The Unix Heritage Society mailing list
 help / color / mirror / Atom feed
* [TUHS] 32V memory management: not quite V7 style swapping -- source code update
@ 2021-06-06 18:03 Paul Ruizendaal via TUHS
  2021-06-06 18:23 ` Clem Cole
  0 siblings, 1 reply; 16+ messages in thread
From: Paul Ruizendaal via TUHS @ 2021-06-06 18:03 UTC (permalink / raw)
  To: TUHS main list

I just noticed that the 32V tape on the TUHS Unix Tree page includes a directory “slowsys”:
https://www.tuhs.org/cgi-bin/utree.pl?file=32V/usr/src/slowsys

This “slowsys” directory appears to contain the 32V kernel with a pure swapping architecture. It is not quite the kernel described in the well known 32V paper, as it seems to have advanced from a fixed 192KB image size mapping to a more variable mapping of up to 1MB — but otherwise the code appears to be as described in the July 1978 paper.

The directory https://www.tuhs.org/cgi-bin/utree.pl?file=32V/usr/src/sys contains the scatter loading, partial swapping version of the 32V kernel.

Paul
 

^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [TUHS] 32V memory management: not quite V7 style swapping -- source code update
  2021-06-06 18:03 [TUHS] 32V memory management: not quite V7 style swapping -- source code update Paul Ruizendaal via TUHS
@ 2021-06-06 18:23 ` Clem Cole
  2021-06-06 18:30   ` Larry McVoy
                     ` (2 more replies)
  0 siblings, 3 replies; 16+ messages in thread
From: Clem Cole @ 2021-06-06 18:23 UTC (permalink / raw)
  To: Paul Ruizendaal; +Cc: TUHS main list

[-- Attachment #1: Type: text/plain, Size: 975 bytes --]

Paul,

You got me thinking and I'm curious if anyone really knows historically how
many sites ran a 32V system?   In those days (late 70s/early 80s) the
universities that knew and and even many sites inside the Bell System, the
Vaxen I ran 4.1BSD (say the Marx's brothers at Whippany along with the Vax
in the underseas research lab were we put the AP I did for my thesis).
There were a couple in Summit I know, and probably Homdel and I'm guessing
in some of the operating companies, but I never got the feeling 32V was
popular.  The folks with Vaxen that I knew, if you were able to run BSD
(4.1 and eventually 4.2), did.  Later on the only non-'pure-joy' systems I
knew were a couple of Ultrix systems because they wanted the support from
DEC and IIRC were using FORTRAN and wanted the DEC compiler which only ran
on Ultrix or VMS.  Inside of AT&T, I personally think I knew more folks
with VMS (Fortran being the key anchor)  than those that ran 32V.
ᐧ

[-- Attachment #2: Type: text/html, Size: 1582 bytes --]

^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [TUHS] 32V memory management: not quite V7 style swapping -- source code update
  2021-06-06 18:23 ` Clem Cole
@ 2021-06-06 18:30   ` Larry McVoy
  2021-06-06 18:35     ` Clem Cole
  2021-06-06 18:39     ` Warner Losh
  2021-06-06 19:08   ` Paul Ruizendaal via TUHS
  2021-06-07 19:24   ` Greg A. Woods
  2 siblings, 2 replies; 16+ messages in thread
From: Larry McVoy @ 2021-06-06 18:30 UTC (permalink / raw)
  To: Clem Cole; +Cc: TUHS main list, Paul Ruizendaal

BSD had networking.  Once you had that, you don't look back.  Sys V (and
prior) so far as I know, didn't get networking until Coherent did their
STREAMS stack that somehow ended up at Lachman - I ported it to a crazy
super computer and to SCO Unix.  SCO was pretty stock AT&T code and let
me tell you, it felt pretty crappy after having used BSD and then SunOS.
It was a giant step backwards.

I just think the BSD folks were moving forward faster.  Rob with start
talking about cat waving its tail, I get it, not everything was better
but a lot was.  Solid networking that performed was very pleasant.

