30 Apr
2025
30 Apr
'25
1:11 p.m.
From: Clem Cole
Yes, that was one of the RTS compilers for the NU
machine. John Romkey
may have done it, as he was the primary person behind PCIP
I decided to poke around in the 'MIT-CSR' dump, since that was the machine
the PC/IP project started on, to see what I could find. Hoo boy! What an
adventure!
In the PC/IP area, I found a 'c86' directory - but it was almost empty. It
did have a shell file, 'grab', which contained:
tftp -g $1 xx "PS:<Wayne>$1"
and a 'graball' file which called 'grab' for the list of compiler
source
files. ('xx' was MIT-XX, the TOPS-20 main time-sharing machint of LCS.)
So I did a Web search for Wayne Gramlich (with whom I hadn't communicated in
many decades), and he popped right up. (Amazing thing, this Internet thingy.
Who'd have ever thought, back in the day, that it would turn into what it
did? Well, probably John Brunner, whom I (sadly) never met, who was there
before any of us.)
I took a chance, and called his number, and he was there, and we had a long
chat. He absolutely didn't do it, although he wrote the loader the project
used ('l68', the source for which I did find.) He's virtually certain
Romkey
didn't (which would have been my guess too; Romkey was like a sophmore when
the project started). His best (_very_ faded) memory was that they started off
with a commercial compiler. (But see below.)
That leaves several mysteries. 1) Why would a commercial compiler not come
with a linker? 2) Why did people who wanted to work with the PC/IP source
need a Bell license?
I did some more poking, and the list of files for the 86 compiler, from
'graball':
trees.c optim.c pftn.c code.c local.c scan.c xdefs.c
table.c reader.c local2.c order.c match.c allo.c comm1.c
manifest mfile1 common macdefs mfile2 mac2defs
matched the file names from 'pcc', as given in "A Tour Through the Portable
C
Compiler":
https://maibriz.de/unix/ultrix/_root/porttour.pdf
(in section "The Source Files"). So whether the 86 compiler was done at MIT
(by someone in RTS), or at a company, it was definitely a 'pcc' descendant.
(Possibly adding to the confusion, we had some other C compilers for various
ISA's in that project [building networking software for various
micro-computers], including an 8080 C compiler from Whitesmiths, Ltd, which I
have also found. It's possible that Wayne's vague memory of a commercial
compiler is of that one?)
I really should reach out to Romkey and Bridgham, to see what they remember.
Later today.
Whether the main motivation for keeping the compiler source on XX was i)
because disk space was short on CSR (we had only a hand-me-down pair of
CalComp Model 215 drives - capacity 58 Mbytes per drive!); ii) to prevent
version skew; or iii) because it was a commercial compiler, and we had to
protect the source (e.g. we didn't have the source to the 8080 compiler, only
the object modules), I have no idea.
Anyway the MIT RTS folks made hardware and PCC back
ends for the 68K,
Z8000 and 8086. I believe that each had separate assemblers, tjt who
sometimes reads this list might know more, as he wrote the 68K assembler.
There is an 'a86' directory on CSR, but it too is empty, except for a
'grab'
command file. That contains only:
tftp -g $1 xx "PS:<novick>$1"
I have no memory of who 'novick' might have been. A Web search for 'novick
mit lcs' didn' turn anything up. (I wonder if it might have been Carol
Novitsky; she was in our group at LCS, and I have a vague memory of her being
associated with the networking software for micro-computers project.)
Anyway, it probably doesn't matter; the c86 'grab' referred to Wayne, but
he
didn't write c86; 'novick' might not have written a86.
Something else to ask Romkey and Bridgham about.
Noel
30 Apr
30 Apr
1:19 p.m.
New subject: Any Interdata war stories?
On Apr 30, 2025, at 9:11 AM, Noel Chiappa
<jnc(a)mercury.lcs.mit.edu> wrote:
Who'd have ever thought, back in the
day, that it would turn into what it
did? Well, probably John Brunner, whom I (sadly) never met, who was there
before any of us.)
The reference here is to Shockwave Rider, by John Brunner in 1975.
I did meet him once. He came by PARC to talk to John Shoch and John Hupp, who
had written a network “worm” after Brunner’s naming. A very prescient author!
“The “worm” programs — early experience with a distributed computation”. Shoch and Hupp
https://dl.acm.org/doi/10.1145/102616.102636
-Larry
1:34 p.m.
New subject: Any Interdata war stories?
It isn’t Unix related, but it is a fun Interdata war story.
I hung out at the MIT Architecture Machine in the early 70’s (the Media Lab before it was
the Media Lab.)
The lab used Interdata minis, with homegrown software. My part was designing and building
various I/O devices, including parts of a network interconnecting the various minis.
There were Interdata model 3, model 5, model 75, and one blazing fast model 85 with
semiconductor memory. The rest were core.
I think the first 7/32 we got had two 32K core modules. There was a microcode bug, such
that in some case the microcode did not disable a non-maskable interrupt, so if the ISR
was so foolish as to cause one, the machine got stuck in an interrupt loop. You couldn’t
clear it by reset or by power cycling the machine, because core! The bad state was in the
memory and was non-volatile. It was possible to clear (sometimes) by swapping the two
core modules with the power off, if the other one didn’t have the poison bits, but if they
did, the only thing that worked was to unplug the memory module with the power on.
Luckily the OS guys figured out how to fix the ISR before we trashed anything permanently.
