suppose TOPS-20 had never happened
Mark Crispin
The KL10 never would have gotten the good pager; it would have only had
the crufty KI10 style memory map whose main function was to avoid The Big
BLT of the KA10 (a.k.a. shuffling).
The PDP-10 would have been cancelled not long after the axing of the
PDP-15 and introduction of VAX in 1977. VAX and PDP-10 would not have
overlapped by more than a couple of years.
VAX 8600 would have happened years sooner, and would have killed the KL10
if it wasn't already dead.
Let's not forget that VMS was technologically superior to TOPS-10 and ran
on much cheaper hardware. VAX had far more expansion potential with 4GB
address space compared to the 1MB on the non-extended PDP-10.
TOPS-10 SMP would certainly not have happened. Without TOPS-20, the
PDP-10 would have been unambiguously a technological dead end and not
worth throwing good money after bad. The sensible thing would be to buy
more VAXen.
TOPS-10 and the PDP-10 would have been utterly forgotten by the mid 1980s,
other than in similar contexts to RSX-11 and RSTS; the kludgy old system
we had before we upgraded to VMS.
Nobody would have developed Common Lisp or C for the PDP-10. EMACS would
have languished in obscurity.
UNIX never would have acquired important technology elements, particularly
those we associate with BSD.
IMAP would not have had shared-access semantics; it would have had the
same single-access lock requirement as POP3. Shared access to mail was
something that you did on TOPS-20, not on UNIX.
I, for one, am glad that TOPS-20 happened; even if that feeling is not
universal.
-- Mark --
http://staff.washington.edu/mrc
Science does not emerge from voting, party politics, or public debate.
Si vis pacem, para bellum.
John Sauter
> Suppose that TOPS-20 had never happened.
>
> Nobody would have developed Common Lisp or C for the PDP-10. EMACS
> would have languished in obscurity.
>
Mark, I disagree with some of your predictions, but most are just my
opinion versus yours. In one case, though, I have support for my
contrary opinion: Common Lisp.
One of the contributors to Common Lisp was Tony Hearn, see
<http://www8.informatik.uni-erlangen.de/html/lisp/lisppeople.html>.
He started doing Lisp work on the PDP-6, and by the time I left Stanford
he was already thinking that there should be a standard syntax. If Guy
Steele had not chaired the committee, Tony certainly could have, and he
was not motivated by TOPS-20 for the very good reason that it did not
exist in the late 1960s.
John Sauter (John_...@systemeyescomputerstore.com)
jmfb...@aol.com
Most of the above is wrong. I haven't thought about the last two items.
>
>
>I, for one, am glad that TOPS-20 happened; even if that feeling is not
>universal.
Why do you keep interpreting what I write as a hate-tops20 opinion?
I do not hate TOPS-20. Nor do I love it. I do not even love TOPS-10.
They were a piece of software that sold hardware which provided
computing needs which our customers were willing to pay for.
PDP-11s also provided computing services and they also needed
software to augment the customer's requirements.
/BAH
John Everett
>I do not hate TOPS-20. Nor do I love it. I do not even love TOPS-10.
Some good advice I received many years ago:
"Never fall in love with an inanimate object; it won't love you back."
;-)
--
jeverett3<AT>sbcglobal<DOT>net (John V. Everett)
Sarr J. Blumson
: Suppose that TOPS-20 had never happened.
...
: The PDP-10 would have been cancelled not long after the axing of the
: PDP-15 and introduction of VAX in 1977. VAX and PDP-10 would not have
: overlapped by more than a couple of years.
I certainly don't want to get into a "where was DEC selling most of the
10s" discussion - I don't have a clue for one thing :-) - but ADP was
certainly buying significant quantities of both KLs and KSs right up to
the bitter end and running TOPS-10 on both of them. To the best of my
knowledge we never looked at TOPS-20 at all, and started looking at VAXen
only after Jupiter died/was murdered.
...
: Nobody would have developed Common Lisp or C for the PDP-10. EMACS would
: have languished in obscurity.
A minor point, but we did C for TOPS-10, too.
...
: I, for one, am glad that TOPS-20 happened; even if that feeling is not
: universal.
A separate question. I'm glad the Tenex happened. I'm glad that TOPS-20
happened. I'm glad that TOPS-10 happened. I can't think of anything that
happened (in this subject area) that I'm not glad happened. Even COBOL. :-)
--
--------
Sarr Blumson sarr.b...@alum.dartmouth.org
http://www-personal.umich.edu/~sarr/
fishtop_records
wrote:
> I can't think of anything that
> happened (in this subject area) that I'm not glad happened. Even COBOL. :-)
My favorite all time SPR answer was about Cobol on the 20. (Did anyone
run Cobol on the 10? I never did)
We found a non-trivial compiler bug, I hacked the source down to one
line, something roughly like
compute A time B plus 42 divided by 17 giving answer.
The response, published as usual, said:
"The problem with your program is that you are using Cobol."
Signed as all Cobol stuff was by Nixon, who seemed forever cursed to
support Cobol-10 (it was not close to native Jsys)
because he did it once.
William Schaub
*snip*
>
>: Nobody would have developed Common Lisp or C for the PDP-10. EMACS would
>: have languished in obscurity.
>
> A minor point, but we did C for TOPS-10, too.
>
Does a C compiler still exist some place for TOPS-10?
If so I would love to get my hands on it.
Mark Crispin
> Mark, I disagree with some of your predictions, but most are just my
> opinion versus yours. In one case, though, I have support for my
> contrary opinion: Common Lisp.
The first Common Lisp specification was not finalized until 1984. It was
a long and extremely drawn-out process.
