Any info on Acorn's own computing environment?

[BigEd]

We just got chatting about this in the online ABUG...

Any ideas what systems Acorn used internally, over the years? I'm pretty sure they had a VAX early on, and it seems rather likely they had Sun workstations later, but that's more of a guess.

Of course they would have used Acorn kit too, but surely that wouldn't have been the whole engineering environment.

Any concrete information, or recollections?

[BigEd]

In the Acorn Cambridge Workstation flyer, circa 1985, we see Acorn note their benchmarks are based on "In house tests using a DEC VAX 11/750 (without FPA under UNIX 4.2 bsd, Acorn Cambridge Workstation at 8 MHz)."

[BigEd]

This is probably an evolutionary dead-end, but I see

Hauser had even gone to the lengths of hiring three integrated-circuit designers and buying in chip design tools and workstations

What I mean is that those workstations wouldn't necessarily become part of Acorn's engineering setup. Ah, I see more detail:

The Acorn design team used VLSI Technology’s software installed on expensive, Motorola 68000 powered, Apollo Workstations

[paulb]

What I mean is that those workstations wouldn't necessarily become part of Acorn's engineering setup. Ah, I see more detail:

The Acorn design team used VLSI Technology’s software installed on expensive, Motorola 68000 powered, Apollo Workstations

I think that in the threads about Acorn's strategy, ARM development and Unix workstation efforts, we probably noted that Apollo workstations were the industry default for applications like Acorn's own VLSI chip design exercise, and they might well have been the inspiration for Acorn's own workstation plans.

Various companies wanted in on the workstation "action", and it was intriguing to learn that Commodore had hoped to pitch their Z8000-based 900 model as a kind of high-resolution graphics workstation, even though that wouldn't have stood a chance in the market (being pitched at the PDP-11 level when the industry had moved on to competing with the VAX). Meanwhile, ICL had a chance of partnering with Apollo but decided to do the British "plucky underdog" thing and bet on the Three Rivers PERQ, which was something of a technological dead end.

As for what you might use a VAX for, it is relevant that ULA design solutions were available for the VAX from both Ferranti and Qudos. See "Automation Cuts Design Time for Gate Arrays" (Computer Design, March 1984) for Ferranti's solution, which I think succeeded a PDP-11-based solution, and "An emerging market for British engineering tools" (Electronic Business, 15 October 1986) for a mention of Qudos' VAX-based solution. The Qudos solution might well have had some origins within Acorn, given Acorn's earlier "CAD workstation" aspirations, so maybe Acorn's VAX got some exercise in that endeavour.

And as for what else you might use a VAX for, it is highly plausible that it would be useful for software development. Since Acorn seems to have used Sun machines internally, I could imagine that a VAX might have been running Ultrix and be particularly amenable in providing a software development environment.

[jgharston]

And as for what else you might use a VAX for, it is highly plausible that it would be useful for software development.

There's also extant code for a sort of HostFS that used a serial link to the VAX for file storage.

(I wish I could remember where it is, it's probably in an "unsorted" directory somewhere )

[paulb]

There's also extant code for a sort of HostFS that used a serial link to the VAX for file storage.

(I wish I could remember where it is, it's probably in an "unsorted" directory somewhere )

I found it here...

https://mdfs.net/Software/Tube/Serial/

...while looking into using an emulated Electron as a terminal to a Unix system, with the serial connection being used as the filing system transport, probably less elegantly than with your solution.

[BigEd]

Acorn seems to have used Sun machines internally

Any detail on that?

I would expect the VAX in the early days could be doing logic simulation and circuit simulation (for the latter, a floating point unit would be a big help) - at Plessey we were using VAXes for both in the 85/86 timeframe. I would expect the ns32k and then the ARM second processors would be used too.

[markusher]

Graham Toal mentioned on a facebook post
"it could be Steve Furber's circuit-level emulation of the ARM during the design phase, which I guess he must have preferred over using Spice on the Vax"

[paulb]

Acorn seems to have used Sun machines internally

Any detail on that?

It is something I seem to remember from discussions and materials from the early 1990s, but I don't have any sources to hand. Acorn's Internet-facing machines were probably not their own Unix systems, or not all of them, at least.