-L
1:53 p.m.
New subject: Any Interdata war stories?
On Wed, 30 Apr 2025, Noel Chiappa wrote:
That leaves several mysteries. 1) Why would a
commercial compiler not come
with a linker? 2) Why did people who wanted to work with the PC/IP source
need a Bell license?
A possible answer for 1, since we're probably talking about a compiler for
MS-DOS: most releases of MS-DOS and PC DOS, at least through version 3.3,
came with a linker, so maybe they didn't deem it necessary.
-uso.
2:07 p.m.
New subject: PC/IP (was: Any Interdata war stories?)
On Wed, Apr 30, 2025 at 09:11:35AM -0400, Noel Chiappa wrote:
"a version of the portable C Compiler that was modified by Chris Terman
to produce code for an 8086 microprocessor was ported from the RTS VAX/780
to the CSR PDP-11/45."
https://people.csail.mit.edu/saltzer/Multics/MHP-Saltzer-060508/bookcases/R…
"If you think that you need the source code, you should realize that a
prerequisite to compiling the PC/IP programs is that you must have
imported Chris Terman's 8086 version of the UNIX Portable C compiler and
associated loader and assember systems. That importation in turn requires
a UNIX system, a current UNIX license, and negotiation with Chris Terman."
https://web.mit.edu/saltzer/www/publications/pcmemo.pdf
From: Clem
Cole
Yes, that was one of the RTS compilers for the NU
machine. John Romkey
may have done it, as he was the primary person behind PCIP
I decided to poke around in the 'MIT-CSR' dump, since that was the machine
the PC/IP project started on, to see what I could find. Hoo boy! What an
adventure!
In the PC/IP area, I found a 'c86' directory - but it was almost empty. It
did have a shell file, 'grab', which contained:
tftp -g $1 xx "PS:<Wayne>$1"
and a 'graball' file which called 'grab' for the list of compiler
source
files. ('xx' was MIT-XX, the TOPS-20 main time-sharing machint of LCS.)
So I did a Web search for Wayne Gramlich (with whom I hadn't communicated in
many decades), and he popped right up. (Amazing thing, this Internet thingy.
Who'd have ever thought, back in the day, that it would turn into what it
did? Well, probably John Brunner, whom I (sadly) never met, who was there
before any of us.)
I took a chance, and called his number, and he was there, and we had a long
chat. He absolutely didn't do it, although he wrote the loader the project
used ('l68', the source for which I did find.) He's virtually certain
Romkey
didn't (which would have been my guess too; Romkey was like a sophmore when
the project started). His best (_very_ faded) memory was that they started off
with a commercial compiler. (But see below.)
That leaves several mysteries. 1) Why would a commercial compiler not come
with a linker? 2) Why did people who wanted to work with the PC/IP source
need a Bell license?
I did some more poking, and the list of files for the 86 compiler, from
'graball':
trees.c optim.c pftn.c code.c local.c scan.c xdefs.c
table.c reader.c local2.c order.c match.c allo.c comm1.c
manifest mfile1 common macdefs mfile2 mac2defs
matched the file names from 'pcc', as given in "A Tour Through the
Portable C
Compiler":
https://maibriz.de/unix/ultrix/_root/porttour.pdf
(in section "The Source Files"). So whether the 86 compiler was done at MIT
(by someone in RTS), or at a company, it was definitely a 'pcc' descendant.
(Possibly adding to the confusion, we had some other C compilers for various
ISA's in that project [building networking software for various
micro-computers], including an 8080 C compiler from Whitesmiths, Ltd, which I
have also found. It's possible that Wayne's vague memory of a commercial
compiler is of that one?)
I really should reach out to Romkey and Bridgham, to see what they remember.
Later today.
Whether the main motivation for keeping the compiler source on XX was i)
because disk space was short on CSR (we had only a hand-me-down pair of
CalComp Model 215 drives - capacity 58 Mbytes per drive!); ii) to prevent
version skew; or iii) because it was a commercial compiler, and we had to
protect the source (e.g. we didn't have the source to the 8080 compiler, only
the object modules), I have no idea.
Anyway the MIT RTS folks made hardware and PCC
back ends for the 68K,
Z8000 and 8086. I believe that each had separate assemblers, tjt who
sometimes reads this list might know more, as he wrote the 68K assembler.
There is an 'a86' directory on CSR, but it too is empty, except for a
'grab'
command file. That contains only:
tftp -g $1 xx "PS:<novick>$1"
I have no memory of who 'novick' might have been. A Web search for 'novick
mit lcs' didn' turn anything up. (I wonder if it might have been Carol
Novitsky; she was in our group at LCS, and I have a vague memory of her being
associated with the networking software for micro-computers project.)
Anyway, it probably doesn't matter; the c86 'grab' referred to Wayne, but
he
didn't write c86; 'novick' might not have written a86.
Something else to ask Romkey and Bridgham about.
Noel 
1 May
1 May
2:36 a.m.
New subject: PC/IP (was: Any Interdata war stories?)
So did Chris compiler go back to the Nu project? And thank you to Al for
the Trix sources. As I said. It was national 16032 not z8000.
FYI I had the whitesmith compiler at one point also. It generated code for
a funky assembler called “anat” for “a natural assembler” which PJ conjured
up for writing the compiler. It was was sort of a mix between an IL and
8080 assembler if my memory is correct. I’d love to see that
distribution both the doc and the compiler again to May with on SIMH. I
suspect I might appreciate it more than I did in those days. I hated it
then but as I learned more about compilers and architectures PJ might have
been on to something.
