Subject: Re: UNIX manpages & questions regarding roff(1) history.
From: Ken Thompson
Date: 23/10/2011 09:49
To: Kristaps Dzonsons

the roffs all descend from runoff from mit's ctss system.  i wrote rf
(then roff) for the pdp-7 for some early manual work someone else
(probably dennis) wrote roff in c on the pdp-11.  this version of roff
was written to do patent applications.  the patent application was the
justification for getting the pdp-11.

joe ossanna wrote nroff, which was kinda programmable via macros. then
with a large set of macros it could emulate roff. he then added fonts
and sizes to make troff for the newly interfaced typesetter.  brian
kernighan (i think) combined nroff and troff to be the same program.

i did the start of the man pages and set up the title/synopsis/...
stuff. the loose prototype for the man pages was the multics man (MSPM).
the mspm was much wordier and less structured, but many sections had
similar headings.

this is my best recollection, but certainly could be wrong.

ken

On Fri, Oct 21, 2011 at 4:16 PM, Kristaps Dzonsons wrote:
> > Ken,
> >
> > While writing an article on UNIX manpage composition (to address the
> > rather sad state of affairs in the Linux world), I've been
> > side-tracked in producing an accurate history of the UNIX manpage.
> > The various [secondary] Internet sources are ambiguous and, in turns
> > out, at times wildly wrong.
> >
> > (Back-story: I wrote mandoc(1) [1], now in most BSD UNIX systems, as
> > a domain-specific compiler for the mdoc and man roff macro packages,
> > and the unpublished mdoc book [2].  This is for a LinuxJournal
> > article, to bring some proper manpageness to the masses.)
> >
> > So in short, after combing through the extant V1-7 AT&T UNIX sources
> > [5] and those for Multics [3] and CTSS [4] before, I still have some
> > questions unanswered by the corpus of source material.  I wonder if
> > you can divulge me with some questions regarding the first versions
> > of roff(1).
> >
> > But first, please accept my condolences regarding Dennis Ritchie's
> > death.  It was shocking news.  I've never really used anything
> > beyond UNIX-ish systems, and to find a part of this picture suddenly
> > missing affects me deeply.  I'm sorry.
> >
> > My question lies in V1 AT&T UNIX roff(1).  In secondary sources, I
> > find frequent mention of Ossanna writing the PDP-11 versions of V1
> > AT&T UNIX roff(1), then nroff(1) in V2, then troff(1) in V4.  But
> > then in a recent email exchange, Doug McIlroy fingered you and
> > Ritchie as writing the original roff(1) for V1, with Ossanna only
> > taking over for nroff(1) and troff(1).  The source for V2 and V4
> > nroff(1) and troff(1) is gone, but the manuals remain and indicate
> > all three of you (jfo, ken, dmr).  Can you straighten the record?
> >
> > (I'll separately mail Kernighan regarding ditroff(1) and the
> > C-language rewrite in general.)
> >
> > Assuming it was your initial work and not Ossanna's, the next
> > question is the prior work.  It's asserted that V1 AT&T UNIX roff(1)
> > is a re-write of McIlroy's BCPL roff.  But there's also mention of
> > Morris' GE-635 roff.  Which was the jumping-off point for UNIX's
> > roff?
> >
> > Lastly: there's mysterious reference to a pre-UNIX PDP-7 rf by
> > Ritchie.  Do you know anything of this?  Sources, references?
> >
> > Well... one less-important question that's just been bothering me
> > for a long time: the V1 AT&T UNIX Programmer's Manual, regarding the
> > structure of manpages, is very specific.  Were there documents
> > before this that you drew from?  I've long been curious whether
> > there's more to the history of the NAME/SYNOPSIS layout and
> > bold/underline and [optional] conventions.
> >
> > I appreciate any thoughts you might have on these missing pieces of
> > the historical record.
> >
> > Thank you,
> >
> > Kristaps
> >
> >
> > [1] http://mdocml.bsd.lv/ [2] http://manpages.bsd.lv/ [3] From the
> > Multics Internet Server.  [4] From Paul Pierce's CTSS source
> > listings.  [5] From the TUHS archive and Ken Wellsch's raw tape
> > dumps.
> >