[paulb]

Any concrete information, or recollections?

Back to the VAX, I just looked through the "AcornInternalDocuments" collection that someone shared somewhere at some point (sorry, no recollection of that), and there is this intriguing mention in a file called ARX (about the state of ARX in 1987):

The 'JuneOS' integration is proceeding to plan - the UK half of the system
will be sent to ARC on Friday, (crossing with the US half). The system
will be constructed from these parts on both VAXes and the two systems
will be kept in step (for bug fixes) using electronic mail. The operating
system kernel and the filing system are proceeding on schedule, the window
system is slightly behind schedule (but not critical).

[BigEd]

That's nice - sounds like one VAX per site!

[gtoal]

I was one of the few Vax users. What do you want to know? The primary software written on it was a VLSI design package written by Lee Smith as the primary coder, Richard Cownie as his backup, and myself as the 'toolsmith'. (We used the Fred Brooks "Mythical Man Month" style of programming team). The package was a language for connecting modules (automatic wiring - eg https://gtoal.com/acorn/panos/gtoal/m6809/all/mux3.all ) plus a very nice polygon manipulation package for the actual layout.

We had several utilities on the machine too such as Spice, though I ported Spice to the 16032 and used that more often.

I'm not aware of our CAD software being used much in the ARM project.

We did have a few minor side projects hosted on the Vax. Someone already mentioned my 'hostfs' file system for Panos (https://gtoal.com/acorn/files/HostFS-README.html ... vax end at https://gtoal.com/acorn/files/newfsys.mod with the Beeb code at https://gtoal.com/acorn/files/hostfs/ or https://gtoal.com/acorn/files/hostfs2/ ) Richard Cownie developed a few utilities on the Vax as well - toansii and topcc spring to mind, maybe also his executable compression program for ARM binaries.

I think there was also a UUCP connection to the Arx vax, over PSS, but I didn't have much to do with Arx who kept themselves to themselves so I may be wrong about the connection being over uucp. The vax had 16? serial lines connected mostly to BBC Micro terminals throughout the building. (Using my terminal rom that supported a graphical VLSI editor ( https://gtoal.com/acorn/panos/gtoal/newview/ ). It had the ECFG dithering built in that later ended up in the extension graphics rom - you can see it in a demo with the "Circuit" program at https://gtoal.com/acorn/files/ECFG2/ ).

There was also a couple of nice HP X/Y plotters - a small one and a big one. I still have a plot file for one of those plotters with my 6809 memory management chip design on it. ( https://gtoal.com/acorn/panos/gtoal/m68 ... hpplotfile )

[BigEd]

Thanks Graham, that's excellent info! And great to see some of your archives too.

Were the 16032 second processors fairly widely used as personal workstations?

What OS were the Vaxes running? What languages were you generally programming in?

What happened next: was there more than one VAX on site, or maybe microVAXes appeared, or perhaps Unix workstations of some kind?

We must get that 6809 MMU design plotted out!

[gtoal]

> Thanks Graham, that's excellent info! And great to see some of your archives too.

> Were the 16032 second processors fairly widely used as personal workstations?

Not *widely* but there were a few. Harry Oldham had one. I had one.

> What OS were the Vaxes running? What languages were you generally programming in?

4.2 BSD from BSDI. The majority of programming was in Modula II

> What happened next: was there more than one VAX on site, or maybe microVAXes appeared, or perhaps Unix workstations of some kind?

I only remember one Vax in Cambridge. We used 16032's a bit - especially for Spice. I don't remember any other local workstations. Some of the fabs we went to had graphical workstations. We got by on Beeb's talking to the Vax. Because so much of our VLSI was done using textual descriptions (All, Mlap etc) fancy editing stations weren't needed nearly as much except for the relatively small individual components like pads, multiplexers etc - anything larger was composed out of those textually rather than graphically. Even the low level objects were usually parameterised text rather than drawn so they could trivially adapt to different fab lines.