On Sun, Jun 06, 2021 at 02:23:49PM -0400, Clem Cole wrote:
> Paul,
> 
> You got me thinking and I'm curious if anyone really knows historically how
> many sites ran a 32V system?   In those days (late 70s/early 80s) the
> universities that knew and and even many sites inside the Bell System, the
> Vaxen I ran 4.1BSD (say the Marx's brothers at Whippany along with the Vax
> in the underseas research lab were we put the AP I did for my thesis).
> There were a couple in Summit I know, and probably Homdel and I'm guessing
> in some of the operating companies, but I never got the feeling 32V was
> popular.  The folks with Vaxen that I knew, if you were able to run BSD
> (4.1 and eventually 4.2), did.  Later on the only non-'pure-joy' systems I
> knew were a couple of Ultrix systems because they wanted the support from
> DEC and IIRC were using FORTRAN and wanted the DEC compiler which only ran
> on Ultrix or VMS.  Inside of AT&T, I personally think I knew more folks
> with VMS (Fortran being the key anchor)  than those that ran 32V.
> ???

-- 
---
Larry McVoy            	     lm at mcvoy.com             http://www.mcvoy.com/lm 

^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [TUHS] 32V memory management: not quite V7 style swapping -- source code update
  2021-06-06 18:30   ` Larry McVoy
@ 2021-06-06 18:35     ` Clem Cole
  2021-06-06 18:40       ` Warner Losh
  2021-06-06 18:53       ` Clem Cole
  2021-06-06 18:39     ` Warner Losh
  1 sibling, 2 replies; 16+ messages in thread
From: Clem Cole @ 2021-06-06 18:35 UTC (permalink / raw)
  To: Larry McVoy; +Cc: TUHS main list, Paul Ruizendaal

[-- Attachment #1: Type: text/plain, Size: 2002 bytes --]

4.2 had networking, 4.1 did not.   32V did not either.  I'm asking 32V vs
4.1
ᐧ

On Sun, Jun 6, 2021 at 2:30 PM Larry McVoy <lm@mcvoy.com> wrote:

> BSD had networking.  Once you had that, you don't look back.  Sys V (and
> prior) so far as I know, didn't get networking until Coherent did their
> STREAMS stack that somehow ended up at Lachman - I ported it to a crazy
> super computer and to SCO Unix.  SCO was pretty stock AT&T code and let
> me tell you, it felt pretty crappy after having used BSD and then SunOS.
> It was a giant step backwards.
>
> I just think the BSD folks were moving forward faster.  Rob with start
> talking about cat waving its tail, I get it, not everything was better
> but a lot was.  Solid networking that performed was very pleasant.
>
> On Sun, Jun 06, 2021 at 02:23:49PM -0400, Clem Cole wrote:
> > Paul,
> >
> > You got me thinking and I'm curious if anyone really knows historically
> how
> > many sites ran a 32V system?   In those days (late 70s/early 80s) the
> > universities that knew and and even many sites inside the Bell System,
> the
> > Vaxen I ran 4.1BSD (say the Marx's brothers at Whippany along with the
> Vax
> > in the underseas research lab were we put the AP I did for my thesis).
> > There were a couple in Summit I know, and probably Homdel and I'm
> guessing
> > in some of the operating companies, but I never got the feeling 32V was
> > popular.  The folks with Vaxen that I knew, if you were able to run BSD
> > (4.1 and eventually 4.2), did.  Later on the only non-'pure-joy' systems
> I
> > knew were a couple of Ultrix systems because they wanted the support from
> > DEC and IIRC were using FORTRAN and wanted the DEC compiler which only
> ran
> > on Ultrix or VMS.  Inside of AT&T, I personally think I knew more folks
> > with VMS (Fortran being the key anchor)  than those that ran 32V.
> > ???
>
> --
> ---
> Larry McVoy                  lm at mcvoy.com
> http://www.mcvoy.com/lm
>

[-- Attachment #2: Type: text/html, Size: 2954 bytes --]

^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [TUHS] 32V memory management: not quite V7 style swapping -- source code update
  2021-06-06 18:30   ` Larry McVoy
  2021-06-06 18:35     ` Clem Cole
@ 2021-06-06 18:39     ` Warner Losh
  2021-06-06 19:23       ` Clem Cole
  1 sibling, 1 reply; 16+ messages in thread
From: Warner Losh @ 2021-06-06 18:39 UTC (permalink / raw)
  To: Larry McVoy; +Cc: TUHS main list, Paul Ruizendaal

[-- Attachment #1: Type: text/plain, Size: 2006 bytes --]

Outside of Bell Labs the thing was "order the V32 tape. Never install it.
Send copies of the license to Berkeley. Run 4BSD."