Sent from a handheld expect more typos than usual
On Wed, Apr 30, 2025 at 10:07 AM Jonathan Gray <jsg(a)jsg.id.au> wrote:
On Wed, Apr 30, 2025 at 09:11:35AM -0400, Noel Chiappa
wrote:
Romkey
machine
From:
Clem Cole
> Yes, that was one of the RTS compilers for the NU machine. John may have
done it, as he was the primary person behind PCIP
I decided to poke around in the 'MIT-CSR' dump, since that was the the PC/IP project started on, to see what I could
find. Hoo boy! What an
adventure!
In the PC/IP area, I found a 'c86' directory - but it was almost empty.
It
did have a shell file, 'grab', which
contained:
tftp -g $1 xx "PS:<Wayne>$1"
and a 'graball' file which called 'grab' for the list of compiler
source
files. ('xx' was MIT-XX, the TOPS-20 main time-sharing machint of LCS.)
So I did a Web search for Wayne Gramlich (with whom I hadn't
communicated in
many decades), and he popped right up. (Amazing
thing, this Internet
thingy.
Who'd have ever thought, back in the day,
that it would turn into what it
did? Well, probably John Brunner, whom I (sadly) never met, who was there
before any of us.)
I took a chance, and called his number, and he was there, and we had a
long
chat. He absolutely didn't do it, although
he wrote the loader the
project
used ('l68', the source for which I did
find.) He's virtually certain
Romkey
didn't (which would have been my guess too;
Romkey was like a sophmore
when
the project started). His best (_very_ faded)
memory was that they
started off
with a commercial compiler. (But see below.)
That leaves several mysteries. 1) Why would a commercial compiler not
come
with a linker? 2) Why did people who wanted to
work with the PC/IP source
need a Bell license?
I did some more poking, and the list of files for the 86 compiler, from
'graball':
trees.c optim.c pftn.c code.c local.c scan.c xdefs.c
table.c reader.c local2.c order.c match.c allo.c comm1.c
manifest mfile1 common macdefs mfile2 mac2defs
matched the file names from 'pcc', as given in "A Tour Through the
Portable C
Compiler":
https://maibriz.de/unix/ultrix/_root/porttour.pdf
(in section "The Source Files"). So whether the 86 compiler was done at
MIT
(by someone in RTS), or at a company, it was
definitely a 'pcc'
descendant.
(Possibly adding to the confusion, we had some other C compilers for
various
ISA's in that project [building networking
software for various
micro-computers], including an 8080 C compiler from Whitesmiths, Ltd,
which I
have also found. It's possible that
Wayne's vague memory of a commercial
compiler is of that one?)
I really should reach out to Romkey and Bridgham, to see what they
remember.
Later today.
Whether the main motivation for keeping the compiler source on XX was i)
because disk space was short on CSR (we had only a hand-me-down pair of
CalComp Model 215 drives - capacity 58 Mbytes per drive!); ii) to prevent
version skew; or iii) because it was a commercial compiler, and we had to
protect the source (e.g. we didn't have the source to the 8080 compiler,
only
the object modules), I have no idea.
> Anyway the MIT RTS folks made hardware and PCC back ends for the
68K,
> Z8000 and 8086. I believe that each had
separate assemblers, tjt
who
> sometimes reads this list might know
more, as he wrote the 68K
assembler.
There is an 'a86' directory on CSR, but it too is empty, except for a
'grab'
command file. That contains only:
tftp -g $1 xx "PS:<novick>$1"
I have no memory of who 'novick' might have been. A Web search for
'novick
mit lcs' didn' turn anything up. (I
wonder if it might have been Carol
Novitsky; she was in our group at LCS, and I have a vague memory of her
being
associated with the networking software for
micro-computers project.)
Anyway, it probably doesn't matter; the c86 'grab' referred to Wayne,
but he
didn't write c86; 'novick' might
not have written a86.
Something else to ask Romkey and Bridgham about.
Noel
"a version of the portable C Compiler that was modified by Chris Terman
to produce code for an 8086 microprocessor was ported from the RTS VAX/780
to the CSR PDP-11/45."
https://people.csail.mit.edu/saltzer/Multics/MHP-Saltzer-060508/bookcases/R…
"If you think that you need the source code, you should realize that a
prerequisite to compiling the PC/IP programs is that you must have
imported Chris Terman's 8086 version of the UNIX Portable C compiler and
associated loader and assember systems. That importation in turn requires
a UNIX system, a current UNIX license, and negotiation with Chris Terman."
https://web.mit.edu/saltzer/www/publications/pcmemo.pdf
3:38 a.m.
New subject: PC/IP (was: Any Interdata war stories?)
Chris was part of the Nu project.
"Was a principal developer of the NuMachine"
"developed a family of portable C compilers for the (then) newly
available microprocessors. These compilers were widely distributed as
the first C implementations for the x86 and 68K processors."
https://people.csail.mit.edu/cjt/resume.html
On Wed, Apr 30, 2025 at 10:36:01PM -0400, Clem Cole wrote:
So did Chris compiler go back to the Nu project? And
thank you to Al for
the Trix sources. As I said. It was national 16032 not z8000.