> We must get that 6809 MMU design plotted out!
It would be nice! I pinned my copy to the wall at UMIST when I took a year out to (try to) do an MSc in VLSI there, and someone stole it /

By the way, re:
> The Acorn design team used VLSI Technology’s software installed on expensive, Motorola 68000 powered, Apollo Workstations

I have *no idea* what this refers to. I saw some Apollos at one of the fab plants we visited but never saw any at Acorn.
G

[paulb]

Back to the VAX, I just looked through the "AcornInternalDocuments" collection that someone shared somewhere at some point (sorry, no recollection of that)

See "Acorn documents" for these documents. I was prompted to dig around a bit after having read "A Backup of Historical Proportions" about Xerox PARC's archives, where Jim Mitchell's departure for Acorn Computers is noted.

[paulb]

Thank you for your recollections!

By the way, re:
> The Acorn design team used VLSI Technology’s software installed on expensive, Motorola 68000 powered, Apollo Workstations

I have *no idea* what this refers to. I saw some Apollos at one of the fab plants we visited but never saw any at Acorn.

Was it possible that those workstations were confined to the ARM development operation, being required for the VLSI Technology tools, and therefore not exactly on general display? Also, may I ask what the nature of your own VLSI work was? Sorry if that passed me by!

[markusher]

Thanks Graham and Paul.
Particularly interesting is the OS the VAX was running, typically this would be VMS or Ultrix (DEC Unix based on BSD).
I'm running a MicroVaxII at the moment and am tempted to try and recreate a basic environment as was at Acorn. Connectivity to Econet, BBCs as terminal devices and the compiler setup etc.

In the Acorn Unix Econet Device Driver and Network Device document from the RISC-iX Group
http://8bs.com/othrdnld/manuals/econet/ ... Device.zip

many mentions are made (13.1.2, 13.2, 14.2.4)of the VAX system being used as a fileserver (NFS) to the BBC Econet network, ostensibly through a gateway machine with Econet and Ethernet cards. Probably an A4000/A5000 type machine or possibly a RiscPC.

Initially, I've been using Termulator or VAX (Uni Liverpool) via a console link but will certainly now look at NewView and HostFS etc.

[SteveBagley]

many mentions are made (13.1.2, 13.2, 14.2.4)of the VAX system being used as a fileserver (NFS) to the BBC Econet network, ostensibly through a gateway machine with Econet and Ethernet cards. Probably an A4000/A5000 type machine or possibly a RiscPC.

More likely to be a RISCiX machine such as an R140/R260 (even a R225), I'd have thought -- 1991/1992 seems very late for the type of experiements in that report and 1994/95 (the RiscPC era) exceptionally so, especialy as Acorn was pushing AUN and Ethernet by then over Econet…

Steve

[arg]

many mentions are made (13.1.2, 13.2, 14.2.4)of the VAX system being used as a fileserver (NFS) to the BBC Econet network, ostensibly through a gateway machine with Econet and Ethernet cards. Probably an A4000/A5000 type machine or possibly a RiscPC.

More likely to be a RISCiX machine such as an R140/R260 (even a R225), I'd have thought -- 1991/1992 seems very late for the type of experiements in that report and 1994/95 (the RiscPC era) exceptionally so, especialy as Acorn was pushing AUN and Ethernet by then over Econet…

I suspect that is the ancestor of a thing I know as "ecofs".

EcoFS, when I received it under the counter on a tape in about 1996, was a C program running on the R140. I subsequently ported it to FreeBSD and enhanced it in various ways.

The source code had CVS logs of its recent history and a file-head comment:

Code: Select all

/*
    Title    :  fs - root module of Econet fileserver emulator
    Author   :  Brian Knight, Acorn Software Tools Group
    Date     :  Mon Mar  3 13:56:23 1986
    LastEdit :  Tue Jul 28 16:31:44 1987 by brian
    C version:  Thu Oct 12 1989 Brian Knight, Active Book Company Ltd.
*/

....

 * Revision 2.0  89/11/12  21:38:20  brian
 * Initial revision of C version

That, and various other comments suggests it was originally written in Modula2 and later converted line-by-line to C.

There's a 'cnotes' giving the spec at the point I received it, but also a 'notes' which appears to describe the Modula2 version:

Code: Select all

Econet Fileserver Emulator
--------------------------

This is a Modula2 program which emulates a level 3 Econet fileserver, using
a subtree of the host operating system's filing system to store its files and
directories.  Most of the code is intended to be portable (though has only
been run on Vax Unix 4.2bsd so far).  All of the operating system dependent
code is in a single module (FSOSInt).