Warner

On Sun, Jun 6, 2021, 12:30 PM Larry McVoy <lm@mcvoy.com> wrote:

> BSD had networking.  Once you had that, you don't look back.  Sys V (and
> prior) so far as I know, didn't get networking until Coherent did their
> STREAMS stack that somehow ended up at Lachman - I ported it to a crazy
> super computer and to SCO Unix.  SCO was pretty stock AT&T code and let
> me tell you, it felt pretty crappy after having used BSD and then SunOS.
> It was a giant step backwards.
>
> I just think the BSD folks were moving forward faster.  Rob with start
> talking about cat waving its tail, I get it, not everything was better
> but a lot was.  Solid networking that performed was very pleasant.
>
> On Sun, Jun 06, 2021 at 02:23:49PM -0400, Clem Cole wrote:
> > Paul,
> >
> > You got me thinking and I'm curious if anyone really knows historically
> how
> > many sites ran a 32V system?   In those days (late 70s/early 80s) the
> > universities that knew and and even many sites inside the Bell System,
> the
> > Vaxen I ran 4.1BSD (say the Marx's brothers at Whippany along with the
> Vax
> > in the underseas research lab were we put the AP I did for my thesis).
> > There were a couple in Summit I know, and probably Homdel and I'm
> guessing
> > in some of the operating companies, but I never got the feeling 32V was
> > popular.  The folks with Vaxen that I knew, if you were able to run BSD
> > (4.1 and eventually 4.2), did.  Later on the only non-'pure-joy' systems
> I
> > knew were a couple of Ultrix systems because they wanted the support from
> > DEC and IIRC were using FORTRAN and wanted the DEC compiler which only
> ran
> > on Ultrix or VMS.  Inside of AT&T, I personally think I knew more folks
> > with VMS (Fortran being the key anchor)  than those that ran 32V.
> > ???
>
> --
> ---
> Larry McVoy                  lm at mcvoy.com
> http://www.mcvoy.com/lm
>

[-- Attachment #2: Type: text/html, Size: 2682 bytes --]

^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [TUHS] 32V memory management: not quite V7 style swapping -- source code update
  2021-06-06 18:35     ` Clem Cole
@ 2021-06-06 18:40       ` Warner Losh
  2021-06-06 18:52         ` Richard Salz
  2021-06-06 18:53       ` Clem Cole
  1 sibling, 1 reply; 16+ messages in thread
From: Warner Losh @ 2021-06-06 18:40 UTC (permalink / raw)
  To: Clem Cole; +Cc: TUHS main list, Paul Ruizendaal

[-- Attachment #1: Type: text/plain, Size: 2176 bytes --]

On Sun, Jun 6, 2021, 12:36 PM Clem Cole <clemc@ccc.com> wrote:

> 4.2 had networking, 4.1 did not.   32V did not either.  I'm asking 32V vs
> 4.1
> ᐧ
>

My info is from the 4.2 time frame...

Warner

On Sun, Jun 6, 2021 at 2:30 PM Larry McVoy <lm@mcvoy.com> wrote:
>
>> BSD had networking.  Once you had that, you don't look back.  Sys V (and
>> prior) so far as I know, didn't get networking until Coherent did their
>> STREAMS stack that somehow ended up at Lachman - I ported it to a crazy
>> super computer and to SCO Unix.  SCO was pretty stock AT&T code and let
>> me tell you, it felt pretty crappy after having used BSD and then SunOS.
>> It was a giant step backwards.
>>
>> I just think the BSD folks were moving forward faster.  Rob with start
>> talking about cat waving its tail, I get it, not everything was better
>> but a lot was.  Solid networking that performed was very pleasant.
>>
>> On Sun, Jun 06, 2021 at 02:23:49PM -0400, Clem Cole wrote:
>> > Paul,
>> >
>> > You got me thinking and I'm curious if anyone really knows historically
>> how
>> > many sites ran a 32V system?   In those days (late 70s/early 80s) the
>> > universities that knew and and even many sites inside the Bell System,
>> the
>> > Vaxen I ran 4.1BSD (say the Marx's brothers at Whippany along with the
>> Vax
>> > in the underseas research lab were we put the AP I did for my thesis).
>> > There were a couple in Summit I know, and probably Homdel and I'm
>> guessing
>> > in some of the operating companies, but I never got the feeling 32V was
>> > popular.  The folks with Vaxen that I knew, if you were able to run BSD
>> > (4.1 and eventually 4.2), did.  Later on the only non-'pure-joy'
>> systems I
>> > knew were a couple of Ultrix systems because they wanted the support
>> from
>> > DEC and IIRC were using FORTRAN and wanted the DEC compiler which only
>> ran
>> > on Ultrix or VMS.  Inside of AT&T, I personally think I knew more folks
>> > with VMS (Fortran being the key anchor)  than those that ran 32V.
>> > ???
>>
>> --
>> ---
>> Larry McVoy                  lm at mcvoy.com
>> http://www.mcvoy.com/lm
>>
>