The only Common Lisp implementation that I know of for the PDP-10 was not
written until 1985. It ran in extended addressing, and immediately after
starting consumed 1023 memory pages.
By that time, the PDP-10 would have been long-forgotten history, and
without TOPS-20 there certainly would not have been extended addressing.
I doubt very much that Tony Hearn would have wasted his time implementing
Common Lisp on a dead architecture. He, like everybody else, would have
focused on VMS and/or UNIX.
As it was, it was due to the dedication of Chuck Hedrick, that there was a
Common Lisp for the PDP-10 at all. Sadly, it was too late to have much of
an impact; most of the Lispers were already gone. The few that remained
on PDP-10s used the legacy Lisps (most notably, MacLisp and Interlisp) and
were only the PDP-10 only because they were using legacy Lisps.
fishtop_records
know that I would have not studied Operating Systems in grad school
from Peter Denning. His thesis on working sets drove Tenex and it
begat Tops-20. When I had experience and heard that Dr Denning was
teaching in my area, I signed up.
I do tend to believe that without Tops-20, LCG would have died much
earlier. I have no idea how much earlier. Its not clear to me that the
Vax 8600 would have be built at all if it wasn't a last gasp attempt
by LCG to convince customers that there was a future with Dec.
But its hard to tell of some other vendor would have implemented
working set virtual memory. Perhaps if they had, Cutler would not have
become a cult figure, leading to his polluting NT with lame ideas from
RSX -- the ideas that curse us today, long after all TOPS flavors had
died.
I don't know, did other firms have the concept of a shell separate
from the OS? Clearly Exec was a big win, rather than the compil
binding that was part of TOPS-10.
Mark Crispin
> On the prime question, I have no idea what would have happened. I do
> know that I would have not studied Operating Systems in grad school
> from Peter Denning. His thesis on working sets drove Tenex and it
> begat Tops-20. When I had experience and heard that Dr Denning was
> teaching in my area, I signed up.
Yes. Denning is a god.
> I do tend to believe that without Tops-20, LCG would have died much
> earlier. I have no idea how much earlier.
I would say before 1980.
> Its not clear to me that the
> Vax 8600 would have be built at all if it wasn't a last gasp attempt
> by LCG to convince customers that there was a future with Dec.
I disagree. A big VAX was needed, and although the VAX 8600 wasn't big
enough it served for that purpose. LCG did its best to kill the 8600
(Venus) they way they did the 4050 (Jupiter). I heard tell of what Kotok
had to do to save Venus.
> But its hard to tell of some other vendor would have implemented
> working set virtual memory.
BSD UNIX was well underway by the late 1970s. The BSD/TOPS-20 fusion that
we know as modern-day BSD UNIX (and, to some degree, Linux) was later.
> Perhaps if they had, Cutler would not have
> become a cult figure, leading to his polluting NT with lame ideas from
> RSX -- the ideas that curse us today, long after all TOPS flavors had
> died.
Far be it from me to defend Cutler and his ideas; but NT isn't that bad.
You have to dig vigorously and deep, but underneath all that is a real
operating system.
> I don't know, did other firms have the concept of a shell separate
> from the OS? Clearly Exec was a big win, rather than the compil
> binding that was part of TOPS-10.
ITS, Multics, and later UNIX all had shells. Shells were pretty much
standard in 1970s operating system design. A kernel-based command decoder
(such as TOPS-10, OS/8, etc.) was 1960s design.
I'm not familiar enough with these to be certain, but I'm pretty sure that
at least some of the PDP-11 operating systems had shells. Certainly VMS
did.
The Tenex/TOPS-20 EXEC was certainly a big win, but it was still quite
weak compared to the UNIX shell. A fair categorization of UNIX vs.
TOPS-20 was that UNIX was an incredibly weak kernel with an incredibly
powerful shell, whereas TOPS-20 was an incredibly powerful kernel with an
incredibly weak shell.
By the way, DEC did not invent COMPIL. For extra credit, who did, and
what was it called?
Eric Smith
> I'm not familiar enough with these to be certain, but I'm pretty sure
> that at least some of the PDP-11 operating systems had shells.
RSTS/E does. I'm not sure about the others.
Al Kossow
> I don't know, did other firms have the concept of a shell separate
> from the OS?
Genie (SDS-930 timesharing system at Cal circa 1965 that became
the SDS-940) has it.
BBN had 940 systems prior to Tenex
--
Posted via a free Usenet account from http://www.teranews.com
glen herrmannsfeldt
(snip)
> My favorite all time SPR answer was about Cobol on the 20. (Did anyone
> run Cobol on the 10? I never did)
I didn't, but I used to use the COBOL compiler for tape testing.
It was the largest file I knew about, so a few copies onto a
tape, and then read them back, was a tape test.
> We found a non-trivial compiler bug, I hacked the source down to one
> line, something roughly like
> compute A time B plus 42 divided by 17 giving answer.
> The response, published as usual, said:
> "The problem with your program is that you are using Cobol."
This was only a story by the time I got to using TOPS-10, but
supposedly someone sent in a bug report about using QUEUE to
run commands while not logged on. (If I remember, it is the
RUN option of QUEUE.) The SPR form has an option for
publish/don't publish and the person sending it in
decided to check publish (ha ha, they wouldn't do that).
It seems that it did get published before anyone got around
to checking what the problem actually was, resulting in
a very fast bug fix!
If documentation still exists, it might be better than
the fourth hand story as I remember it.
-- glen
glen herrmannsfeldt
(snip)
>> Its not clear to me that the
>> Vax 8600 would have be built at all if it wasn't a last gasp attempt
>> by LCG to convince customers that there was a future with Dec.