The following things are not currently implemented:

- Command lines "*CAT", "*LOAD" and "*SAVE"
- Code 26 - read disc free space
- Code 28 - set fileserver date
- Code 29 - create file
- Code 30 - read user free space
- Code 31 - set user free space

Differences
-----------
- '(', ')' and ',' are allowed in names.
- Wildcards cause a scan of any component of a path name, not just the last.
  e.g. *ACCESS *.* WR/R   will affect all objects in subdirectories of the 
  current directory, not just those in the first subdirectory.
- Directory entries are not sorted into alphabetical order.

Bugs
----
- You cannot have fileserver objects called '/' or '//'.
- Directory names are always converted to lower case
- *USERS doesn't hide privileged users
- Nothing stops you creating a directory with >256 entries - these cause
  trouble in Examine (1 byte offset).

So probably it originally ran on the Vax with some other machine (likely BBC) proxying access to the Econet. For my FreeBSD version, I wrote a driver with the same interface as the R140/RiscIX Econet driver, but I also built a version that proxied the Econet functions over a TCP connection to another machine (FreeBSD or RiscIX) where the actual Econet interface was - thus allowing one-hop access to the same fileserver rather than the performance hit of traversing multiple Econet bridges. My proxy didn't share any historic code, but almost certainly worked much the same way as the original Vax version.

[arg]

My other comment on the overall "environment" is that the stuff discussed so far started in sometime around 1983-84. Up to that point, it was very much 'eat your own dogfood' so far as software tools went, while lots of stuff was still manual.

Certainly for everything up to and including the BBC, circuit diagrams and PCB layout were all manual pencil/tape on film.

Electron ULA was hand-drawn schematics plus some kind of tool that ran on the mainframe at the University Computer Lab. There was a tool being written summer of 1982 for doing ULA physical layout graphically on BBC micros, but I am not sure if it was ever directly used in anger; maybe it went on to be the basis of the tools described above by Graham Toal.

[Coeus]

We must get that 6809 MMU design plotted out!

Here's a PNG version from hp2xx

[BigEd]

Oh yikes that’s unusual. Thanks though!

[BeebMaster]

I'm interested in what Acorn's admin/secretarial setup would have been throughout the ages, things like correspondence, press releases, drafts of manuals etc, maybe even internal only documents, like employment contracts, holiday request forms, memos, expense claims forms, etc. Were they using their own computers, at least from the point when they had invented computers which could do word-processing, and software, and what printers were used?

My earliest letter from Acorn is this one, which looks to have been done on what we used to call a desktop publisher of the day, hopefully running on an Acorn.

[paulb]

I'm interested in what Acorn's admin/secretarial setup would have been throughout the ages, things like correspondence, press releases, drafts of manuals etc, maybe even internal only documents, like employment contracts, holiday request forms, memos, expense claims forms, etc.

We know that they used Longman Logotron's Eureka for spreadsheets and evidently Lotus 1-2-3 before that:

"Acorn says Eureka", Acorn User, December 1993.

[arg]

I'm interested in what Acorn's admin/secretarial setup would have been throughout the ages, things like correspondence, press releases, drafts of manuals etc, maybe even internal only documents, like employment contracts, holiday request forms, memos, expense claims forms, etc. Were they using their own computers, at least from the point when they had invented computers which could do word-processing, and software, and what printers were used?

In 1981/82, the secretaries still had electric typewriters, and dictation by executives to be typed up by the secretary was still common practice.

AFAIK, at that point there were no computers other than Acorn ones in-house. Payroll was done on a mainframe by a bureau - we got payslips that had been printed on some kind of lineprinter, but not in house. There was a Telex machine in reception and telexes were sent by typing on it. At the end of 1981, when Acorn had just moved into the Fulbourn Road office, there were (Econet-equipped) Atoms everywhere but BBCs were still scarce and you had to justify having one; that gradually switched during 1982. There also wasn't any software for BBCs to start with - if you had a BBC, it was as an adjunct to the System 3 that you used for serious work (connected to the BBC user port with a ribbon cable that provided some kind of filesystem on the BBC using the S3's floppy).

