Any info on Acorn's own computing environment?

[piersw]

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.

Much later on, in the mid to late 90s, Acorn had three Sun workstations that I can recall, all running Solaris. Valhalla was the main server that ran email and NFS mounts. Starman was used for cross-compiling Java, but Valhalla ended up being used for compiling too as it was much faster (that was understandably unpopular). I've no idea what Starman (and the other whose name I've forgotten) was originally purchased for, though it was someone's personal workstation immediately before Java.

Most RISC OS developers used RiscPCs. NCOS was forked, continued to be largely developed on RiscPCs, but moved to cross compilation for ROM builds (for speed). The reunification inevitably took forever. Galileo was developed on individual PC workstations (presumably Linux). All management and admin staff I knew of used Windows PCs (for the apps), and payroll was outsourced. Even lowest tier management ended up with Windows PCs as they were fond of gantt charts.

[jgharston]

Dredging my memory, back in AFE in Hong Kong the original BBC-based stock market terminal software was originally developed on 6502 CoPros on Beebs, and then moved to 65Tube on RISC OS. The RISC OS stock market terminal software was developed on RISC OS 2 machines, the PC-based version was initially developed with PCEm for writing and hardware PCs for testing.

We had a RISCiX machine running as a server for shared services, but almost all user data - including development sources - was on the personal machines of the developers. When I was there we had started a rudimentary form of centralised source archiving. Checking my network map all the admin staff were on A-series RISC OS machines.

[paulb]

Galileo was developed on individual PC workstations (presumably Linux).

Were any documents or resources related to Galileo preserved or even published? There were promotional articles published by Acorn directly (or ART, or whoever) and also at least one article in Byte, but beyond that, nothing of substance that I recall. As I may have remarked previously, the more cynical of us who were following Acorn's activities in the mid-1990s, by that time being regularly treated to rather aspirational remarks by its management, wondered if it was much more than vapourware, but I have read assertions of it being quite far developed.

[piersw]

Were any documents or resources related to Galileo preserved or even published? There were promotional articles published by Acorn directly (or ART, or whoever) and also at least one article in Byte, but beyond that, nothing of substance that I recall. As I may have remarked previously, the more cynical of us who were following Acorn's activities in the mid-1990s, by that time being regularly treated to rather aspirational remarks by its management, wondered if it was much more than vapourware, but I have read assertions of it being quite far developed.

I doubt anything was preserved, unless by the individuals involved. I wasn't involved, but sat near them as part of the NC team, but on RO since it was what was going to ship sooner.

ART's and the Clan's comments regarding Galileo were very different to reality, but that doesn't mean it wasn't being heavily invested in. It was just very obviously (to anyone inside) intended to be solely for its embedded video market and would never be useful for the desktop/enthusiast market. I don't really understand why ART/Clan made the comments they did - I cringed each time I heard it mentioned.

I don't recall most of the details of it, but when I joined in 1996 there was a functional, but proof-of-concept QoS kernel. The team ramped up in size rapidly but I can't be certain when it ramped down. The vast majority, if not all, of the engineers were new recruits, and were really quite heavily anti-RISC OS (probably rightly, given their target market), and reimplemented everything from scratch.

I only really remember seeing their vector graphics library and font manager, which were clean-room implementations in C (or C++?). The vector graphics library had no input from the existing Draw module, nor did the font manager which didn't yet implement hinting and used non-RO fonts (so presumably would have relied on standard font shapes rather than 45 degree circles). However, it worked like modern antialiased graphics libraries (and GDraw) rather than supersampling like the RO font manager.

I think Galileo died when Intel canned the SA1500 and Acorn's SA1501 was useless. Firepath (sort of the SA1500's successor) was always intended to run Linux, I seem to recall. The [software part of the] Firepath team were primarily the former Galileo team (minus the kernel engineer who chose to leave), plus a very small group of RO people, including me. After various funding options didn't pan out, the split happened with Firepath going to E14 and the rest to Pace.

Since my last post, I remembered the Galileo engineers' PC workstations ran x86 Solaris not Linux.

[edited to clarify I meant software for Firepath, as I wouldn't wish to annoy the Bristolians here.]

[paulb]

I doubt anything was preserved, unless by the individuals involved. I wasn't involved, but sat near them as part of the NC team, but on RO since it was what was going to ship sooner.

Someone who claimed to have worked on Galileo did surface on a Wikipedia talk page, but I haven't been able to find that again.

ART's and the Clan's comments regarding Galileo were very different to reality, but that doesn't mean it wasn't being heavily invested in. It was just very obviously (to anyone inside) intended to be solely for its embedded video market and would never be useful for the desktop/enthusiast market. I don't really understand why ART/Clan made the comments they did - I cringed each time I heard it mentioned.

I guess that Acorn/ART weren't really focusing on new product development for the company's traditional markets, but I imagine that they didn't want to communicate that to customers and enthusiasts. So, there were always suggestions that the investments in Online Media and the network computing division would somehow be transferable to Acorn's computer line-up, although I seem to recall that the potential impact of Galileo was very cautiously described. But certainly, any ideas that Galileo would rectify RISC OS's deficiencies were left to enthusiasts and their fevered speculation.