> I disagree. A big VAX was needed, and although the VAX 8600 wasn't big
> enough it served for that purpose. LCG did its best to kill the 8600
> (Venus) they way they did the 4050 (Jupiter). I heard tell of what
> Kotok had to do to save Venus.
Was it that DEC/Digital didn't understand how to be a big
computer company? As I understand it, DEC started when IBM meant
computer, and that usually meant big. DEC could sell small computers
cheaper, easier to use, that could be used to do some of what
IBM would otherwise do.
Then IBM started selling smaller systems, and DEC started selling
larger systems until they were direct competitors.
(snip)
>> I don't know, did other firms have the concept of a shell separate
>> from the OS? Clearly Exec was a big win, rather than the compil
>> binding that was part of TOPS-10.
> ITS, Multics, and later UNIX all had shells. Shells were pretty much
> standard in 1970s operating system design. A kernel-based command
> decoder (such as TOPS-10, OS/8, etc.) was 1960s design.
Then there was IBM's VM where not only the command interpreter but
just about everything else was done in user space. Write your
own OS and IPL it into your virtual machine.
-- glen
Morten Reistad
fishtop_records <pfar...@pfarrell.com> wrote:
>On the prime question, I have no idea what would have happened. I do
>know that I would have not studied Operating Systems in grad school
>from Peter Denning. His thesis on working sets drove Tenex and it
>begat Tops-20. When I had experience and heard that Dr Denning was
>teaching in my area, I signed up.
If this !Tops20 world also removes this paper from wide knowledge,
then Linux would have been strongly affected. The successors to this
paper strongly affected the initial Linux internals. Otherwise Linus
may have listened to Tanenbaum and done something like QNX or Mach.
>I do tend to believe that without Tops-20, LCG would have died much
>earlier. I have no idea how much earlier. Its not clear to me that the
>Vax 8600 would have be built at all if it wasn't a last gasp attempt
>by LCG to convince customers that there was a future with Dec.
>
>But its hard to tell of some other vendor would have implemented
>working set virtual memory. Perhaps if they had, Cutler would not have
>become a cult figure, leading to his polluting NT with lame ideas from
>RSX -- the ideas that curse us today, long after all TOPS flavors had
>died.
The two other scalable architectures could have won out; Multics style
segmentation (like in primos) or message passing between virtual machines
(VM/CMS, Mach, to some extent QNX).
>I don't know, did other firms have the concept of a shell separate
>from the OS? Clearly Exec was a big win, rather than the compil
>binding that was part of TOPS-10.
Multics has a separate shell, but it is implemented differently.
It is a stack invocation, not a process.
-- mrr
fishtop_records
> Was it that DEC/Digital didn't understand how to be a big
> computer company?
Word! They had no clue. Rumour was than whenever KO thought they would
compete with IBM, he would kill the project.
Its only urban legend, but it rings true.
> Then there was IBM's VM where not only the command interpreter but
> just about everything else was done in user space. Write your
> own OS and IPL it into your virtual machine.
Yes, AMS was mostly an IBM shop, they bought TOPS-20 systems to
compete in delivering interactive systems.
The IBM fanboys argued that VM/CMS was as good as TOPS-20, that IPLing
your own OS was a great thing, etc.
But, when it came to developing solution for paying customers, the
TOPS-20 folks delivered a lot more in less time for less money.
There never was a big market for VM/CMS, altho we had the IBM gear to
support it.
fishtop_records
> In article <be635a4d-0c8f-4ef8-9fbe-8cd382125...@o6g2000hsd.googlegroups.com>,
>
> fishtop_records <pfarr...@pfarrell.com> wrote:
> >On the prime question, I have no idea what would have happened. I do
> >know that I would have not studied Operating Systems in grad school
> >from Peter Denning. His thesis on working sets drove Tenex and it
> >begat Tops-20. When I had experience and heard that Dr Denning was
> >teaching in my area, I signed up.
>
> If this !Tops20 world also removes this paper from wide knowledge,
> then Linux would have been strongly affected. The successors to this
> paper strongly affected the initial Linux internals. Otherwise Linus
> may have listened to Tanenbaum and done something like QNX or Mach.
I really didn't mean !tops20 would evaporate the paper, but working on
TOPS-20 for years sure made me aware of it.
Which is why I took CS 501 or whatever Operating Systems was from
Denning.
Of course Cutler, who worked for DEC, never showed that he read it.
> The two other scalable architectures could have won out; Multics style
> segmentation (like in primos) or message passing between virtual machines
> (VM/CMS, Mach, to some extent QNX).
>
> >I don't know, did other firms have the concept of a shell separate
> >from the OS? Clearly Exec was a big win, rather than the compil
> >binding that was part of TOPS-10.
>
> Multics has a separate shell, but it is implemented differently.
> It is a stack invocation, not a process.
I worked with Tom Van Vleck, a keeper of the Multic flame, and chatted
recently when MIT released the source code.
While the Honeywell 36 bit machines had some similarities to TOPS-20,
it sounds like it was too weird to be ported to something like SIMH
fishtop_records
> > But its hard to tell of some other vendor would have implemented
> > working set virtual memory.
>
> BSD UNIX was well underway by the late 1970s. The BSD/TOPS-20 fusion that
> we know as modern-day BSD UNIX (and, to some degree, Linux) was later.
Did any of them have the microcode assist to keep the 'in use' bit of
the page table supported? or do they do that as sepate code
in the kernel instructions?