[-- Attachment #2: Type: text/html, Size: 3676 bytes --]

^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [TUHS] 32V memory management: not quite V7 style swapping -- source code update
  2021-06-06 18:40       ` Warner Losh
@ 2021-06-06 18:52         ` Richard Salz
  2021-06-06 19:02           ` Clem Cole
  0 siblings, 1 reply; 16+ messages in thread
From: Richard Salz @ 2021-06-06 18:52 UTC (permalink / raw)
  To: Warner Losh; +Cc: TUHS main list, Paul Ruizendaal

[-- Attachment #1: Type: text/plain, Size: 42 bytes --]

Didn't 4.1c have some kind of networking?

[-- Attachment #2: Type: text/html, Size: 71 bytes --]

^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [TUHS] 32V memory management: not quite V7 style swapping -- source code update
  2021-06-06 18:35     ` Clem Cole
  2021-06-06 18:40       ` Warner Losh
@ 2021-06-06 18:53       ` Clem Cole
  1 sibling, 0 replies; 16+ messages in thread
From: Clem Cole @ 2021-06-06 18:53 UTC (permalink / raw)
  To: Larry McVoy; +Cc: TUHS main list, Paul Ruizendaal

[-- Attachment #1: Type: text/plain, Size: 2655 bytes --]

4.2 and cheap LAN connections and later WAN connections drove the hockey
stick - no doubt and your are right, until the Stream networking work
(which was originally Steve Rago and the folks in Summit, IIRC), PWB 3/PWB
4.0 and later the commercially System V was definitely at a huge
disadvantage and even when the Streams networking code shows us, the
performance was not nearly as good plus by that time, all the code had
started to assume sockets (and sockets had been ported to DOS), so Streams
was still born.
ᐧ

On Sun, Jun 6, 2021 at 2:35 PM Clem Cole <clemc@ccc.com> wrote:

> 4.2 had networking, 4.1 did not.   32V did not either.  I'm asking 32V vs
> 4.1
> ᐧ
>
> On Sun, Jun 6, 2021 at 2:30 PM Larry McVoy <lm@mcvoy.com> wrote:
>
>> BSD had networking.  Once you had that, you don't look back.  Sys V (and
>> prior) so far as I know, didn't get networking until Coherent did their
>> STREAMS stack that somehow ended up at Lachman - I ported it to a crazy
>> super computer and to SCO Unix.  SCO was pretty stock AT&T code and let
>> me tell you, it felt pretty crappy after having used BSD and then SunOS.
>> It was a giant step backwards.
>>
>> I just think the BSD folks were moving forward faster.  Rob with start
>> talking about cat waving its tail, I get it, not everything was better
>> but a lot was.  Solid networking that performed was very pleasant.
>>
>> On Sun, Jun 06, 2021 at 02:23:49PM -0400, Clem Cole wrote:
>> > Paul,
>> >
>> > You got me thinking and I'm curious if anyone really knows historically
>> how
>> > many sites ran a 32V system?   In those days (late 70s/early 80s) the
>> > universities that knew and and even many sites inside the Bell System,
>> the
>> > Vaxen I ran 4.1BSD (say the Marx's brothers at Whippany along with the
>> Vax
>> > in the underseas research lab were we put the AP I did for my thesis).
>> > There were a couple in Summit I know, and probably Homdel and I'm
>> guessing
>> > in some of the operating companies, but I never got the feeling 32V was
>> > popular.  The folks with Vaxen that I knew, if you were able to run BSD
>> > (4.1 and eventually 4.2), did.  Later on the only non-'pure-joy'
>> systems I
>> > knew were a couple of Ultrix systems because they wanted the support
>> from
>> > DEC and IIRC were using FORTRAN and wanted the DEC compiler which only
>> ran
>> > on Ultrix or VMS.  Inside of AT&T, I personally think I knew more folks
>> > with VMS (Fortran being the key anchor)  than those that ran 32V.
>> > ???
>>
>> --
>> ---
>> Larry McVoy                  lm at mcvoy.com
>> http://www.mcvoy.com/lm
>>
>

