaps be disappointed that he is neither the
author of nor cited "alongside [the] lot of complete nonsense" to which
I alluded.  This may be found at
	computing.unn.ac.uk/staff/cgnr1/liege_quantum03.pdf
it cites Baez, Barr, Bell, Birkhoff, Bishop, Bridges, Dirac, Dummett,
Ehresmann, Einstein, Freyd, Heyting, Hilbert, Johnstone, Leibniz,
Mac Lane, Peirce, Scedrov, Troelstra, Turing, Wells and me.

Re: DVI, PDF and TAC

[Bob Rosebrugh]

With regrets that this response is not briefer...

On Thu, 1 Nov 2007, Paul Taylor wrote:

> Does anyone know of another journal that publishes primarily in DVI?

Many electronic journals in mathematics post dvi files. Like TAC, most
of these post, and also archive, all of dvi, ps and pdf.

TAC's policy on dvi has evolved since 1995. That policy will continue to
change, no doubt at a slower rate than some would wish. None of us knows
what the digital world will look like in 10 years, but careful choices
made for TAC over a dozen years ago have been validated.

> Maybe Bob Rosebrugh could tell us how many downloads there have
> been in the various formats from the main TAC web site at MTA.

Inevitably nowadays, most of the web traffic on sites like TAC's is for
caching, so such figures for any TAC site mean nothing. If Paul's counts
record human usage, then what is surprising is how many of the downloads
were *not* pdf.

...

> Why does he create so much inconvenience for TAC authors for such
> a trivial benefit to himself?

This request (for a single source file) is seen as, at worst, a trivial
inconvenience by most authors, and it simplifies the lives of editors who
volunteer their time and knowledge.

Note that TAC's submission requirements for authors describe what we would
like to see. Sometimes we don't. We are grateful on the many occasions
when authors comply. In practice, TAC editors are flexible and are working
with authors who gladly cooperate in publishing a visually pleasing
article. We much prefer diagrams based on xypic, but a glance at recent
numbers shows that leeway is available.

...

> It would appear that they are still trying to impose this standard.

Paul has made a heroic search for a conspiracy, but, alas, has not found
one. I don't particularly remember writing the sentence he quotes from
early 1990, and hadn't heard of his diagram package back then. My 1990
suggestion was not motivated by an intention to exclude him 15 years
later. The misperception of malice is regretted.

Bob Rosebrugh