(I have to admit I've not even looked inside a kernel in 15 years.)
jmfb...@aol.com
John Everett <jeve...@sbcglobal.DEFEAT.UCE.BOTS.net> wrote:
>On Fri, 30 Nov 07 13:00:32 GMT, jmfb...@aol.com wrote:
>
>>I do not hate TOPS-20. Nor do I love it. I do not even love TOPS-10.
>
>Some good advice I received many years ago:
>
>"Never fall in love with an inanimate object; it won't love you back."
>
>;-)
<grin> Right.
/BAH
jmfb...@aol.com
<96d33358-c7e8-4d4c...@p69g2000hsa.googlegroups.com>,
fishtop_records <pfar...@pfarrell.com> wrote:
>On Nov 30, 10:46 am, "Sarr J. Blumson" <s...@rygar.gpcc.itd.umich.edu>
>wrote:
>> I can't think of anything that
>> happened (in this subject area) that I'm not glad happened. Even COBOL. :-)
>
>My favorite all time SPR answer was about Cobol on the 20. (Did anyone
>run Cobol on the 10? I never did)
Yes.
>
>We found a non-trivial compiler bug, I hacked the source down to one
>line, something roughly like
> compute A time B plus 42 divided by 17 giving answer.
>
>The response, published as usual, said:
>
>"The problem with your program is that you are using Cobol."
And the implied end of that sentence was "...on a -20." :-)
>
>Signed as all Cobol stuff was by Nixon, who seemed forever cursed to
>support Cobol-10 (it was not close to native Jsys)
>because he did it once.
He didn't usually do the compiler work. He wrote SORT and
was given the ISAM problems.
/BAH
jmfb...@aol.com
fishtop_records <pfar...@pfarrell.com> wrote:
>know that I would have not studied Operating Systems in grad school
>from Peter Denning. His thesis on working sets drove Tenex and it
>begat Tops-20. When I had experience and heard that Dr Denning was
>teaching in my area, I signed up.
>
>I do tend to believe that without Tops-20, LCG would have died much
>earlier.
No, it wouldn't. You people are concentrating on software. DEC
was in the hardware business, not the software business.
>I have no idea how much earlier. Its not clear to me that the
>Vax 8600 would have be built at all if it wasn't a last gasp attempt
>by LCG to convince customers that there was a future with Dec.
LCG was long gone by the time the 8600 was started.
>
>But its hard to tell of some other vendor would have implemented
>working set virtual memory. Perhaps if they had, Cutler would not have
>become a cult figure, leading to his polluting NT with lame ideas from
>RSX -- the ideas that curse us today, long after all TOPS flavors had
>died.
>
>I don't know, did other firms have the concept of a shell separate
>from the OS? Clearly Exec was a big win, rather than the compil
>binding that was part of TOPS-10.
/BAH
jmfb...@aol.com
I remember talking about the bug. Wasn't that the MPB QUEUE?
The real fix was writing GALAXY. So the time frame we're talking
about would have been 1973 or 1974.
/BAH
Al Dykes
fishtop_records <pfar...@pfarrell.com> wrote:
I got to know Mr. Nixon via the compute verb, also.
At BigBank, we bought a major general ledger package that ran on lots
of platforms. It was written in Cobol with lots of compute statements.
DEC's cobol generated single-precision math and there weren't enough
bits to represent our bottom-line, circa 1980. The error was in the
low two digits (pennies). The GL vendor had never seen this before.
Eventually we identified the cause and wrote an SPR. The response
acknowledged the incompatibility and said it would be "fixed in next
release" with no date scheduled for that release. Unacceptable.
It was one of those times when it was good to work for a company so
large that we could "buy" our vendor.
We got our outside auditors to write a letter stating that because
they had to hand-fudge the annual report, they would have to put an
astrix on the number and a footnote blaming DEC for it.
It was then that I was put in touch with Mr. Nixon. He delivered a
patched compiler for us in short order.
fishtop_records
> In article <be635a4d-0c8f-4ef8-9fbe-8cd382125...@o6g2000hsd.googlegroups.com>,
> fishtop_records <pfarr...@pfarrell.com> wrote:
> >earlier.
>
> No, it wouldn't. You people are concentrating on software. DEC
> was in the hardware business, not the software business.
Interesting, in the VMS birthday thread, you said that G Bell and
others did not understand that
dec's software helped sell the hardware. While that doesn't put you in
the software business, its
kinda of like selling a car without an enginer.
> LCG was long gone by the time the 8600 was started.
That is not what LCG said at the time.
fishtop_records
> >The response, published as usual, said:
>
> >"The problem with your program is that you are using Cobol."
>
Right. Since neither the cobol compiler nor the runtime were ever
converted, they were pure UUO 10 code.
And it wasn't an IO problem, it was logic failure. Far as I knew, most
of the instructions functioned the same on a KL, no
matter if you have 20 or 10 flavor. A movem worked the same.
> >Signed as all Cobol stuff was by Nixon, who seemed forever cursed to
> >support Cobol-10 (it was not close to native Jsys)
> >because he did it once.
>
> He didn't usually do the compiler work. He wrote SORT and
> was given the ISAM problems.
I think it was more that no one did much, and he got stuck with it.
Rich Alderson
> Mark, I disagree with some of your predictions, but most are just my
> opinion versus yours. In one case, though, I have support for my
> contrary opinion: Common Lisp.
> One of the contributors to Common Lisp was Tony Hearn, see
> <http://www8.informatik.uni-erlangen.de/html/lisp/lisppeople.html>.
I remember Tony Hearn as the proponent of "Standard Lisp", which was used as
the underpinning for a mathematics package (REDUCE? I forget after 25 years or
so), and which was one of the minor influences on Common Lisp.
