> From: Nemo
> I have read that one of the first groups in AT&T to use early Unix was
> the legal dep't, specifically to use *roff to write patent applications.
> Can anyone elaborate on this or supply references?
Are you familiar with the description in Dennis M. Ritchie, "The Evolution of
the Unix Time-sharing System":
https://www.bell-labs.com/usr/dmr/www/hist.htm
(in the section "The first PDP-11 system")? Not a great deal of detail, but...
> It would also be interesting to learn how the writers were taught *roff,
> what editors were used
I'm pretty sure 'ed' was the only editor available at that point.
Noel
> From: Clem Cole
> Programmer's Workbench - aka PWB was John Mashey and team in Whippany.
> They took a V6 system and make some changes
I was suprised to find, reading the article on it in the Unix BSTJ issue, that
the system changes were less than I'd thought. Some of the stuff in the PWB1
release that we have (see previous message) is _not_ described in that article
(which is fairly detailed), which further compounds the lack of clarity over
who/what/when between V6 and V7.
> Noel may know how it made it to MIT
That I _do_ know! There was some sort of Boy Scouts group at Bell (not sure
exactly where) and one of the members went to MIT. I think he was doing
undergraduate research work in the first group at MIT to have Unix (Steve
Ward's), but anyway he had some connection there; and I think also had a
summer job at Bell. He was the Bell->MIT conduit.
> PWB 2.0 was released a few years later and was based on the UNIX/TS
> kernel and some other changes and it was around this time that the UNIX
> Support Group was formed
??? If PWB1 was in July '77, and PWB2 was some years later, USG couldn't have
been formed 'around [that] time' because there's that USG document from
January '76?
Noel
> From: Jon Forrest <nobozo(a)gmail.com>
> John Mashey had a lot to do with PWB so maybe he can say a few words
> about it if he's on here.
It would be great to have some inside info about the relationship among the
Research, USG and PWB systems. Clearly there was comunication, and things got
passed around, but we know so little about what was happening during the
period between V6 and V7 when a lot happened (e.g. the changes to C, just
mentioned).
E.g. check out the PWB1 version of exec():
https://minnie.tuhs.org//cgi-bin/utree.pl?file=PWB1/sys/sys/os/sys1.c
It's been changed from V6 to copy the arguments into swap space, _not_ buffers
allocated from the system buffer pool (which is how V6 does it). So, who did
this originally - did the PWB people do it, or was it something the research
people did, that PWB picked up?
I used to think it was USG, but there's a 'Unix Program Description' document
prepared by USG, dated January 1976, and it's still clearly using the V6
approach. The PWB1 release was allegedly July, 1977:
https://minnie.tuhs.org//cgi-bin/utree.pl?file=PWB1
(Which is, AFAIK, the _only_ set of sources we have for after V6 and before V6
- other than the MIT system, which seems to be basically PWB1.)
So who did the exec() changes, originally?
And I could list a bunch more like this...
Noel
I never really learned VI. I can stumbled through it in ex mode if I have
to. If there's no EMACS on the UNIX system I'm using, I use ed.
You get real good at regular expressions. Some of my employees were
pretty amazed at how fast I could make code changes with just ed.
Here's part of the story.
> From: "Doug McIlroy" <doug(a)cs.dartmouth.edu>
> To:<tuhs(a)minnie.tuhs.org>
> Sent:Fri, 16 Dec 2016 21:09:16 -0500
> Subject:[TUHS] How Unix made it
to the top
>
> It has often been told how the Bell Labs law department became the
> first non-research department to use Unix, displacing a newly acquired
> stand-alone word-processing system that fell short of the department's
> hopes because it couldn't number the lines on patent applications,
> as USPTO required. When Joe Ossanna heard of this, he told them about
> nroff and promised to give it line-numbering capability the next day.
> They tried it and were hooked. Patent secretaries became remote
> members of the fellowship of the Unix lab. In due time the law
> department got its own machine.
Come to think of it, they must already have had a machine, probably
leased from the commercial word-processing company, for they had DEC
tapes they needed to convert to Unix format. Several of us in the Unix
lab spent a memorable afternoon decoding the proprietary format. It was
finally broken when we computed a bitwise autocorrelation function. It
had a big peak at seven. The tapes were pure ASCII rather than bytewise
ASCII--a lot of work for very little data compression.
As for training, the secretaries had to learn nroff and ed plus the
usual lot of ls, mkdir, mv, cp, rm. The patent department had to invest
in modems and order phone lines to plug them into. I don't know what
terminals they used.