FYI I had the whitesmith compiler at one point also. It generated code for
a funky assembler called “anat” for “a natural assembler” which PJ conjured
up for writing the compiler. It was was sort of a mix between an IL and
8080 assembler if my memory is correct. I’d love to see that
distribution both the doc and the compiler again to May with on SIMH. I
suspect I might appreciate it more than I did in those days. I hated it
then but as I learned more about compilers and architectures PJ might have
been on to something.
Sent from a handheld expect more typos than usual
On Wed, Apr 30, 2025 at 10:07 AM Jonathan Gray <jsg(a)jsg.id.au> wrote:
> On Wed, Apr 30, 2025 at 09:11:35AM -0400, Noel Chiappa wrote:
> > > From: Clem Cole
> >
> > > Yes, that was one of the RTS compilers for the NU machine. John
> Romkey
> > > may have done it, as he was the primary person behind PCIP
> >
> > I decided to poke around in the 'MIT-CSR' dump, since that was the
> machine
> > the PC/IP project started on, to see what I could find. Hoo boy! What an
> > adventure!
> >
> > In the PC/IP area, I found a 'c86' directory - but it was almost
empty.
> It
> > did have a shell file, 'grab', which contained:
> >
> > tftp -g $1 xx "PS:<Wayne>$1"
> >
> > and a 'graball' file which called 'grab' for the list of
compiler source
> > files. ('xx' was MIT-XX, the TOPS-20 main time-sharing machint of
LCS.)
> >
> > So I did a Web search for Wayne Gramlich (with whom I hadn't
> communicated in
> > many decades), and he popped right up. (Amazing thing, this Internet
> thingy.
> > Who'd have ever thought, back in the day, that it would turn into what it
> > did? Well, probably John Brunner, whom I (sadly) never met, who was there
> > before any of us.)
> >
> > I took a chance, and called his number, and he was there, and we had a
> long
> > chat. He absolutely didn't do it, although he wrote the loader the
> project
> > used ('l68', the source for which I did find.) He's virtually
certain
> Romkey
> > didn't (which would have been my guess too; Romkey was like a sophmore
> when
> > the project started). His best (_very_ faded) memory was that they
> started off
> > with a commercial compiler. (But see below.)
> >
> > That leaves several mysteries. 1) Why would a commercial compiler not
> come
> > with a linker? 2) Why did people who wanted to work with the PC/IP source
> > need a Bell license?
> >
> >
> > I did some more poking, and the list of files for the 86 compiler, from
> > 'graball':
> >
> > trees.c optim.c pftn.c code.c local.c scan.c xdefs.c
> > table.c reader.c local2.c order.c match.c allo.c comm1.c
> > manifest mfile1 common macdefs mfile2 mac2defs
> >
> > matched the file names from 'pcc', as given in "A Tour Through
the
> Portable C
> > Compiler":
> >
> > https://maibriz.de/unix/ultrix/_root/porttour.pdf
> >
> > (in section "The Source Files"). So whether the 86 compiler was done
at
> MIT
> > (by someone in RTS), or at a company, it was definitely a 'pcc'
> descendant.
> >
> > (Possibly adding to the confusion, we had some other C compilers for
> various
> > ISA's in that project [building networking software for various
> > micro-computers], including an 8080 C compiler from Whitesmiths, Ltd,
> which I
> > have also found. It's possible that Wayne's vague memory of a
commercial
> > compiler is of that one?)
> >
> > I really should reach out to Romkey and Bridgham, to see what they
> remember.
> > Later today.
> >
> > Whether the main motivation for keeping the compiler source on XX was i)
> > because disk space was short on CSR (we had only a hand-me-down pair of
> > CalComp Model 215 drives - capacity 58 Mbytes per drive!); ii) to prevent
> > version skew; or iii) because it was a commercial compiler, and we had to
> > protect the source (e.g. we didn't have the source to the 8080 compiler,
> only
> > the object modules), I have no idea.
> >
> >
> > > Anyway the MIT RTS folks made hardware and PCC back ends for the
> 68K,
> > > Z8000 and 8086. I believe that each had separate assemblers, tjt
> who
> > > sometimes reads this list might know more, as he wrote the 68K
> assembler.
> >
> > There is an 'a86' directory on CSR, but it too is empty, except for a
> 'grab'
> > command file. That contains only:
> >
> > tftp -g $1 xx "PS:<novick>$1"
> >
> > I have no memory of who 'novick' might have been. A Web search for
> 'novick
> > mit lcs' didn' turn anything up. (I wonder if it might have been
Carol
> > Novitsky; she was in our group at LCS, and I have a vague memory of her
> being
> > associated with the networking software for micro-computers project.)
> >
> > Anyway, it probably doesn't matter; the c86 'grab' referred to
Wayne,
> but he
> > didn't write c86; 'novick' might not have written a86.
> >
> > Something else to ask Romkey and Bridgham about.
> >
> > Noel
>
> "a version of the portable C Compiler that was modified by Chris Terman
> to produce code for an 8086 microprocessor was ported from the RTS VAX/780
> to the CSR PDP-11/45."
>
>
https://people.csail.mit.edu/saltzer/Multics/MHP-Saltzer-060508/bookcases/R…
>
> "If you think that you need the source code, you should realize that a
> prerequisite to compiling the PC/IP programs is that you must have
> imported Chris Terman's 8086 version of the UNIX Portable C compiler and
> associated loader and assember systems. That importation in turn requires
> a UNIX system, a current UNIX license, and negotiation with Chris Terman."
> https://web.mit.edu/saltzer/www/publications/pcmemo.pdf
>
4:20 a.m.