--
Rich Alderson "You get what anybody gets. You get a lifetime."
ne...@alderson.users.panix.com --Death, of the Endless
Christopher C. Stacy
> John Sauter <John_...@systemeyescomputerstore.com> writes:
>
>> Mark, I disagree with some of your predictions, but most are just my
>> opinion versus yours. In one case, though, I have support for my
>> contrary opinion: Common Lisp.
>
>> One of the contributors to Common Lisp was Tony Hearn, see
>> <http://www8.informatik.uni-erlangen.de/html/lisp/lisppeople.html>.
>
> I remember Tony Hearn as the proponent of "Standard Lisp", which was used as
> the underpinning for a mathematics package (REDUCE? I forget after 25 years or
> so), and which was one of the minor influences on Common Lisp.
Common Lisp is the common compromise of the MACLISP-descended dialects,
intended to defeat INTERLISP from becoming the mandary dialect for DOD
projects. It is essentially a stripped-down version of Lisp Machine Lisp,
with one adoption (lexical scoping) from Scheme, and major influence from
Xerox (LOOPS) in its object-oriented model. Common Lisp is what unified
MIT LispM Lisp with the VAX and SUN (Unix and VMS) MACLISP-based dialects.
The PDP-10 (and MACLISP itself) were almost entirely irrelevent at the point
when Common Lisp was begun, and 100% irrelevent by the time it was done.
The orignal "Common Lisp" was not a real standard, and I don't think that
anyone attempted to implement ANSI Common Lisp on the PDP-10. The Lisp
community had begun moving off of PDP-10s years before Common Lisp was begun,
Rutgers notwisthstanding.
jmfb...@aol.com
Good for you.
>
>It was then that I was put in touch with Mr. Nixon. He delivered a
>patched compiler for us in short order.
Yep. :-) He was a firefighter; a very good firefighter.
/BAH
jmfb...@aol.com
fishtop_records <pfar...@pfarrell.com> wrote:
>On Dec 1, 7:15 am, jmfbah...@aol.com wrote:
>> In article
<be635a4d-0c8f-4ef8-9fbe-8cd382125...@o6g2000hsd.googlegroups.com>,
>> fishtop_records <pfarr...@pfarrell.com> wrote:
>> >I do tend to believe that without Tops-20, LCG would have died much
>> >earlier.
>>
>> No, it wouldn't. You people are concentrating on software. DEC
>> was in the hardware business, not the software business.
>
>
>Interesting, in the VMS birthday thread, you said that G Bell and
>others did not understand that
>dec's software helped sell the hardware.
They didn't.
> While that doesn't put you in
>the software business, its
>kinda of like selling a car without an enginer.
I know. We have been talking about the situation as it was back
then, not what it should have been.
>
>> LCG was long gone by the time the 8600 was started.
>
>That is not what LCG said at the time.
It was not LCG. When the VAX development was started up in
Marlboro, LCG had been transferred to Central Engineering.
That was Bell's balliwick.
/BAH
jmfb...@aol.com
<521a25fb-f3a7-4585...@e23g2000prf.googlegroups.com>,
fishtop_records <pfar...@pfarrell.com> wrote:
>On Dec 1, 7:11 am, jmfbah...@aol.com wrote:
>> In article
>> <96d33358-c7e8-4d4c-a777-a7357b343...@p69g2000hsa.googlegroups.com>,
>> >The response, published as usual, said:
>>
>> >"The problem with your program is that you are using Cobol."
>>
>> And the implied end of that sentence was "...on a -20." :-)
>
>
>Right. Since neither the cobol compiler nor the runtime were ever
>converted, they were pure UUO 10 code.
>
>And it wasn't an IO problem, it was logic failure. Far as I knew, most
>of the instructions functioned the same on a KL, no
>matter if you have 20 or 10 flavor. A movem worked the same.
So far we've been talking about how head-wedged the OS development
groups were. We were perfect compared to the language groups :-)
>
>> >Signed as all Cobol stuff was by Nixon, who seemed forever cursed to
>> >support Cobol-10 (it was not close to native Jsys)
>> >because he did it once.
>>
>> He didn't usually do the compiler work. He wrote SORT and
>> was given the ISAM problems.
>
>I think it was more that no one did much, and he got stuck with it.
No. The problem became a firefight. It was escalated internally
to number one, top, ultimate priority. You, as a customer, could
have escalated it using the procedures that were in place but
you chose another method instead.
/BAH
John Sauter
>
> By the way, DEC did not invent COMPIL. For extra credit, who did, and
> what was it called?
>
> -- Mark --
Mark did not explicitly disallow me from answering, but I have waited
over 24 hours anyway, to give someone else a chance at the extra credit.
I cannot resist the urge to answer any longer.
COMPIL was invented by Bill Weir, a graduate student in the Stanford
Artificial Intelligence Project, in the late 1960s. It was called RPG,
which stood for Rapid Program Generation. The program was coded in
FAIL, a one-pass assembler written by Phil Petit, also a graduate
student in the Stanford AI Project who shared an office with Bill.
Bill requested assistance from two people. Dick Gruen, a DEC employee,
modified the MACRO assembler and the FORTRAN compiler to interface to
Bill's RPG. I modified the operating system to allow RPG to retrieve
the command line that invoked it.
When DEC made RPG part of the TOPS-10 product they changed the name to
COMPIL and made it assemble in the standard assembler, MACRO. Around
the same time Robert Clements rewrote the terminal services part of
TOPS-10, and included the ability to read the invoking command line in
his re-design.