[-- Attachment #2: Type: text/html, Size: 4212 bytes --]

^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [TUHS] 32V memory management: not quite V7 style swapping -- source code update
  2021-06-06 18:52         ` Richard Salz
@ 2021-06-06 19:02           ` Clem Cole
  2021-06-06 19:18             ` arnold
  0 siblings, 1 reply; 16+ messages in thread
From: Clem Cole @ 2021-06-06 19:02 UTC (permalink / raw)
  To: Richard Salz; +Cc: TUHS main list, Paul Ruizendaal

[-- Attachment #1: Type: text/plain, Size: 2223 bytes --]

Yes -- this is what I just sent to Larry on  private exchange.

The BSD release order - 3.0 BSD is the replacement for 32V, 4.0 is the
first real update and 4.1 was the so-called 'FASTVAX' stuff Joy did as part
of the UNIX/VMS bake off with Stanford.  4.0 and 4.1 are very similar.
 4.1 lived 'in the wild' about 3-4 years as the standard 'BSD' and any one
with Vax (certainly at a university) got the BSD tape.  That is the code
base that Rob Gurwitz used to write the original BBN TCP -- which did not
use open (/dev/ip  /dev/tcp/hostname or /dev/udp/hostname) like MIT's
ChaosNet.

4.1a was Bill's first attempt at rewriting Gurwitz's code and creating
sockets.   Only a few people had it (I was one) - but it did make it to a
number of the IETF beta testers -- the MIT folk I know had the tape - Noel
might know whom there they had it.   4.1b replaced that about 3-6 months
later but I'm not sure it left UCB very far -- I ran it on one of the CAD
machines as 4.1c was being created.  IIRC B was the first to have Kirks FFS
in it and we were pushing the FFS code with our cad programs that beat on
the FS pretty hard.    BBN might have had 4.1b, as I think that is where
Gurwitz & Walsh started the BBN2 code.   4.1c came pretty soon thereafter
as the 'beta' for 4.2 and that also went to IETF folks.   Joy left for Sun,
so I know Masscomp, Sun and DEC all had copies of 4.1c.   San wnj, Kirk and
Sam completed the beta, and got the formal release creating the BSD 4.2
release (which as Larry points out was the first truly networked UNIX) and
that went just anyone with a UCB license (and all hell broke loose because
- Henry Spencer's famous '*4.2 is just like Unix, only different*' quote).
 BBN released their alternative Network stack for sockets (*a.k.a.* BBN2 or
Bob Walsh version) and that fight started.  About a year or so later Kirk
and Sam got 4.3 out which was putting back in a number 4.1-ism so not so
much code broke, I think Van Jacobson's stuff up the hill at LBL went and
some of the new BBN stuff.  4.4 was at least a year after that.

ᐧ

On Sun, Jun 6, 2021 at 2:52 PM Richard Salz <rich.salz@gmail.com> wrote:

> Didn't 4.1c have some kind of networking?
>

[-- Attachment #2: Type: text/html, Size: 3341 bytes --]

^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [TUHS] 32V memory management: not quite V7 style swapping -- source code update
  2021-06-06 18:23 ` Clem Cole
  2021-06-06 18:30   ` Larry McVoy
@ 2021-06-06 19:08   ` Paul Ruizendaal via TUHS
  2021-06-07 19:24   ` Greg A. Woods
  2 siblings, 0 replies; 16+ messages in thread
From: Paul Ruizendaal via TUHS @ 2021-06-06 19:08 UTC (permalink / raw)
  To: TUHS main list

[-- Attachment #1: Type: text/plain, Size: 2639 bytes --]

Clem,

> You got me thinking and I'm curious if anyone really knows historically how many sites ran a 32V system?