New subject: PC/IP (was: Any Interdata war stories?)
On 4/30/25 8:38 PM, Jonathan Gray wrote:
Chris was part of the Nu project.
"Was a principal developer of the NuMachine"
"developed a family of portable C compilers for the (then) newly
available microprocessors. These compilers were widely distributed as
the first C implementations for the x86 and 68K processors."
https://people.csail.mit.edu/cjt/resume.html
I found most of the yearly LCS reports have been digitized to DTIC
which answered a bunch of my questions about who was doing what at
that time
I've archived them at http://bitsavers.org/pdf/mit/lcs/progress_reports
2 May
2 May
2:56 p.m.
New subject: PC/IP
On 5/1/25 12:20 AM, Al Kossow wrote:
On 4/30/25 8:38 PM, Jonathan Gray wrote:
Some background, though the MIT LCS progress reports should cover much
of this. I won't attempt to put any dates.
Chris Terman was one of the graduate students in the RTS group. Since
VT-52 terminals were relatively scarce, he designed and built his own
with a larger screen - something like 40 lines by maybe 120 or 132
characters, called the "Termanal". I don't remember if it used an 8080
to handle the control sequences in the data stream or something else.
He then got interested in designing a terminal that could display bit
map graphics, to be comparable to the graphics used on the Lisp Machines
just being built by the MIT-AI lab. I had stumbled across one of the LCS
progress reports that credits Professor Steve Ward and one of the
undergraduate staff, Rae McClellan in assisting the design of this bit
graph which was named the "Nu Terminal" (I don't think it was the "Nu
Termanal"). This used an 8086. A couple of these were built. One of the
undergraduate students, Jon Sieber, had been a member of an Explorer
Post in Murray Hill where Dennis Ritchie was the advisor. Jon would
regularly bring UNIX tapes from the Research Lab and included things
like early versions of the Portable C Compiler and the Circuit Design
Aids. Chris used the Circuit Design Aids to design wire-wrap boards for
the Nu Terminal and the RTS lab got a semi-automatic wire wrap machine.
Some students and staff took turns doing the actual wire wrapping. My
contribution was writing some simple software that simulated a paper
tape reader for the wire wrap machine.
An undergraduate student, Mike Patrick, did his bachelor's thesis
writing a table driven assembler and constructed tables for the 8086 and
I think an 8080. Later there were drivers for the Zylog Z8000, the
National Semiconductor NS16000 and the Motorola 68000. I contributed a
small bit of code for doing optimal choice of short vs long branches (to
branch to an address more than +/- 127 bytes, you had to branch around a
longer jump instruction).
Chris Terman did the work of modifying the Portable C Compiler to
generate code for the 8086, the Z8000, NS16000 and MC68000. I think we
may have built one machine with the Z8000, but quickly settled on using
the MC68000, primarily because of the 32-bit support (one progress
report says that Zenith was supposed to build multiple Z8000 based
machines, but I don't remember those. The NS16000 had better memory
management, but I don't think we ever actually received any CPU chips.
Anyway, these compilers were what was distributed, and the MC68000
compiler in particular was used by almost all the companies that came
out the MC68000-based Unix machines. Apollo was a notable exception, but
Apollo wrote their own operating system from scratch rather than Unix.
Side note: Bill Poduska came to visit Steve Ward and before the visit
Steve was all excited, but was disappointed that Bill was not going to
use Unix.
Before the RTS group used Unix, they had written a small timesharing
system for the PDP-11/45 that was used in the 6.031 introductory
computer science course taught by Mike Dertouzos. Chris was involved in
maintaining that, though I think Steve Ward was probably the main
implementor. Chris had also spent too many hours changing address
jumpers on Unibus and other controllers as well as tweaking Unix mkconf
files, and thought that while the 4BSD autoconfiguration was an
improvement, there should be a better way. Chris and Steve designed the
Nu bus, and the Nu Bus was used in the MC68000 boards. Eventually it was
picked up by Apple.
Chris was one of many students who took the Mead/Conway LSI design
course and ended up abandoning his research on portable compilers in
favor of simulating LSI designs. He was also a co-founder of Symbolics
and designed the controller for their laser printer before returning to
MIT as a Lecturer and sponsored research staff.
There were also proposed follow-on software projects related to the Nu
terminal. One was Trix. Steve Ward said he didn't know what an "ics"
was, but Multics clearly had too many, and Unix had too few, hence Trix.
Jack Test was hired to do a lot of the development. Wikipedia has a
reasonable summary of Trix, as far as I remember, but I had left RTS to
join Masscomp in late 1981/early 1982, and I know Jack Test was an early
employee of Alliant Computer so he left Trix probably in 1982.
Chris was part of the Nu project.
"Was a principal developer of the NuMachine"
"developed a family of portable C compilers for the (then) newly
available microprocessors. These compilers were widely distributed as
the first C implementations for the x86 and 68K processors."
https://people.csail.mit.edu/cjt/resume.html
I found most of the yearly LCS reports have been digitized to DTIC
which answered a bunch of my questions about who was doing what at
that time
I've archived them at http://bitsavers.org/pdf/mit/lcs/progress_reports
3:15 p.m.