>From this distant point in time it seems that it all happened in a couple
of weeks. Joe Ossanna did most of the teaching, and no doubt supplied
samples to copy. As far as I know the only other instructional materials
would have been man pages and the nroff manual (forbiddingly terse,
though thorough). He may have made a patent-macro package, but I doubt
it; I think honor for the first real macro package goes to Lesk's -ms.
Doug
Larry’s question about PWB made me think it might be useful to this list
for some of this to be written down.
When you write the story of UNIX, licensing is a huge part of it (both good
and bad). As I have tried to explain before the 1956 consent decree and
the later 1980 Judge Green ruling, as well as how the AT&T legal department
set up the licenses really cast huge shadows that almost seem trite today;
but seem to have been forgotten.
In fact later licensing would become so important, one of the more infamous
UNIX wars was based on it (if you go back to the original OSF Founding
Principles – two of them are ‘Fair and Stable Licensing Terms’). As we
all know, because of the original 1956 decree, AT&T was not allowed to be
in the computer business and so when people came calling both to use it
(Academically and Commercially) and to relicense it; history has shown that
AT&T’s management killed the golden goose. I’d love to hear the views of
Doug, Steve, Ken and other who were inside looking out.
FWIW: These are my thoughts from an Academic and Commercial user back in
the day. AT&T’s management was clearly concerned about the consent decree
and the original licenses show it. UNIX was licensed for free to academic
institutions for research use (sans a small tape copying fee) and the bits
were ‘abandoned on your door step.’ This style of license, along with
the publishing of the ideas behind really did get the ideas out and the
academic community loved it. We used it and we were able to share
everything.
The academic license was fine until people want to start to use in a
commercial setting (Rand Corp). Again, AT&T legal is worried about being
perceived in the computer business, so the original commercial use license
shows it. AT&T licensing basically uses the academic license but add the
ability to actually use it for commercial purposes. Then the first
Universities start to want to use UNIX more like a commercial system [Dan
Klein and I would go on strike and force CMU to purchase first commercial
use license for an Academic setting, following by Case Western].
As AT&T management realized the UNIX IP did seem to be some value, just
like the transistor had been earlier, it seems like they wanted to find a
way to keep it under their control. I remember having a dinner
conversation with Dennis at a USENIX about this topic. Steve has expressed
they told many folks to treated it as a ‘trade secret’ (which is strange to
me since the cat as already out of the bag by then and the ideas (IP)
behind UNIX had already been extensively published (we even had USENIX
formed to discuss ideas).
By the time Judge Green allows AT&T to be in the computer business I think
AT&T management completely misunderstood the value of what they had. The
AT&T legal team had changed the commercial rules in every new UNIX release
a new license was created, and thus firms like DEC, HP, IBM *et al* were
getting annoyed because they had begun to invest in the technology
themselves and the feeling inside of those firms was that AT&T management
was changing the ground rules after the game started.
IMO a funny thing happened (bad karma), it seems like the tighter AT&T
management seems to try to control things in the UNIX community, the less
control the community gave them. Clearly, the new features of the
technology started to be driven by BSD. But the license was the one place
they could control and they tried. In fact, by the time of the SVR4 it
all came to a head and OSF was formed because most firms were unwilling to
give AT&T the kind of control they were asking in the that license [as
Larry has previously expressed, Sun made a Faustian deal WRT to SVR4]. In
the end, the others were shipping from an SVR3 license or had bought it
out.
> From: Clem cole
> Thinking about this typesetter C may have been later with ditroff.
Not so sure about that; we had that C at MIT, but only regular troff (which
had been hacked to drive a Varian).
> From: Arnold Skeeve
> It seems to be shortly after the '78 release of V7.
No, typesetter C definitely pre-dated V7. The 'PWB1' system at MIT had the new
C.
Looking at the documentation files for the extension (e.g. addition of
'long's), none of them have dates in them (alas), but my hard-copy printout of
one is dated "May 8 1978", and it was several years old at that point.
(Also, several sources give '79 for V7 - Salus says 'June 1979').
Noel
> From: Clem Cole
> Their is a open question about the need to support self modifying code
> too. I personally don't think of that as important as the need for
> conditional instructions which I do think need to be there before you
> can really call it a computer.
Here's one way to look at it: with conditional branching, one can always
write a program to _emulate_ a machine with self-modifying code (if that's
what floats your boat, computing-wise) - because that's exactly what older,
simple microcoded machines (which don't, of course, have self-modifying code
- their programs are in ROM) do.
Noel
Way back on this day in 1941, Conrad Zuse unveiled the Z3; it was the
first programmable automatic computer as we know it (Colossus 1 was not
general-purpose). The last news I heard about the Z3 was that she was
destroyed in an air-raid...
This pretty much started computing, as we know it.
-- Dave