I don’t know the answer to that, but my outsider impression is that 32V and 32-bit Sys III had a very small installed base, and hence that the innovations in it are not all that well known or understood.

However, in searching the web for things related to “scatter loading” I find some references to Xenix. It may be that some of the 32V/SysIII stuff lived in high volume during the mid-80’s through that code base.

——

Going on a tangent, I am wondering about “mmap”. John Reiser wrote that he based the idea on PMAP from Tenex/TOPS-20, and fretted over a system call with 6 parameters. This would be 1980 at the latest.

In their 1981 discussion paper “Proposals for the enhancement of Unix on the VAX” Joy and Fabry propose some mmap like ideas, but without this API specifically. In the 1983 “4.2BSD system manual” they propose the mmap call in more or less its current form, but it is not implemented. It remained unimplemented until SunOS 4 in 1988 and 4.3BSD Reno in 1990.

It makes me wonder how close to the current “mmap" the code in the demand paging version of 32V was; it might have been quite close.

By the way, from that discussion paper it seems that CSRG had access to the demand paging version of the 32V code in 1981 but unfortunately it does not seem to have survived there (I've checked the DVD that Kirk McKusick provides, but the 32V code on there seems the same as what Keith Bostic provided to TUHS).

Paul

> On Jun 6, 2021, at 8:23 PM, Clem Cole <clemc@ccc.com> wrote:
> 
> Paul,
> 
> You got me thinking and I'm curious if anyone really knows historically how many sites ran a 32V system?   In those days (late 70s/early 80s) the universities that knew and and even many sites inside the Bell System, the Vaxen I ran 4.1BSD (say the Marx's brothers at Whippany along with the Vax in the underseas research lab were we put the AP I did for my thesis).  There were a couple in Summit I know, and probably Homdel and I'm guessing in some of the operating companies, but I never got the feeling 32V was popular.  The folks with Vaxen that I knew, if you were able to run BSD (4.1 and eventually 4.2), did.  Later on the only non-'pure-joy' systems I knew were a couple of Ultrix systems because they wanted the support from DEC and IIRC were using FORTRAN and wanted the DEC compiler which only ran on Ultrix or VMS.  Inside of AT&T, I personally think I knew more folks with VMS (Fortran being the key anchor)  than those that ran 32V.
> ᐧ


[-- Attachment #2: Type: text/html, Size: 4382 bytes --]

^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [TUHS] 32V memory management: not quite V7 style swapping -- source code update
  2021-06-06 19:02           ` Clem Cole
@ 2021-06-06 19:18             ` arnold
  2021-06-06 19:35               ` Clem Cole
  0 siblings, 1 reply; 16+ messages in thread
From: arnold @ 2021-06-06 19:18 UTC (permalink / raw)
  To: rich.salz, clemc; +Cc: tuhs, pnr

Clem Cole <clemc@ccc.com> wrote:

> About a year or so later Kirk
> and Sam got 4.3 out

4.3 was ~ 3 years after 4.2 (1983 -> 1986). It was mainly a huge
number of bug fixes and performance improvements.

> which was putting back in a number 4.1-ism so not so much code broke,

For example? As I remember it, it was essentially a super-bug-fixed
version of 4.2. Nothing more.

> 4.4 was at least a year after that.

It was many years after that, circa 1992.

Arnold

^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [TUHS] 32V memory management: not quite V7 style swapping -- source code update
  2021-06-06 18:39     ` Warner Losh
@ 2021-06-06 19:23       ` Clem Cole
  0 siblings, 0 replies; 16+ messages in thread
From: Clem Cole @ 2021-06-06 19:23 UTC (permalink / raw)
  To: Warner Losh; +Cc: TUHS main list, Paul Ruizendaal

[-- Attachment #1: Type: text/plain, Size: 1276 bytes --]

On Sun, Jun 6, 2021 at 2:39 PM Warner Losh <imp@bsdimp.com> wrote:

> Outside of Bell Labs the thing was "order the V32 tape. Never install it.
> Send copies of the license to Berkeley. Run 4BSD."
>
Right - certainly my behavior and what I preached at the time.  But it was
also true for >>some<< sites inside the labs - frankly many from the ones I
knew (Mary Ann what did run in Columbus).   There were definitely copies of
BSD in IH, as I know sent some of the stuff from my own work there. I never
really understood who 'had' to run 32/V or (PWB 3.0 and PWB 4.0) and like
Redman and the Marx machines were 'allowed' to be BSD.