The name was a joke, of course. RPG is the name of a well-known IBM
programming language, and therefore completely unsuitable for an
offering from DEC. In addition, Dick Gruen was also known as Richard P.
Gruen.
Bill Weir is one of the unrecognized pioneers of interactive computing.
In addition to COMPIL he also wrote a text editor, trying to make it
easier to use than TECO. His vision was that you should be able to
insert or delete a line of text anywhere in the file, without having to
explicitly write out the file and start the editing session over again
to access a page before the current page. In support of his vision he
asked me to add a feature to the file system: the ability to add a block
in the middle of a file. I said no: it was too hard. As a result, Bill
called his editor "stopgap -- works by recopying". The editor became
well-known under the same SOS, for Son of Stopgap.
After Bill left Stanford he went to work for one of the commercial
timesharing companies, perhaps Tymnet. I encountered him around 1975
when he came to Maynard to lecture on how he did long-distance
interactive terminals. I remember sitting in an office window on 5-5,
listening intently. The details escape me, but it was something
involving such meta-characters as red balls, green balls, and character
gobblers. His ideas were used in the design of the CTERM protocol, I
believe.
Does anyone know where Bill Wier is now?
John Sauter (John_...@systemeyescomputerstore.com)
Bob Clements
> My favorite all time SPR answer was about Cobol on the 20. (Did anyone
> run Cobol on the 10? I never did)
>
> We found a non-trivial compiler bug, I hacked the source down to one
> line, something roughly like
> compute A time B plus 42 divided by 17 giving answer.
>
> The response, published as usual, said:
>
> "The problem with your program is that you are using Cobol."
>
> Signed as all Cobol stuff was by Nixon, who seemed forever cursed to
> support Cobol-10 (it was not close to native Jsys)
> because he did it once.
I just talked to Dave on the phone. He says he doesn't remember that
SPR specifically, but it sounds like something he might have written.
He also thinks he would have produced the needed patch.
I don't remember him working on the Cobol compiler, but it was probably
after I left for BBN in January of 1972. He says he worked on Cobol for
about a year, prompted by some new government requirements (industry
wide) for Cobol. So he got drafted to implement whatever it was.
He is proud of the fact that, for some months, the DEC Cobol compiler
was the only one to pass that set of tests. Not even IBM's did.
Re: the comments on processor speed -- It wasn't just the late machines
that were unimpressive. During the brief time I was back in hardware
on the KI-10 design, I had to plead with the circuit guys and the logic
guys to stop blaming each other for the slow speed of the main register
shift loop and solve the problem. It was shifting at the same rate as
the KA-10, for crying out loud.
On another note, I haven't fired up a Usenet account in over five years.
I'm saddened to see how all of Usenet has deteriorated and how it is
being flooded by trash posts in every group I've looked at. Also, I'm
sad to see the flamewars in this group. But I guess that's life in the
21st century. I probably won't stay around long. The "from" address
on this post is not valid. If anyone wants to send me anything, I have
an account at gmail.com where the username is my initials, a dot, and
"ka10".
Best,
/Rcc
Lexington MA
DEC '64-'72, BBN '72-'03, currently probably retired
jmfb...@aol.com
Bob Clements <fake.a...@k1bc.com> wrote:
>fishtop_records wrote:
>
>> My favorite all time SPR answer was about Cobol on the 20. (Did anyone
>> run Cobol on the 10? I never did)
>>
>> We found a non-trivial compiler bug, I hacked the source down to one
>> line, something roughly like
>> compute A time B plus 42 divided by 17 giving answer.
>>
>> The response, published as usual, said:
>>
>> "The problem with your program is that you are using Cobol."
>>
>> Signed as all Cobol stuff was by Nixon, who seemed forever cursed to
>> support Cobol-10 (it was not close to native Jsys)
>> because he did it once.
>
>I just talked to Dave on the phone. He says he doesn't remember that
>SPR specifically, but it sounds like something he might have written.
>He also thinks he would have produced the needed patch.
It's good to hear that you both are active. 'ey, Bob.
>
>I don't remember him working on the Cobol compiler, but it was probably
>after I left for BBN in January of 1972. He says he worked on Cobol for
>about a year, prompted by some new government requirements (industry
>wide) for Cobol. So he got drafted to implement whatever it was.
>He is proud of the fact that, for some months, the DEC Cobol compiler
>was the only one to pass that set of tests. Not even IBM's did.
>
>
>Re: the comments on processor speed -- It wasn't just the late machines
>that were unimpressive. During the brief time I was back in hardware
>on the KI-10 design, I had to plead with the circuit guys and the logic
>guys to stop blaming each other for the slow speed of the main register
>shift loop and solve the problem. It was shifting at the same rate as
>the KA-10, for crying out loud.
>
>
>On another note, I haven't fired up a Usenet account in over five years.
>I'm saddened to see how all of Usenet has deteriorated and how it is
>being flooded by trash posts in every group I've looked at.
That's a change over the last year in the one's I'm haunting. The
signal::noise ratio is much lower.
> Also, I'm
>sad to see the flamewars in this group. But I guess that's life in the
>21st century. I probably won't stay around long.
Thanks for dropping in.
> The "from" address
>on this post is not valid. If anyone wants to send me anything, I have
>an account at gmail.com where the username is my initials, a dot, and
>"ka10".
>
>Best,
>/Rcc
>Lexington MA
>DEC '64-'72, BBN '72-'03, currently probably retired
Aw, that's too bad. Sorry to hear that.
/BAH
Bob Clements
> COMPIL was invented by Bill Weir, a graduate student in the Stanford
> Artificial Intelligence Project, in the late 1960s. It was called RPG,
> which stood for Rapid Program Generation. [...]