New subject: PC/IP
Thank you, Tom, for the definitive answers to much of this. I remembered
that the Z8000 was mixed up in that mess, but it was missing from Al's Trix
tape. Do you know if a Z8000 back end or set of support tools was ever
built, and if so, does anyone know if they survived? It does look like Al
has 8086 [Terman compiler]. 68K (of a few flavors) and an NS16032
(author's unknown). One of the tools you mentioned from MIT seems to have
survived, although Dennis and I saved the official UNIX Circuit Design
System release in the mid-1990s. Warren has had that TUHS archives ever
since, but I'm not I ever saw you tools other than things like you 68K
assembler, and I guess is was our fiiend Wayne that wrote the linker (which
until this thread I did not now).
BTW: Again, it proves how interwoven the people and tech (i.e., open source
culture) were in the 1970s; i.e., it's not a new thing. The PDPs were
running the Stanford Circuit Design System (SUDS) and the 11's often
at USCD. The people came and went. For instance,the former Wayne was a
year ahead of me at CMU before he headed to MIT for a Master's and PhD,
ᐧ
On Fri, May 2, 2025 at 10:57 AM Tom Teixeira <tjteixeira(a)earthlink.net>
wrote:
On 5/1/25 12:20 AM, Al Kossow wrote:
On 4/30/25 8:38 PM, Jonathan Gray wrote:
Some background, though the MIT LCS progress reports should cover much
of this. I won't attempt to put any dates.
Chris Terman was one of the graduate students in the RTS group. Since
VT-52 terminals were relatively scarce, he designed and built his own
with a larger screen - something like 40 lines by maybe 120 or 132
characters, called the "Termanal". I don't remember if it used an 8080
to handle the control sequences in the data stream or something else.
He then got interested in designing a terminal that could display bit
map graphics, to be comparable to the graphics used on the Lisp Machines
just being built by the MIT-AI lab. I had stumbled across one of the LCS
progress reports that credits Professor Steve Ward and one of the
undergraduate staff, Rae McClellan in assisting the design of this bit
graph which was named the "Nu Terminal" (I don't think it was the "Nu
Termanal"). This used an 8086. A couple of these were built. One of the
undergraduate students, Jon Sieber, had been a member of an Explorer
Post in Murray Hill where Dennis Ritchie was the advisor. Jon would
regularly bring UNIX tapes from the Research Lab and included things
like early versions of the Portable C Compiler and the Circuit Design
Aids. Chris used the Circuit Design Aids to design wire-wrap boards for
the Nu Terminal and the RTS lab got a semi-automatic wire wrap machine.
Some students and staff took turns doing the actual wire wrapping. My
contribution was writing some simple software that simulated a paper
tape reader for the wire wrap machine.
An undergraduate student, Mike Patrick, did his bachelor's thesis
writing a table driven assembler and constructed tables for the 8086 and
I think an 8080. Later there were drivers for the Zylog Z8000, the
National Semiconductor NS16000 and the Motorola 68000. I contributed a
small bit of code for doing optimal choice of short vs long branches (to
branch to an address more than +/- 127 bytes, you had to branch around a
longer jump instruction).
Chris Terman did the work of modifying the Portable C Compiler to
generate code for the 8086, the Z8000, NS16000 and MC68000. I think we
may have built one machine with the Z8000, but quickly settled on using
the MC68000, primarily because of the 32-bit support (one progress
report says that Zenith was supposed to build multiple Z8000 based
machines, but I don't remember those. The NS16000 had better memory
management, but I don't think we ever actually received any CPU chips.
Anyway, these compilers were what was distributed, and the MC68000
compiler in particular was used by almost all the companies that came
out the MC68000-based Unix machines. Apollo was a notable exception, but
Apollo wrote their own operating system from scratch rather than Unix.
Side note: Bill Poduska came to visit Steve Ward and before the visit
Steve was all excited, but was disappointed that Bill was not going to
use Unix.
Before the RTS group used Unix, they had written a small timesharing
system for the PDP-11/45 that was used in the 6.031 introductory
computer science course taught by Mike Dertouzos. Chris was involved in
maintaining that, though I think Steve Ward was probably the main
implementor. Chris had also spent too many hours changing address
jumpers on Unibus and other controllers as well as tweaking Unix mkconf
files, and thought that while the 4BSD autoconfiguration was an
improvement, there should be a better way. Chris and Steve designed the
Nu bus, and the Nu Bus was used in the MC68000 boards. Eventually it was
picked up by Apple.
Chris was one of many students who took the Mead/Conway LSI design
course and ended up abandoning his research on portable compilers in
favor of simulating LSI designs. He was also a co-founder of Symbolics
and designed the controller for their laser printer before returning to
MIT as a Lecturer and sponsored research staff.
There were also proposed follow-on software projects related to the Nu
terminal. One was Trix. Steve Ward said he didn't know what an "ics"
was, but Multics clearly had too many, and Unix had too few, hence Trix.
Jack Test was hired to do a lot of the development. Wikipedia has a
reasonable summary of Trix, as far as I remember, but I had left RTS to
join Masscomp in late 1981/early 1982, and I know Jack Test was an early
employee of Alliant Computer so he left Trix probably in 1982.
Chris was part of the Nu project.
"Was a principal developer of the NuMachine"
"developed a family of portable C compilers for the (then) newly
available microprocessors. These compilers were widely distributed as
the first C implementations for the x86 and 68K processors."
https://people.csail.mit.edu/cjt/resume.html
I found most of the yearly LCS reports have been digitized to DTIC
which answered a bunch of my questions about who was doing what at
that time
I've archived them at http://bitsavers.org/pdf/mit/lcs/progress_reports
3:28 p.m.
