Nils-Peter Nelson, sqtroff, and groff history

[G. Branden Robinson]

Some people might find the following messages, first from Nils-Peter
Nelson and then Liam R. E. Quin, of interest.  They're posts from 1996
to the comp.text USENET group.  (Je me souviens, DejaNews!)

I added some gaps between paragraphs.

---snip---
>Where does one get the real troff and DWB system these days?

This is a relation of what happened to DWB, not a justification.

<history>
DWB is the collective name for over 80 Unix commands that
generally relate to troff. DWB was under continuous development
in Bell Labs from the 1970's through 1994, first in Bell Labs Research,
then in Unix System Laboratories, then in my own group in Bell Labs.
USL abandoned DWB when it abandoned tools in general and focused
on the basic operating system (1987). They signed an agreement
with SoftQuad, which was sensible for external folks but left
AT&T documenters in the position of buying AT&T software from another
company at a greatly increased price. In addition, Brian Kernighan
of Research continued to improve troff, but SoftQuad did not
have the rights to his improvements. I picked up the ball to
reduce cost for internal AT&T folks and preserve compatibility
with Research. Our own funding was from license fees.

<flame>In 1994, our biggest internal customer took our software,
installed it on all their machines, then declined to pay.</flame>

Because this happened in the second half of the year,
and I was supposed to make revenue match cost, I had no choice but
to disband the group and get the cost down to zero.

We all found other jobs in AT&T, though Jaap Akkerhuis elected
to return to his native Netherlands.

We left the source code in the custody of another organization,
which has since disbanded for other reasons. There is one person
assigned part-time, in case interest re-awakens.</history>

There are currently at least three common versions of troff,
all stemming from the original Bell Labs version by Joe Ossanna:
an ancient variation from Bill Joy that, remarkably, still gets
shipped by Sun ("no /dev/cat"); the SoftQuad version, based on
USL's DWB 2.0; the DWB3.x versions from my group, in cooperation
with Brian Kernighan. This version is widespread
inside Bell Labs, Lucent and AT&T and includes
many PostScript printer support commands and X Window stuff
like the Picasso drawing program. To answer the orginal question,
you can rotate tables and Picasso can rotate text, but not grap.
Since you could choose to pipe grap into picasso (instead of pic)
it's a trivial enhancement.

I'd claim DWB 3.4 is "the real troff" since it is in synch
with the original Research effort.

To get "the real troff":
AT&Ters can get free binary versions (but not source!)
from babel.ho.lucent.com,

Everyone else can buy source (but not binary!) from Lucent
Software Solutions, 1-800-462-8146.

There is no support group anywhere. There is no one to complain
to if this explanation distresses you. There is no one who
understands it better than I do, if you are unsatisfied with
my explanation. I seldom read this newsgroup
so please send follow-ups to n...@research.att.com
(lucent.com works, too).

---snip---

> USL abandoned DWB when it abandoned tools in general and focused
> on the basic operating system (1987). They signed an agreement
> with SoftQuad, which was sensible for external folks but left
> AT&T documenters in the position of buying AT&T software from another
> company at a greatly increased price. In addition, Brian Kernighan
> of Research continued to improve troff, but SoftQuad did not
> have the rights to his improvements.

I can add a little to this from my own perspective. I should say
at the outset that this is my own opinion, and not SoftQuad policy or
anything official.

In 1987, I worked for the UK dealer of SoftQuad, not for SoftQuad.
As far as I could tell, the people at SoftQuad didn't realise when they
bought the DWB rights that they were not also getting rights to
Brian Kernighan's work. They certainly didn't realise that the
DWB distribution and the ditroff distribution that were both then
available were different from each other! I don't think it's a question
of blame or stupidity or malice so much as of miscommunication somewhere.
After all, it had taken me several months to understand this, calling
offices in Europe and the United States, back when I worked at
Warwick University, which was trying to see if it wanted DWB.

At any rate, SoftQuad continued to develop sqtroff for several years.
The SoftQuad version has kerning, long character names, long macro,
register, string and special character names, and an ASCII intermediate
format that's moderately readable and amenable to awk/perl/sed.

I think it was in 1988 that I discovered that James Clark was working
on groff, and pressured SoftQuad, who very kindly sent him a manual
for the then latest sqtroff. As a result, groff is fairly compatible
with sqtroff 2.9.1, the then shipping version.

