Subject: Re: Reconstructing UNIX manpage history: ditroff(1)
From: Brian Kernighan 
Date: 23/10/2011 01:40
To: Kristaps Dzonsons

Hi, Kristaps --

Maybe I can answer some questions, though memory keeps getting dimmer.

After Joe Ossanna died in late 1977, troff was static for probably close
to two years, since no one had the time and courage to touch it.  It was
entirely in C at that point, somewhat over 9000 lines as I remember.  I
simply modified it, gradually getting around some of the limitations on
fonts, making use of dynamic memory, and generating "device-independent"
output for devices whose properties were specified in dynamically loaded
files.  I also added some simple graphics commands to permit lines,
circles, and other curves; this was necessary for the Pic language that
I was playing with.  I also played with characters that had longer names
and that only had numbers, and for Plan 9 I did some tools that made it
possible to handle Japanese, but that experiment wasn't picked up much.

I'm pretty sure that I only talked about a "device independent troff";
the name "ditroff" came from somewhere else, and I've never been fond of
it.  At some point I merged troff and nroff into a single source, though
with conditional compilation and probably something in the makefile to
keep the pieces organized.  I don't now remember whether it was actually
a single executable or two different ones with a lot of shared code.

As to my own formatter, when I was a grad student at Princeton, I wrote
a simple formatter that I called Roff ("an abbreviated version of
Runoff") and used it to print my thesis.  I had encountered Jerry
Saltzer's Runoff at MIT (during the summer of 1966), and then seen
programs like Doug McIlroy's roff and Ossanna's nroff in 1967 and 1968
during summers at Bell Labs.  My roff was a 1000-line Fortran program
that interpreted a handful of commands like .sp for a blank line (very
much like all the other *roff's) and printed filled and justified lines
of text.  The one point of novelty was that since the input was on
punched cards and thus all in upper case, it did automatic
capitalization of sentences, with some multi-punch sequences to handle
exceptions.

It was fine for my thesis, and it was picked up by a student agency and
used for another 5 or more years as a service: pay some money and get
neatly printed text.  The code has long since disappeared (it was a deck
of cards) but I still have a copy of the original documentation (on
paper only).

Hope this helps.  Good luck with chasing everything down.  It's hard to
do good history, but it's important.  Let me know if I can do anything
more.

Brian

On Sat, 22 Oct 2011, Kristaps Dzonsons wrote:

> Ken,
>
> While researching an article on the "mdoc" roff macro package, I was
> discouraged by the inconsistency of secondary sources regarding
> initial development of the UNIX manpage.  It seemed every source had
> slightly different (and at times conflicting) dates, developers,
> places, and so on. So I've set about reconstructing the history from
> primary sources, from Saltzer's CTSS source code onward.
>
> I wonder if you have time to answer a few questions about your pivotal
> role in this process, as there are some holes in the primary record.
>
> There are four main gaps:
>
> (1) Bob Morris' involvement has no source record at all and neither
> Saltzer nor McIlroy can recall the details of his circa-1967 runoff on
> the GE-635 (or "pre-Multics" arch in general);
> (2) Ritchie's pre-UNIX PDP-7 "rf" utility, possibly based on Morris'
> port, has no source and no real record of existence beyond hearsay;
> (3) the exact authorship of the PDP-11 roff/nroff/troff (V1, V2, V4
> AT&T UNIX), although McIlroy's fairly sure and I'm waiting to hear
> from Thompson; and lastly,
> (4) the hand-off of Ossanna's C-language troff(1), after his untimely
> death, to your own, then to what would become ditroff(1).
>
> I've written off (1) as totally unknown and (2) as something Thompson
> may know.  (3) is shaping up, but could use corroboration since the
> available source archives are incomplete.
>
> As for (4), can you give an account of this sequence?  How much of the
> C-language troff(1) had been written, and did you use it as a basis or
> start directly into your own?  When was ditroff(1) named as such, and
> was there ever a distinct C-language troff(1) and one that would later
> be named ditroff(1)?
>
> Furthermore, McIlroy hints that you'd written a formatter while at
> Princeton before working on Ossanna's.  Did this work go anywhere, and
> was it related to the roff set of tools?
>
> I understand that this is a lot of questioning, but the historical
> record is quite sparse.  (Yes, I have more questions, but the above
> would make a complete record from RUNOFF to runoff to roff, nroff,
> troff, groff, and so on.  It's a start...)
>
> As to my own interest, I wrote mandoc(1) (http://mdocml.bsd.lv/), now
> used in a few BSD UNIX systems instead of GNU troff; then the
> [unpublished] mdoc book, http://manpages.bsd.lv/.  I've noticed that
> many GNU/Linux types are in a balkanised mentality of texinfo (bad)
> vs. "man" macro (inelegant) vs. "ad hoc on-line HOWTO" (awful).  This
> segue into history stems from a LinuxJournal article on writing mdoc
> manuals to, I hope, encourage some consistency.
>
> Lastly, please accept my sincere condolences regarding Ritchie's
> death.  It was shocking news, and I'm sorry.
>
> Thank you and take care,
>
> Kristaps
>