New subject: PC/IP
On 5/2/25 8:15 AM, Clem Cole wrote:
Thank you, Tom, for the definitive answers to much of
this.
This is a huge help to me. I came to Apple just before the Mac II was released and know
all of the people that were involved with designing our PC version of Nubus. One of the
people was Tony Masterson who came to Apple from TI, and I think I remember him telling
me he was a MIT grad student. I've traced the path from Western Digital to TI for
Nubus
George White seems to be the MIT - Western Digital connection.
http://bitsavers.org/pdf/ti/nubus/White_-_Nubus_Computer_Design_19840615.pdf
Getting back to Unix, TI used Nubus to build the Explorer Lisp machines, with roots
back to the MIT CADR based on the Nu Machine (aka TI 1500 Unix system). HP bought
the 1500 product line, which was then EOL'ed.
3 May
3 May
9:16 a.m.
New subject: PC/IP
On Fri, May 02, 2025 at 08:28:03AM -0700, Al Kossow wrote:
On 5/2/25 8:15 AM, Clem Cole wrote:
James Gula was another person who moved from MIT to WD to TI.
"Jim Gula has left the LCS staff to join WD; while we regret losing him,
we feel that he will serve important roles both in the MIT/WD interface
and in making the Nu into a successful product."
LCS Progress Report 18, July 1980 - June 1981, p 212
"In December, it was concluded that the research interests of the TRIX
project were best served by a re-engineering of its internal structure
rather than by polishing of the existing implementation. In order to
reconcile this plan with the need for a viable Nu programming
environment for other projects, Version 7 UNIX was ported to the Nu
during the Spring by Gula, Sieber, Terman, and Test."
LCS Progress Report 18, July 1980 - June 1981, p 213
"John Seamons of Lucasfilm brought up Jim Gula's MIT Nu Unix on the Sun.
We have an Ethernet based version of this Unix running at Stanford"
Vaughan Pratt on fa.works, Jan 1982
https://groups.google.com/g/fa.works/c/WHpSvlbG0A8/m/IUdSUIwJqAgJ
https://www.saildart.org/WORKS.MSG[UP,DOC]28
James Gula on LinkedIn:
Engineering Manager, Texas Instruments, 1983 - 1986
Engineering manager for the Nu Machine Project
Software Engineering Manager, Western Digital, 1981 - 1983
Managed the UNIX port to the Nu Machine project
Technical Staff, MIT LCS, 1979 - 1981
Port of UNIX to the Nu Machine and related software.
Thank you, Tom, for the definitive answers to
much of this.
This is a huge help to me. I came to Apple just before the Mac II was released and know
all of the people that were involved with designing our PC version of Nubus. One of the
people was Tony Masterson who came to Apple from TI, and I think I remember him telling
me he was a MIT grad student. I've traced the path from Western Digital to TI for
Nubus
George White seems to be the MIT - Western Digital connection.
http://bitsavers.org/pdf/ti/nubus/White_-_Nubus_Computer_Design_19840615.pdf
6:29 p.m.
New subject: PC/IP
"John Seamons of Lucasfilm brought up Jim
Gula's MIT Nu Unix on the Sun.
We have an Ethernet based version of this Unix running at Stanford"
Vaughan Pratt on fa.works, Jan 1982
https://groups.google.com/g/fa.works/c/WHpSvlbG0A8/m/IUdSUIwJqAgJ
https://www.saildart.org/WORKS.MSG[UP,DOC]28
Since I've never seen the design for the original Nu 68K I wonder if
that was where the "SUN" segment/page MMU came from, since it looked
similar to what was in the CADR, and if the sniffing of the stack to
see if it needed to grow on function calls came from the Terman compiler.
7:01 p.m.
New subject: PC/IP
There were a couple of solutions, but they were all similar, which might be
challenging to figure out now. The first gen of the the micros needed an
external MMU and by the time of the Z8000/68K family/NS16032/and even
the Intel devices were already several different MMU from mini's and
mainframes which the different microprocessor MMU swiped different ideas
and added a few other there own (particular to them). I think poking
around the Asilimor Workshop Archives is likely to be the most fruitful. I
would love to find a copy of Forest Basket's paper where he proposed using
2 68000's as 'executor' and 'fixer' as Apollo and Masscomp would
do [many
of those were from Asilomar - Forest's was]. Yale Patt and a few of his
students had a few MMU papers for some of the chips around that time,
IIRC. Sadly, my copies of that stuff from a few of those Asilomar
conferences were lost.
ᐧ
On Sat, May 3, 2025 at 2:29 PM Al Kossow <aek(a)bitsavers.org> wrote:
"John Seamons of Lucasfilm brought up Jim
Gula's MIT Nu Unix on the Sun.
We have an Ethernet based version of this Unix running at Stanford"
Vaughan Pratt on fa.works, Jan 1982
https://groups.google.com/g/fa.works/c/WHpSvlbG0A8/m/IUdSUIwJqAgJ
https://www.saildart.org/WORKS.MSG[UP,DOC]28
Since I've never seen the design for the original Nu 68K I wonder if
that was where the "SUN" segment/page MMU came from, since it looked
similar to what was in the CADR, and if the sniffing of the stack to
see if it needed to grow on function calls came from the Terman compiler.
7:14 p.m.
