Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!sq!ian
From: i...@sq.UUCP
Newsgroups: comp.text
Subject: The rumors of my death are greatly exaggerated
Message-ID: <1987Feb6.161011.10075@sq.uucp>
Date: Fri, 6-Feb-87 16:10:11 EST
Article-I.D.: sq.1987Feb6.161011.10075
Posted: Fri Feb  6 16:10:11 1987
Date-Received: Tue, 10-Feb-87 18:57:43 EST
References: <505@geac.UUCP> <112@tg.UUCP> <513@geac.UUCP>
Reply-To: i...@sq.UUCP (Ian F. Darwin)
Organization: SoftQuad Inc., Toronto
Lines: 102
Keywords: troff SoftQuad
Checksum: 41599
Summary: troff lives

Some contributors to the discussions in this newsgroup seem to convey
the impression that troff is dying of old age.  Paraphrasing Mark
Twain, the rumors of troff's death are greatly exaggerated.

Troff is alive and well and living in the UNIX environment.  Several
sites are doing significant troff development. Brian Kernighan at
Bell Labs Research continues to come up with pre-processors, such as
``chem'', a program to insert chemical drawings with elements such as
benzene rings into place in a text, just as eqn puts equations into
place in a text.  Berkeley has added several extensions to troff.
Several other universities are developing or modifying troff.  A dozen
or so companies offer ditroff drivers for popular laser printers and
typesetters (Image Network sells one for the LN03).  Troff has even
been ported to MS-DOS by Elan Computer Group.  And then there is SoftQuad.

SoftQuad set out to make production quality typesetting software for
UNIX.  ``Production quality typesetting'' in this regard means
the type of type (so to speak) that would be acceptable in an old-style
craft printing house.  There is quite literally nothing available that
is up to these typographer's standards. Not Scribe. Not TeX. Not AT&T's
ditroff. Not Mac-Anything. Not PageMaker.  All these are fine products
in their own field, have been used to typeset many papers and theses
and a small mountain of books.  But none of them will stand up in a job
printing shop with high standards and commercial pressures.

The founders of SoftQuad considered several packages as possible
starting points (including Scribe, TeX, and ditroff), and settled (in
1984) on ditroff.  They set out to make it usable in commercial
typography.  Two of the most important problems were the lack of a
hyphenation dictionary facility and the lack of kerning.

Avi Naiman (who used to work with us) mentioned a paper we did at
USENIX in January, 1985 describing some of the work. By that time the
company had in place a ditroff with proper hyphenation and with contour
kerning. They had also totally rewritten the ditroff intermediate
language to be easier to pass through UNIX software tools like awk. In
the ditroff intermediate language, for example, the input string
``hello'' might appear as ``ch 19e17l10l10on423 0'' while in ours it
appears as ``hello''.  Both ditroff and sqtroff intermediate languages
contain only ascii characters (unlike SCRIBE and DVI files which contain
``unprintable'' characters: h\220G\034ello). Production publishing shops
need to be able to extract (from a trial formatting run) arbitrary
information beyond what most formatters provide. Having the
intermediate language in a readable form (that admittedly owes some of
its form to PostScript) allows simple shell scripts with programs like
grep or awk to process the intermediate file to extract this information.

We also have made life easier for those who must ``meddle in the
affairs of troff''. We've added a comprehensive trace facility
that lets you see what a troff macro really does. We allow
names longer than two characters, so you no longer need to
write macros with names that look like bird droppings.
And we've cleaned up the code - a lot.

For more technical details on what we've done, see the USENIX
paper (Dallas, January 1985, page 165) or contact us.
Note that the address shown in that paper has changed.

The remainder of this notice deals with licensing and commercial
issues; if you are an anti-commercial techie, this may hurt your ears;
so reach for your interrupt key now!

Despite the urgings of our marketing people, we try very hard not to
ship a product until it is ready. Thus we did not ship any formal
released products until June, 1986, although a few friends and
associates received our software before that.  Since then, a number of
books has been produced at sites using our software, and many papers
and theses typeset (if you were at the UniForum trade show, you might
have seen the SCO Third Party Software catalog - it was typeset with
SoftQuad troff). Despite the enhancements to troff and the changes to
the intermediate language, most old troff files will work with our product.

As was pointed out, our LN03 driver has not been shipped yet.
Admittedly our sales staff did accept a purchase order for it several
months ago (the customer, GEAC, was told before ordering that no
specific delivery date had been set for the LN03).  We are running the
LN03 software in-house, but we aren't yet happy with the fonts, so we
can't ship it as production software. When we do ship a preliminary
version of something, we try to so identify it.  Our customers have a
right to expect prompt shipment of orders that we accept, but because
we currently ship only binary products, customers also have a right to
expect it to work ``out of the box''. When it is ready, we will ship it.

