Subject: Re: UNIX manpage history: CTSS RUNOFF
From: Jerry Saltzer 
Date: 22/10/2011 04:10
To: Kristaps Dzonsons

Kristaps,

Recording this historical stuff accurately is getting harder with each
passing year, as the principals from the 1960s have increasing trouble
recalling details, or worse, die off.

To get to your questions, TYPSET and RUNOFF were designed, implemented,
and made public at MIT in the fall of 1964.  In addition to the
documents you found, there are some other things on-line that you might
find helpful.  One is on my publications page, at

     http://web.mit.edu/Saltzer/www/publications/pubs.html

Look for "section AH.9.01" of the CTSS programmer's guide, which at the
bottom of its first page lists some of the prior art that was
incorporated in TYPSET and RUNOFF.  The dot-at-line-beginning convention
was inherited from the Memo, Modify, and Ditto command set of CTSS,
implemented a year or two earlier.  I don't know if that convention was
original with the designers of that command set or if that idea came
from elsewhere.  The user's guide to Memo, Modify, and Ditto is in the
first edition of the CTSS programmer's guide.  You may be able to find
that manual on-line at bitsavers.org

In the description of AH.9.01 is a link to my 1964 memorandum
CC-244/MAC-M-193 describing RUNOFF.  The page containing that memo also
has a link to my Programming Staff Note PSN-40, which is of interest
only if you want to trace the development of specific RUNOFF commands.

Starting in 1965 and 1966, several Bell Labs staff members, including
Bob Morris, Doug McIlroy, Ken Thompson, and Dennis Ritchie were working
on the Multics project, which was implemented using CTSS as a staging
system, so they all knew about RUNOFF and had access to the sources,
which were in the MAD language, an early partial implementation of
ALGOL.  My understanding was that Doug did the conversion of RUNOFF from
the CTSS extended character set to ASCII, using BCPL, under CTSS.  He
chose to name the port "roff", I think to distinguish the ASCII version
from the CTSS extended character set version.  When Bell Labs management
abandoned participation in the Multics project in 1969, Ken Thompson and
Dennis Ritchie went on to design Unix (effectively a much-stripped down
and simplified Multics) and C (an advance on B which itself was an
advance on BCPL) and ported McIlroy's roff into that system so that Unix
would have an immediate killer app.  If I recall correctly, Joe Ossanna
was also involved and a significant contributor to the roff-nroff-troff
series of commands, but I lack first-hand knowledge of that path.

Bell Labs had a GE-635 computer, but as far as I know it ran only GECOS,
a batch-processing system.  Bob and Doug together designed and
implemented the first PL/I compiler for Multics (staged on CTSS), but I
was unaware that Bob was involved in any of the RUNOFF or roff ports,
and I was surprised to see roff mentioned  in his obituary.  Your
suggestion that he might have ported it to GE-635 assembler language
seems quite far-fetched to me.  It might be a good idea to contact Doug
to get that part of the story straight.

Quite separately, someone at Dartmouth College did a port of RUNOFF to
DTSS, and Stu Madnick ported RUNOFF to CP/CMS at IBM, renaming it
"SCRIPT".  Therein began another branch--Goldfarb, Mosher and Lorie at
IBM found the RUNOFF/SCRIPT language so frustratingly unsystematic that
they developed a much more systematic extension for SCRIPT named GML.
GML is the parent of SGML and HTML, and grandparent of today's XML.  But
that is another story.

If you have any other questions, don't hesitate to ask.

                                            Jerry Saltzer

On Oct 21, 2011, at 3:44 PM, Kristaps Dzonsons wrote:

> > Professor,
> > 
> > While composing an article on UNIX manpages, I've been side-tracked
> > in creating a history of UNIX manpages from primary sources.  (The
> > Internet is a mess of secondary sources, most of them, it turns out,
> > completely wrong.)  The domain of my side-track is from your CTSS
> > RUNOFF to the next-generation tools (after GNU troff).
> > 
> > In short, I wonder if you have time for a few questions on this
> > matter.  Fortunately, the CTSS source listings managed by Paul
> > Pierce contain your RUNOFF sources, which constitute the primary
> > sources for my pre-UNIX period of study.
> > 
> > First question: Bob Morris' roff (nee runoff) clone on the GE-635
> > (DTSS, I'm guessing).  This bit of code seems completely lost (or?).
> > Did you work with Bob on this utility?  If so, am I correct in
> > assuming it was in GE-635 assembler?  Can you give a ballpark year
> > of its production?
> > 
> > Second, and a bit more vaguely, were you inspired by any prior
> > utilities or ideas for developing the line-based, dot-controlled
> > RUNOFF language?
> > 
> > That about wraps it up!  Beyond digging through sources from the
> > CTSS listings to McIlroy's BCPL source, Kernighan's troff, etc., I'm
> > trying to reconstruct the chain of influence from RUNOFF onward --
> > hence the last question.
> > 
> > Thank you,
> > 
> > Kristaps