New subject: PC/IP
On 5/3/25 12:01 PM, Clem Cole wrote:
There were a couple of solutions, but they were all
similar, which might be challenging to figure out now. The first gen of the the micros
needed an external MMU and by the time of the Z8000/68K family/NS16032/and even the Intel
devices were already several different MMU from
mini's and mainframes which the different microprocessor MMU swiped different ideas
and added a few other there own (particular to them). I
think poking around the Asilimor Workshop Archives is likely to be the most fruitful. I
would love to find a copy of Forest Basket's paper
where he proposed using 2 68000's as 'executor' and 'fixer' as
Apollo and Masscomp would do [many of those were from Asilomar - Forest's
was]. Yale Patt and a few of his students had a few MMU papers for some of the chips
around that time, IIRC. Sadly, my copies of that
stuff from a few of those Asilomar conferences were lost.
I hope the scanning and preservation efforts of the past 20 years continues.
As a naive engineer coming to the Valley in 1984 with my only prior exposure
being Usenet, the degree of tribal knowledge and interconnectedness took
a while to realize. I'm still learning things, for example what is coming
to light here, even after working at the Computer History Museum for almost
20 years now along with realizing how much of Valley folklore we don't have
in our archives.
4 May
4 May
6:48 a.m.
New subject: PC/IP
On Sat, May 03, 2025 at 11:29:14AM -0700, Al Kossow wrote:
Perhaps the Nu paper mentions an MMU?
Referenced by the SUN documents, but not available online.
S. A. Ward and C. J. Terman, "An approach to personal computing"
Proceedings of COMPCON, IEEE, February 1980, pp. 460-465.
IEEE doesn't have it online at
https://www.computer.org/csdl/proceedings/1000109
The CHM has the proceedings:
https://www.computerhistory.org/collections/catalog/102713995
CompCon80; VLSI: New Architectural Horizons; 1980.
--
"The one question mark for the CPU board design was the memory
management unit, to which Andy, Forest (who sent us a design while at
PARC), and I all made significant contributions."
Vaughan Pratt, in:
From the Valley of Heart's Delight to the Silicon Valley:
A Study of Stanford University's Role in the Transformation
Appendix A: The Founding of Sun Microsystems
http://infolab.stanford.edu/pub/cstr/reports/csl/tr/97/713/CSL-TR-97-713.ps
"So we switched to the 68000. I think we ended up with a 6810 as the
original processor. I designed a memory-mapping system for that
processor, which barely had the capabilities to do memory mapping. The
original SUN workstation had a really fascinating memory mapping system.
I was a little annoyed with Andy, because years later I discovered that
he had patented that memory mapping system."
Oral History of Forest Baskett, p 13
https://archive.computerhistory.org/resources/access/text/2017/12/102717243…
https://patents.google.com/patent/US4527232A/en
https://patents.google.com/patent/US4550368A/en
"John Seamons of Lucasfilm brought up Jim
Gula's MIT Nu Unix on the Sun.
We have an Ethernet based version of this Unix running at Stanford"
Vaughan Pratt on fa.works, Jan 1982
https://groups.google.com/g/fa.works/c/WHpSvlbG0A8/m/IUdSUIwJqAgJ
https://www.saildart.org/WORKS.MSG[UP,DOC]28
Since I've never seen the design for the original Nu 68K I wonder if
that was where the "SUN" segment/page MMU came from, since it looked
similar to what was in the CADR, and if the sniffing of the stack to
see if it needed to grow on function calls came from the Terman compiler.
3 May
3 May
1:49 p.m.
New subject: PC/IP
On 5/2/25 11:15 AM, Clem Cole wrote:
Thank you, Tom, for the definitive answers to much of
this. I
remembered that the Z8000 was mixed up in that mess, but it was
missing from Al's Trix tape. Do you know if a Z8000 back end or set
of support tools was ever built, and if so, does anyone know if they
survived? It does look like Al has 8086 [Terman compiler]. 68K (of a
few flavors) and an NS16032 (author's unknown). One of the tools you
mentioned from MIT seems to have survived, although Dennis and I saved
the official UNIX Circuit Design System release in the
mid-1990s. Warren has had that TUHS archives ever since, but I'm not I
ever saw you tools other than things like you 68K assembler, and I
guess is was our fiiend Wayne that wrote the linker (which until this
thread I did not now).
BTW: Again, it proves how interwoven the people and tech (i.e., open
source culture) were in the 1970s; i.e., it's not a new thing. The
PDPs were running the Stanford Circuit Design System (SUDS) and the
11's often at USCD. The people came and went. For instance,the
former Wayne was a year ahead of me at CMU before he headed to MIT for
a Master's and PhD,
ᐧ
I'm pretty sure a Z8000 back end was produced because I remember that we
built at least one Z8000 board. One of the LCS progress reports mentions
that Zenith had committed to build some Z8000 systems, back when "office
automation" was a thing. However, I have no idea what happened to the tools.
I don't remember anyone but Chris producing back ends, but it's possible
someone else did the NS16032. But I don't remember anything else about
the NS16000 systems.
The tools Chris and others produced in support of the Mead/Conway LSI
course were also widely distributed, but I'm not sure what the
mechanism. Since those were completely unencumbered by Unix, there was
probably less formality, but I expect the MIT license was included.
19
days inactive
23
days old
16 comments
7 participants
participants (7)
-
Al Kossow -
Clem Cole -
jnc＠mercury.lcs.mit.edu -
Jonathan Gray -
Lawrence Stewart -
Steve Nickolas -
Tom Teixeira