What I came to realize is that if a site ran VMS inside the labs, it was
because they either had a commercial app that required it, or a had some
really important FORTRAN code that needed the DEC tools.    In the case of
the later, since the stuff I worked on had a commercial Fortran compiler
associated (great-great-grandfather to PGI compilers), those sites could
use BSD so they could get the distribution (I must have written about 25-30
tapes before I left).  But my stuff only ran on BSD 4.1/4.1a/4.1b/4.1c   -
I think Rick Spieklemeyers updated it when 4.2 and 4.3 finally finished but
I had left UCB by then.
ᐧ

[-- Attachment #2: Type: text/html, Size: 2359 bytes --]

^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [TUHS] 32V memory management: not quite V7 style swapping -- source code update
  2021-06-06 19:18             ` arnold
@ 2021-06-06 19:35               ` Clem Cole
  2021-06-06 19:45                 ` arnold
  0 siblings, 1 reply; 16+ messages in thread
From: Clem Cole @ 2021-06-06 19:35 UTC (permalink / raw)
  To: Aharon Robbins; +Cc: TUHS main list, Paul Ruizendaal

[-- Attachment #1: Type: text/plain, Size: 1186 bytes --]

On Sun, Jun 6, 2021 at 3:18 PM <arnold@skeeve.com> wrote:

> For example? As I remember it, it was essentially a super-bug-fixed version
> of 4.2. Nothing more.
>
To be honest I've forgotten all of the details, but the big one was the
ability to (optionall) have the old 4.1 signal semantics and I think there
were a couple of more.   The stuff that went into 4.3 was heavily driven by
the Arpa advisory board.  Many people in the Arpa community had felt that
CSRG was not supplying what they needed.  There were some compromises.  As
part of that the team put much of the CMU Mach memory system at some point
(I thought at 4.3) and as Paul pointed out, that's when mmap was finally
added [which is part of why I think that's when the CMU changes came].   I
also where the great line (whom I wish IO knew who first said it),
described "new BSD system was full of other folks' ideas, but had been peed
on to smell like UCB).

BTW: the other thing that was happening at UCB by then was that while the
CSRG project was funding a lot of stuff, it was not seen as research, but
more of a development for hire by then.  There started to be a lot of
resentment.
ᐧ
ᐧ

[-- Attachment #2: Type: text/html, Size: 2548 bytes --]

^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [TUHS] 32V memory management: not quite V7 style swapping -- source code update
  2021-06-06 19:35               ` Clem Cole
@ 2021-06-06 19:45                 ` arnold
  2021-06-06 20:46                   ` Clem Cole
  0 siblings, 1 reply; 16+ messages in thread
From: arnold @ 2021-06-06 19:45 UTC (permalink / raw)
  To: clemc, arnold; +Cc: tuhs, pnr

Clem Cole <clemc@ccc.com> wrote:

> There were some compromises.  As
> part of that the team put much of the CMU Mach memory system at some point
> (I thought at 4.3) and as Paul pointed out, that's when mmap was finally
> added [which is part of why I think that's when the CMU changes came].

Both of those were 4.4.

Just to set the record straight. :-)

Thanks,

Arnold

^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [TUHS] 32V memory management: not quite V7 style swapping -- source code update
  2021-06-06 19:45                 ` arnold
@ 2021-06-06 20:46                   ` Clem Cole
  0 siblings, 0 replies; 16+ messages in thread
From: Clem Cole @ 2021-06-06 20:46 UTC (permalink / raw)
  To: Aharon Robbins; +Cc: TUHS main list, Paul Ruizendaal

[-- Attachment #1: Type: text/plain, Size: 509 bytes --]

Fair enough -- thanks.
ᐧ

On Sun, Jun 6, 2021 at 3:45 PM <arnold@skeeve.com> wrote:

> Clem Cole <clemc@ccc.com> wrote:
>
> > There were some compromises.  As
> > part of that the team put much of the CMU Mach memory system at some
> point
> > (I thought at 4.3) and as Paul pointed out, that's when mmap was finally
> > added [which is part of why I think that's when the CMU changes came].
>
> Both of those were 4.4.
>
> Just to set the record straight. :-)
>
> Thanks,
>
> Arnold
>