Do you have any insight into claims about Acorn using IBM's Workplace OS, which is clearly what Peter Bondar was alluding to at one point? It sounded like pie in the sky to me, although such a project would have made sense in its own way, at least hypothetically. Obviously, Workplace OS imploded, but Acorn were very slightly exposed to IBM's technologies with the SchoolServer and PowerPC was the hot new thing at the time. It was also becoming clear that, like Apple, Acorn's operating system was in need of replacement or radical renewal.

I don't recall most of the details of it, but when I joined in 1996 there was a functional, but proof-of-concept QoS kernel. The team ramped up in size rapidly but I can't be certain when it ramped down. The vast majority, if not all, of the engineers were new recruits, and were really quite heavily anti-RISC OS (probably rightly, given their target market), and reimplemented everything from scratch.

For a portable operating system, I suppose it made a lot of sense to start from scratch. I remember that QoS was another hot topic at the time and sat nicely with all the talk about ATM and its QoS features that had accompanied Online Media coverage.

I only really remember seeing their vector graphics library and font manager, which were clean-room implementations in C (or C++?). The vector graphics library had no input from the existing Draw module, nor did the font manager which didn't yet implement hinting and used non-RO fonts (so presumably would have relied on standard font shapes rather than 45 degree circles). However, it worked like modern antialiased graphics libraries (and GDraw) rather than supersampling like the RO font manager.

Very interesting. I have said before that Acorn should have reimplemented their user interface toolkits and frameworks in order to target Unix and thus offer themselves and their customers a migration path from RISC OS, not at the lowest level (like Torch, who admittedly didn't have much to base their earliest efforts on) but on top of emerging standards, because there were clearly talented people at Acorn developing things like Draw, the font manager, and so on.

I think Galileo died when Intel canned the SA1500 and Acorn's SA1501 was useless. Firepath (sort of the SA1500's successor) was always intended to run Linux, I seem to recall. The [software part of the] Firepath team were primarily the former Galileo team (minus the kernel engineer who chose to leave), plus a very small group of RO people, including me. After various funding options didn't pan out, the split happened with Firepath going to E14 and the rest to Pace.

I'll have to dig around for SA-1501 information now! Odd that what was ostensibly a portable product was sunk by the demise of a specific product, though.

Since my last post, I remembered the Galileo engineers' PC workstations ran x86 Solaris not Linux.

I suppose that for a supported environment, Solaris x86 made most sense. However, I remember Linux running on Pentium 90 systems already in 1995 being pretty viable replacements for traditional workstations.

[piersw]

Someone who claimed to have worked on Galileo did surface on a Wikipedia talk page, but I haven't been able to find that again.

Well, they did exist. Quite a few were contractors, too (eg. the vector graphics guy whose name I forget despite sitting next to him for a year).

Do you have any insight into claims about Acorn using IBM's Workplace OS, which is clearly what Peter Bondar was alluding to at one point?

Never heard of it. Peter claimed all sorts, and I wasn't in ART, but I'm certain not a single minute of engineer time was spent looking at other OSes. Galileo, TAO, or anything. Peter was trying to find a purpose for ART, which he eventually did with RO being embedded inside CE devices and STBs/TVs.

For some time they had a laughable (as in they would laugh when saying it) hope Galileo would end up magically being something RO would slot on top of, perhaps with a thin emulation layer. But the ART and Galileo teams never spoke (not even the engineers). Remember the company was planned to be split from Online Media onwards, so the fact ART moved back to the same building with OM/AcornNC was purely to save money with long leases.

The same long lease that many people at Pace blamed for the closure of their Cambridge office, and the office lay empty for ten years looking for a replacement tenant.

I'll have to dig around for SA-1501 information now! Odd that what was ostensibly a portable product was sunk by the demise of a specific product, though.

The funding dried up and Galileo's guaranteed market disappeared. FirePath was self-funded and Linux was more mature.

[paulb]

Someone who claimed to have worked on Galileo did surface on a Wikipedia talk page, but I haven't been able to find that again.

Well, they did exist. Quite a few were contractors, too (eg. the vector graphics guy whose name I forget despite sitting next to him for a year).

I don't doubt their existence, but I just need to remember which page it was. Then I might discover the name.

Do you have any insight into claims about Acorn using IBM's Workplace OS, which is clearly what Peter Bondar was alluding to at one point?

Never heard of it. Peter claimed all sorts, and I wasn't in ART, but I'm certain not a single minute of engineer time was spent looking at other OSes. Galileo, TAO, or anything. Peter was trying to find a purpose for ART, which he eventually did with RO being embedded inside CE devices and STBs/TVs.

It sounds like a gradual withdrawal from Acorn's traditional markets, although there was the whole enthusiast marketing initiative, which seems a bit cynical now if you think about it. Some Web-related technologies did float across to the traditional audience, but it would have needed a concerted effort to integrate something like Taos into Acorn's platform strategy, and I guess that there wasn't the business case for it.