Hi, John,
I think I've told this before on this newsgroup --- many moons ago.
But I'll add a couple of comments.
> The program was coded in
> FAIL, a one-pass assembler written by Phil Petit, also a graduate
> student in the Stanford AI Project who shared an office with Bill.
> Bill requested assistance from two people. Dick Gruen, a DEC employee,
> modified the MACRO assembler and the FORTRAN compiler to interface to
> Bill's RPG. I modified the operating system to allow RPG to retrieve
> the command line that invoked it.
That was all at SAIL, not DEC, right?
> When DEC made RPG part of the TOPS-10 product they changed the name to
> COMPIL and made it assemble in the standard assembler, MACRO.
I think I did that, and committed the MACRO/LOADER/FORTRAN/... edits.
> The name was a joke, of course. RPG is the name of a well-known IBM
> programming language, and therefore completely unsuitable for an
> offering from DEC. In addition, Dick Gruen was also known as Richard P.
> Gruen.
And thereby hangs a tale. When Dick gave me the RPG source, he told
me that DEC could have it if and only if we continued to call it RPG.
Larry Portner vetoed that. He said we weren't calling it by a name
that IBM was already using, for a completely different product. It
would just confuse people. I told him that Dick/Bill had made it a
condition of the donation. I don't know whether Larry actually had
any contact with them on the subject. But I got the flak from them.
> Does anyone know where Bill Wier is now?
You've forgotten the spelling of his name. It's William F Weiher.
Google gives only a couple of hits on that, with one saying he is
(or was) at Intuit, but no further info. Leaving out the "F" gets
more hits, not all of which are _this_ Bill Weiher, and nothing that
looks recent.
> John Sauter (John_...@systemeyescomputerstore.com)
/Rcc
Rich Alderson
The Multicians all seem to think that emulating the DPS-8 hardware would be
difficult. From an admittedly cursory reading of the hardware manual, I don't
see where it would be any more difficult than KLH10, though I agree that SimH
is probably too limited. After all, it only provides a KS, not a KL, and
drafts off the PDP-11 emulator for its Unibus peripherals.
Al Kossow
> fishtop_records <pfar...@pfarrell.com> writes:
>
>> On Nov 30, 8:58 pm, Morten Reistad <fi...@last.name> wrote:
>
>>> Multics has a separate shell, but it is implemented differently.
>>> It is a stack invocation, not a process.
>
>> I worked with Tom Van Vleck, a keeper of the Multic flame, and chatted
>> recently when MIT released the source code.
>> While the Honeywell 36 bit machines had some similarities to TOPS-20,
>> it sounds like it was too weird to be ported to something like SIMH
>
> The Multicians all seem to think that emulating the DPS-8 hardware would be
> difficult. From an admittedly cursory reading of the hardware manual, I don't
> see where it would be any more difficult than KLH10, though I agree that SimH
> is probably too limited. After all, it only provides a KS, not a KL, and
> drafts off the PDP-11 emulator for its Unibus peripherals.
>
The problem is the undocumented nature of the rather baroque mass storage and comms
subsystems, not the CPU itself.
John Sauter
> John Sauter wrote:
>
>> COMPIL was invented by Bill Weir, a graduate student in the Stanford
>> Artificial Intelligence Project, in the late 1960s. It was called RPG,
>> which stood for Rapid Program Generation. [...]
>
> Hi, John,
>
> I think I've told this before on this newsgroup --- many moons ago.
> But I'll add a couple of comments.
>
Thank you for filling in gaps, Bob. I don't recall ever seeing the
information you provided below.
>
>
>> The program was coded in
>> FAIL, a one-pass assembler written by Phil Petit, also a graduate
>> student in the Stanford AI Project who shared an office with Bill.
>> Bill requested assistance from two people. Dick Gruen, a DEC employee,
>> modified the MACRO assembler and the FORTRAN compiler to interface to
>> Bill's RPG. I modified the operating system to allow RPG to retrieve
>> the command line that invoked it.
>
> That was all at SAIL, not DEC, right?
Yes.
>
>
>> When DEC made RPG part of the TOPS-10 product they changed the name to
>> COMPIL and made it assemble in the standard assembler, MACRO.
>
> I think I did that, and committed the MACRO/LOADER/FORTRAN/... edits.
>
>> The name was a joke, of course. RPG is the name of a well-known IBM
>> programming language, and therefore completely unsuitable for an
>> offering from DEC. In addition, Dick Gruen was also known as Richard P.
>> Gruen.
>
> And thereby hangs a tale. When Dick gave me the RPG source, he told
> me that DEC could have it if and only if we continued to call it RPG.
> Larry Portner vetoed that. He said we weren't calling it by a name
> that IBM was already using, for a completely different product. It
> would just confuse people. I told him that Dick/Bill had made it a
> condition of the donation. I don't know whether Larry actually had
> any contact with them on the subject. But I got the flak from them.
>
Knowing Dick, I suspect his demand was just ego on his part, with no
insistence from Bill. I know I was not consulted on the name.
Obviously Larry made the right decision--poking IBM in the eye is never
a good idea.
>> Does anyone know where Bill Wier is now?
>
> You've forgotten the spelling of his name. It's William F Weiher.
> Google gives only a couple of hits on that, with one saying he is
> (or was) at Intuit, but no further info. Leaving out the "F" gets
> more hits, not all of which are _this_ Bill Weiher, and nothing that
> looks recent.
>
Thanks for the correction; spelling was never my strong suit.