[-- Attachment #2: Type: text/html, Size: 1261 bytes --]

^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [TUHS] 32V memory management: not quite V7 style swapping -- source code update
  2021-06-06 18:23 ` Clem Cole
  2021-06-06 18:30   ` Larry McVoy
  2021-06-06 19:08   ` Paul Ruizendaal via TUHS
@ 2021-06-07 19:24   ` Greg A. Woods
  2 siblings, 0 replies; 16+ messages in thread
From: Greg A. Woods @ 2021-06-07 19:24 UTC (permalink / raw)
  To: The Unix Heritage Society mailing list

[-- Attachment #1: Type: text/plain, Size: 2670 bytes --]

At Sun, 6 Jun 2021 14:23:49 -0400, Clem Cole <clemc@ccc.com> wrote:
Subject: Re: [TUHS] 32V memory management: not quite V7 style swapping -- source code update
>
> You got me thinking and I'm curious if anyone really knows historically how
> many sites ran a 32V system?   In those days (late 70s/early 80s) the
> universities that knew and and even many sites inside the Bell System, the
> Vaxen I ran 4.1BSD (say the Marx's brothers at Whippany along with the Vax
> in the underseas research lab were we put the AP I did for my thesis).

If my memory serves me correctly, University of Calgary ran 32V on their
first teaching VAX 11/780 for a short while when I was starting my
second year of undergrad (Sept. through to the xmas break, IIRC).  This
would be the fall of 1980.

After that first term I think it was running 3BSD for the rest of the
semester, and finally was running and early 4BSD not long after.

As I recall they had tried to run 3.x right away but had some problems
(possibly with a serial driver?  the initial setup had some 20 or 30
terminals) and in order to not have all us students trying to crowd onto
the old PDP 11/60 with just 12 terminals (which was also still in use by
a bunch of classes), while the VAX sat idle, they just gave up and
(re)installed 32V.  I think my class probably lost a week or two of time
to use the system.  At that time the only other teaching system was the
Multics mainframe, and it was also overloaded with too many users.

I remember being a little dismayed that the BSD C compiler seemed
entirely different from 32V (where it was very V7-like and thus what I
was familiar with from first year).  It wasn't until 4BSD offered me job
control and command history in CSH that I finally became more accepting
of BSD.

I think it was after the 4.x upgrade that they instituted CPU time
limits for students, and I remember discovering that if one caught
SIGXCPU then the limit just kept increasing -- i.e. the hard limit never
worked in the initial release of whatever version it was -- so I wrote a
little program that would catch the signal, then burn CPU in a loop
until the limit was above some requested value, and then it would fork a
shell.  I put that in my ~/.login and had lots of fun until I was
caught.  Then I fessed up and didn't get expelled!  They fixed the bug
of course and as a result of it all I then got to know the sysadmins
better and learned an awful lot more from them than I did in class on
some/many days.

--
					Greg A. Woods <gwoods@acm.org>

Kelowna, BC     +1 250 762-7675           RoboHack <woods@robohack.ca>
Planix, Inc. <woods@planix.com>     Avoncote Farms <woods@avoncote.ca>

[-- Attachment #2: OpenPGP Digital Signature --]
[-- Type: application/pgp-signature, Size: 195 bytes --]

^ permalink raw reply	[flat|nested] 16+ messages in thread

end of thread, other threads:[~2021-06-07 19:25 UTC | newest]

Thread overview: 16+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-06-06 18:03 [TUHS] 32V memory management: not quite V7 style swapping -- source code update Paul Ruizendaal via TUHS
2021-06-06 18:23 ` Clem Cole
2021-06-06 18:30   ` Larry McVoy
2021-06-06 18:35     ` Clem Cole
2021-06-06 18:40       ` Warner Losh
2021-06-06 18:52         ` Richard Salz
2021-06-06 19:02           ` Clem Cole
2021-06-06 19:18             ` arnold
2021-06-06 19:35               ` Clem Cole
2021-06-06 19:45                 ` arnold
2021-06-06 20:46                   ` Clem Cole
2021-06-06 18:53       ` Clem Cole
2021-06-06 18:39     ` Warner Losh
2021-06-06 19:23       ` Clem Cole
2021-06-06 19:08   ` Paul Ruizendaal via TUHS
2021-06-07 19:24   ` Greg A. Woods

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).