Today, you can use groff on linux (which was what the original poster
asked for, I think), and sqtroff on most commercial Unix platforms.

I don't believe we could sell you source for sqtroff, because of
the complexities of whatever remains of our arrangement with AT&T.

We might be able to do a linux port, but for similar reasons, we
probably couldn't give it away. If you're interested in sqtroff for
linux, and can pay a couple of hundred bucks for it (is that plausible?),
please let me know. If enough people reply, maybe we can do a port.
Unfortunately, it's either a fixed royalty, or there's a minimum,
and when the cost of duplicating the manual is added in, and shipping,
I don't think it can be much less than that. If you do reply, please
let me know what version of Linix (ELF or a.out?) you're using, what
kind of support (if any) you'd want/need, and how we'd get it to you
(tape? CD-ROM? ftp?). We do have a version for ESIX, which might
well run under Linux, come to think of it.

Nils-Peter, I don't know if this is good or bad, but as far as I can
tell, SoftQuad never actually got any significant income from the
AT&T DWB deal. Certainly it wasn't anything like as successful as
HoTMetaL PRO has been for us :-)

When I was at Unixsys UK (the dealer for sqtroff), we had quite a few
customers for sqtroff. It was easier for us to sell in the UK than
in North America, it turns out -- I think partly because the cost of
visiting people interested in buying it was much lower, and partly
because of different ways of looking at software. I remember that
at one point, without doing any advertising, sqtroff was selling at
the rate of maybe ten thousand (pounds sterling) worth a month, which
for a company of 8 or 10 people wasn't a bad side-line...

Our (SQ's) income from sqtroff is negligible at this point as far as I know.
We do still get the occasional support question, but they're pretty rare.
Or they seem that way, 'cos there are more people doing HoTMetaL PRO
and Author/Editor and Panorama PRO support... There was someone asking
something about troff fonts earlier this year, though.

> the DWB3.x versions from my group, in cooperation with Brian Kernighan.
> This version is widespread inside Bell Labs, Lucent and AT&T and
> includes many PostScript printer support commands and X Windows stuff
> like the Picasso drawing program.

Hmmm.... I wish we/SQ had been able to take advantage of the improvements.
On the other hand, there are so many other improvements in sqtroff that
were done independently that it's hard to say which version is `best' --
it probably depends on what you want to do with it.
We don't distribute chem or ideal, and I've never even seen Picasso.
On the other hand, our manual won an award :-)

> I'd claim DWB 3.4 is "the real troff" since it is in synch
> with the original Research effort.

Naaah, the V6 troff was the real one :-) :-)

Someone else asked about Elan troff. I can't give an unbiased answer,
of course, but this might help you evaluate the various versions,
if you/re interested.

At Unixsys, we sold at least 3 different versions of troff.

One was sqtroff, and this was best for people who had PostScript printers
and/or wanted to develop their own macros. It used (back then) to
have some problems with -me and -ms, because it derived from the AT&T
version, where -mm was used most. (Those problems are all since
sorted out as far as I know!).
Our web page doesn't really say much about troff, I notice.

Another was Elan troff. This had a slicker install, and included
print filters for the old SysV spooler and the BSD one, and knew
about the HP font cartridges. Well, I said it was a long time ago!
But eroff had very little support for macro writers -- no debugging
trace, for example, and only 2-character names. The documtation
was (then) aimed at people who already knew troff and/or wanted to
use existing macro packages, not at people writing macros.
Eroff had a -x option to show if you were using enhancements not
found in ``bok standard'' troff -- just like my troff patches on
the 1986 or 1987 EUUG tape :-)

On balance, we used to sell sqtroff to people developing macros,
and elan only to people who were using it as part of another
application, and needed backwards compatibility and the lpd stuff
more than anything else. These days the lpd filter is a non-issue,
as most Unix system can print to a PostScript or HP printer out of
the box, and if nothing had been done to either package since,
I'd go with the SoftQuad one. But as I said, I work for SoftQuad.

I've been using sqtroff recently together with David Megginson's
NSGMLS.pm to typeset a book from SGML.

Elan does have a web page for eroff. However, it talks about their
SunView previewer, and mentions it's available for the Apoolo Domain and
the 3B2, and various other computers you can't buy any more, so maybe
they've not been doing much with the shipping version of eroff for the
last few years, just like us.
But I am guessing. Anyone from Elan listening?

Lee
---snip---

Regards,
Branden

signature.asc
Description: PGP signature