In my original posting I credited you with including the ability to read
the invoking command line in your rewrite of the terminal services. I
got that impression from a conversation we had when you visited the KA10
at Sanders in the early 1970s--you said that you had let the design of
the terminal services rewrite settle in your mind before you started
writing code, and you thought that the quality of the software was
better as a result. Am I correct in my belief that it was the COMPIL
cusp that motivated you to include the ability to read the invoking
command line?
Another tidbit: you may not have known that the first version of the
PDP-6 monitor which ran at the Stanford AI Lab was called RCC 2.4L.
Bob Clements
> Am I correct in my belief that it was the COMPIL
> cusp that motivated you to include the ability to read the invoking
> command line?
I have no recollection of doing that. But if I did, that would
certainly have been the motivation.
I recall that CP/M and later MS/DOS had that capability -- simply
by looking at the right place in low memory. A natural tool in
the days before real shells.
> Another tidbit: you may not have known that the first version of the
> PDP-6 monitor which ran at the Stanford AI Lab was called RCC 2.4L.
Well, you know the story of that PDP-6's installation, I'm sure.
I was there for a number of weeks, and certainly did some monitor
builds while I was there. Any local changes or bug fixes would have
been taken back to Maynard. It's quite possible that I actually named
that monitor that way to distinguish it from any built from the
sources we started with. Or maybe not. Again, that's a detail I
don't remember after all this time.
/Rcc
Mark Crispin
> Well, you know the story of that PDP-6's installation, I'm sure.
The story you told at DECUS in 1984? Sometime I or someone else ought to
transcribe it from the video tape.
-- Mark --
http://panda.com/mrc
Democracy is two wolves and a sheep deciding what to eat for lunch.
Liberty is a well-armed sheep contesting the vote.
Bob Clements
> On Tue, 4 Dec 2007, Bob Clements posted:
>> Well, you know the story of that PDP-6's installation, I'm sure.
>
> The story you told at DECUS in 1984?
If it talks about the water hose, the Bermuda shorts and the dry ice,
then yes.
> Sometime I or someone else ought
> to transcribe it from the video tape.
One's words to tend to follow one around, don't they ...
I still need to thank whoever was in charge of DECUS at the time
for the funding for my plane fare and hotel. My employer would
not spring for it; I wasn't doing anything 36-bit-related at
that time.
> -- Mark --
/Rcc
Mark Crispin
> then yes.
and you putting a bit of pressure on Les... ;-)
-- Mark --
http://staff.washington.edu/mrc
Science does not emerge from voting, party politics, or public debate.
Si vis pacem, para bellum.
John Sauter
> Mark Crispin wrote:
>> On Tue, 4 Dec 2007, Bob Clements posted:
>>> Well, you know the story of that PDP-6's installation, I'm sure.
>>
>> The story you told at DECUS in 1984?
>
> If it talks about the water hose, the Bermuda shorts and the dry ice,
> then yes.
>
I wasn't at the 1984 DECUS, unfortunately. I had left 36-bit-land in
1978. However, I was one of the many who helped Bob install the PDP-6
at Stanford: we rolled the cabinets onto the raised floor, and put dry
ice underneath the floor tiles to cool it.
A minor item I haven't seen recounted elsewhere is that Bob told us that
when we "accepted" the PDP-6 he was authorized to take us to dinner.
That was a good business lesson.
John Sauter (John_...@systemeyescomputerstore.com)
bob
> On Nov 30, 8:25 pm, glen herrmannsfeldt <g...@ugcs.caltech.edu> wrote:
>> Was it that DEC/Digital didn't understand how to be a big
>> computer company?
>
> Word! They had no clue. Rumour was than whenever KO thought they would
> compete with IBM, he would kill the project.
> Its only urban legend, but it rings true.
Not urban legend, true, experienced it a couple of times, maybe 4.
>
>> Then there was IBM's VM where not only the command interpreter but
>> just about everything else was done in user space. Write your
>> own OS and IPL it into your virtual machine.
>
> Yes, AMS was mostly an IBM shop, they bought TOPS-20 systems to
> compete in delivering interactive systems.
> The IBM fanboys argued that VM/CMS was as good as TOPS-20, that IPLing
> your own OS was a great thing, etc.
>
> But, when it came to developing solution for paying customers, the
> TOPS-20 folks delivered a lot more in less time for less money.
> There never was a big market for VM/CMS, altho we had the IBM gear to
> support it.
>
bob
jmfb...@aol.com
>fishtop_records wrote:
>> On Nov 30, 8:25 pm, glen herrmannsfeldt <g...@ugcs.caltech.edu> wrote:
>>> Was it that DEC/Digital didn't understand how to be a big
>>> computer company?
>>
>> Word! They had no clue. Rumour was than whenever KO thought they would
>> compete with IBM, he would kill the project.
>> Its only urban legend, but it rings true.
>Not urban legend, true, experienced it a couple of times, maybe 4.
It was for survival.
/BAH
Pat Farrell
>>> Word! They had no clue. Rumour was than whenever KO thought they would
>>> compete with IBM, he would kill the project.
>>> Its only urban legend, but it rings true.
>>Not urban legend, true, experienced it a couple of times, maybe 4.
>
> It was for survival.
As Doctor Feel would ask:
\
How did that work out for them?
jmfb...@aol.com
Pat Farrell
> In article <3oap25-...@tools.pfarrell.com>,
>>> It was for survival.
>>How did that work out for them?
>
> For whom?
Did DEC/Digital survive with this enlightened view?
When Doctor Feel asks, he always knows the answer before hand.
jmfb...@aol.com
Yes, they did survive. It became a billion dolloar company.
Then it could compete with IBM.
/BAH