Memos were typed by the secretaries on typewriters and photocopied. I remember claiming some expenses, but not sure what kind of form I filled in: probably it was a photocopied form and I filled it in with a pen. I never had an employment contract nor a holiday request form. There were not many secretaries; tech people did their own documents.

Technical documents were produced in text files with markup as discussed in this thread. The E editor ran on System 3 and later the BBC; the processor for the markup language was built-in and only supported the rather obscure but very fast dot matrix printer that sat on the end of the landing (having previously stood in the Gents at 4A Market Hill). I think the manufacturer was "BNY". I have vague recollection of support for other printers - there was hardware support for some kind of daisywheel with a card in the System 3 (and an enormous ribbon cable connecting to the printer) - maybe there was support for this in E but I'm not sure. I later often saw documents printed on other printers with visible BNY control codes that the printer didn't support! SJ Research later expanded support for that markup language and wrote drivers for it on centronix-interfaced daisywheel and eventually Apple Laserwriter. Meantime Acorn mostly drifted away from E. The Edit editor (on Master, and the Pascal system in BBCs, and standalone in-house on BBCs but maybe not sold in that format?) had its own similar but different markup language, but probably limited usage. Around that time, product manuals were contracted out to technical authors (Baddesley Associates?) with their own systems. View probably took over for internal non-technical use, but that was after my time and I don't know details.

Internal documents like the "MOS Docn" were in E format and printed on the BNY.

Many documents were produced for the BBC; some would have been done on typewriters (such as those revealed by Richard Russell in the 40-years thread), but some must have been done electronically. I've pulled from my filing cabinet a Teletext/Telesoftware spec dated 11/5/82 - photocopied and bound for a meeting with the BBC. The main part is done on a daisywheel printer; just possible that it was done on a typewriter, but the paragraphs are justified (Herman's secretary had a rather sophisicated electric typewriter, but I don't think it did justified paragraphs) so I think it was done in E and we somehow had a means of printing posh copies to take to the BBC. The appendices of that document are in pen-and-ink!

I'm not sure what John Coll used for the BBC User Guide, but I wouldn't be surprised if it was on actual paper and re-keyed by the printers: this was the early days of computerised typesetting and you couldn't send your own tape/disc to a printing firm to have it typeset on their equipment without being blacked by the NGA union - they'd insist on re-keying it. Remember the Murdoch/Wapping stand-off with the print unions wasn't until 1986.

[paulb]

And as for what else you might use a VAX for, it is highly plausible that it would be useful for software development. Since Acorn seems to have used Sun machines internally, I could imagine that a VAX might have been running Ultrix and be particularly amenable in providing a software development environment.

I don't think I posted this article before, but I remember finding it and then filed it away under National Semiconductor topics, for some reason:

"The Way It Should Be Done", Unix Review, June 1985.

It describes Lucasfilm Games and their development of microcomputer games in a VAX environment. I wonder if any of the observations in the article are pertinent to what went on at Acorn.

[BigEd]

Nice article! Reposted over here.

[gtoal]

We must get that 6809 MMU design plotted out!

Here's a PNG version from hp2xx

That looks surprisingly familiar! I think the colours may be different from what we used though. There's not much benefit of using that design nowadays though as it was primarily a counter for refreshing the DRAM whereas now we'd use a static memory (as indeed I did myself after testing the dram prototype).

[gtoal]

Thanks Graham and Paul.
Particularly interesting is the OS the VAX was running, typically this would be VMS or Ultrix (DEC Unix based on BSD).
I'm running a MicroVaxII at the moment and am tempted to try and recreate a basic environment as was at Acorn. Connectivity to Econet, BBCs as terminal devices and the compiler setup etc.

It was a Vax 750 running BSDI and had something like 16 or 24 serial ports to BBC terminals.

[paulb]

It was a Vax 750 running BSDI and had something like 16 or 24 serial ports to BBC terminals.

That's some quality multi-user computing! Reading retro topics recently, some remarks were made about VAX usage at other companies. It seems that Sinclair and Psion also had VAX-11/750 models, but both Sinclair's and Psion's ran VMS.