We are a source licensee of AT&T's Documentor's WorkBench product.  But
we do not have a monopoly on distribution of DWB.  If your printing needs
don't include top-quality typography, and want to write your own
drivers and support them and put big fixes into troff yourself, you
should certainly be able to order the unsupported DWB Release 1 or 2
source from AT&T.  If you want a quality typesetting product that is
derived from DWB and is supported, we hope you will order from us.

Ian Darwin,
Director of R&D,
SoftQuad, Inc.
720 Spadina Avenue,
Toronto, Ontario
Canada M5S 2T9
+1 416 963-8337

Opinions expressed herein *are* those of SoftQuad, Inc., and the
author.  The product names used herein are trademarks of their
respective manufacturers and/or vendors.

Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!utgpu!utfyzx!sq!ian
From: i...@sq.UUCP
Newsgroups: comp.text
Subject: SoftQuad history: Troff vs TeX
Message-ID: <1987Feb17.154939.3148@sq.uucp>
Date: Tue, 17-Feb-87 15:49:39 EST
Article-I.D.: sq.1987Feb17.154939.3148
Posted: Tue Feb 17 15:49:39 1987
Date-Received: Thu, 19-Feb-87 00:36:52 EST
References: <112@tg.UUCP> <513@geac.UUCP> <1987Feb6.161011.10075@sq.uucp>
Organization: SoftQuad Inc., Toronto
Lines: 67
Keywords: troff SoftQuad
Checksum: 62177

Several people have asked me, in response to my previous posting on
SoftQuad Troff, for a detailed answer on why SoftQuad originally chose
troff over TeX.

``Small is beautiful'' (although I use the term loosely when talking
about troff internals).  Troff proper is smaller (because it is
intentionally designed to delegate gruntwork to pre- and
post-processors).  Because it is smaller, there are fewer design
decisions imbedded in the code.

Troff's operating mode conforms to UNIX program conventions.  It does
not ask you for filenames, nor does it put its output into magic
filenames (InputFileName.dvi, etc).  As a result, it works well in
pipeplines, resulting in an abundance of pre-processors.  These are
relatively easy to write (as opposed, say, to trying to add the
corresponding functionality to TeX).  See Bentley, ``Programming
Pearls: Little Languages'', CACM, August 1986, V29, N8, page 711, for
detailed discussion - I have used up enough net bandwidth already, and
Bentley gets the ideas across better than I might.

The founders of SoftQuad, like the troff software, had a long history
of experience with phototypesetters, which are the machines then used
by most commercial printers.  By contrast, TeX appeared to be primarily
oriented towards raster printers (300 dpi laser printers in particular)
for which it could download CM fonts.  At that time, the CM fonts were,
uh, not as well-formed as they are today.  We needed to get good
quality letter forms; TeX did not produce them in the quality needed,
and appeared to be wired in to use of downloading CM fonts.
While TeX did have kerning, that too depended on the raster fonts; to
drive phototypesetting machines, we needed the ability not only to 
set type using manufacturer's fonts without downloading CM, but to kern
those existing fonts. To my *limited* understanding, this could
only be done by building TFM files (or whatever) for those fonts. It
seemed at least as easy to slightly modify ditroff's font.OUT files as
it would have been to build TFM files from scratch (not using Metafont79).

Also, TeX was written in a slightly-non-standard dialect of Pascal,
rather than in WEB, so there were issues of portability to consider
(recall that the early UNIX TeX tapes - in circulation until quite
recently - could only go to UNIX-source-licensed sites because they
contained modifications to the Pascal compiler needed to make that
version of TeX compile).  Troff by contrast was known to run on many
flavors of UNIX, and was written in fairly portable C.  Of course
today's TeX runs on about the same range of systems as Troff does, so
portability is no longer an issue.

One desirable property of ditroff's intermediate language is that it is
amenable to ``postprocessors'' other than those for specific devices.
For example, you might want to use grep, awk, etc., on the intermediate
language.  Clearly one could hack up TeX to print readable DVI files,
but then it wouldn't be TeX anymore.  We hacked up ditroff to print
more-readable intermediate files, and it isn't ditroff anymore - hence
that product is called SQtroff.  We have backwards compatibility with
all previous versions of troff (including a filter in our standard
software that generates ditroff-compatible intermediate language from
our format, so a user with existing ditroff postprocessors can continue
to use them with our SQtroff).

The decision to use ditroff instead of TeX was made in 1983 (before I
even thought of coming to work for SoftQuad).  Some of the reasons are
no longer applicable, but most of them are.  We would probably make the
same decision today.

Ian F. Darwin,
Director, Research & Development,
SoftQuad, Inc., Toronto, Canada
i...@sq.com, i...@sq.uucp, utzoo!sq!ian