For some time they had a laughable (as in they would laugh when saying it) hope Galileo would end up magically being something RO would slot on top of, perhaps with a thin emulation layer. But the ART and Galileo teams never spoke (not even the engineers). Remember the company was planned to be split from Online Media onwards, so the fact ART moved back to the same building with OM/AcornNC was purely to save money with long leases.

I remember that OM was to be spun out, but this was abandoned for various reasons, perhaps related to market buoyancy but also the lack of a path to broader success. Video on demand was yet another fashionable technology that would supposedly be rolled out pervasively, and there was a certain amount of fibre being laid opportunistically, but it seems rather naive now to think that something like the rather limited Cambridge and London trials of the technology - step #1 in the business plan - would rapidly proceed to step #3 without that irksome step #2 being elaborated. After all, it took quite some time for something like ADSL to become widespread, and decent quality streaming video is a relatively recent thing even now.

The same long lease that many people at Pace blamed for the closure of their Cambridge office, and the office lay empty for ten years looking for a replacement tenant.

I'm far enough off topic already without getting into the effects of the property market on commerce. It all rather seems that Acorn became a kind of incubator for a bunch of other businesses in the end, which may have been lucrative for some, but is a shame given that there would have been enough expertise around its traditional activities to have some kind of viable ongoing business, even if it were merely doing design and product development for other companies, which was presumably the essence of ART.

Of course, for the customers, they were left in the capable hands of Xemplar for their ongoing and future needs. Add sarcasm according to taste. Unlike Acorn, some of its developers did have a cross-platform strategy and thus permitted a kind of migration path. I'm guessing that products like SchoolServer had little input from Acorn and were done under contract, so I can't even bring the discussion back to the topic by suggesting that there were a handful of developers at Acorn using NT on PowerPC, let alone anyone doing the PReP/CHRP stuff that Peter Bondar mentioned.

Many thanks for the behind-the-scenes insights!

[piersw]

It sounds like a gradual withdrawal from Acorn's traditional markets, although there was the whole enthusiast marketing initiative, which seems a bit cynical now if you think about it. Some Web-related technologies did float across to the traditional audience, but it would have needed a concerted effort to integrate something like Taos into Acorn's platform strategy, and I guess that there wasn't the business case for it.

It certainly wan't intended to be cynical - they made a business case for Phoebe and it was built. Unfortunately, it didn't work first time and the timing was bad, so the division closed.

Java floated across, though was never as polished as I would have liked. I believe the source licence that Acorn had to pay was in the order of £1m, in late 90s money, plus several man years of development (Oracle did fund the initial development, though their Java licence was theirs for the NC). I haven't a clue how many were sold, but it wouldn't have made a dent in the costs.

I also don't recall any RO engineers ever being interested in replacing RO on desktops. It was a big enough pain getting apps upgraded to be SA compatible, so anything else would mean no apps.

[paulb]

It certainly wan't intended to be cynical - they made a business case for Phoebe and it was built. Unfortunately, it didn't work first time and the timing was bad, so the division closed.

It says quite a bit about the state of the company and the market, though. Giving the workstation division, as I believe people called it, just one shot to get something out there. Shades of the Acorn Business Computer, I suppose. The unfortunate thing is that there were plenty of opportunities for all sorts of devices running Linux on ARM and other non-x86 architectures towards the end of the 1990s and into the 2000s. It is revealing that one of the fairly well-known sites hosting RISC OS content (or "RiscOS" as they call it) is named after such an appliance (admittedly MIPS-based).

Java floated across, though was never as polished as I would have liked. I believe the source licence that Acorn had to pay was in the order of £1m, in late 90s money, plus several man years of development (Oracle did fund the initial development, though their Java licence was theirs for the NC). I haven't a clue how many were sold, but it wouldn't have made a dent in the costs.

And then, Oracle and others went and used NetBSD instead for the next version of the Network Computer operating system, which was a more amenable target for something like Java and also more practical for an Internet appliance in many respects. In a sense, that technological direction is one that Acorn should have been pursuing anyway, as people like Gordon Taylor had been saying from when the Risc PC came out, although I would have suggested that they might have thought about it earlier still when scheduling RISC iX for discontinuation.

I also don't recall any RO engineers ever being interested in replacing RO on desktops. It was a big enough pain getting apps upgraded to be SA compatible, so anything else would mean no apps.

There would have needed to be a transition from RISC OS as it was to something giving a similar experience, but that would have required an investment that evidently wasn't made. Recently, I have been looking into the Unix activities of Atari and Commodore who saw the Unix market as one avenue to grow out of their steady decline. It seems that Commodore's approach was a bit like Acorn - offer a Unix and leave the rest up to the market - but Atari did at least bundle some Motif application builder development tools, even if the audience seems only to have been actual developers, potentially bound by a pre-release licence, and didn't really offer any migration path from Atari's own proprietary operating system.

[paulb]

I don't doubt their existence, but I just need to remember which page it was. Then I might discover the name.

It was Chris Berry who, as Chrisbtoo, made some remarks on the Talk:Acorn Computers page. The Galileo (operating system) page perhaps reflects the public perception of the effort, as opposed to the corporate reality, but then the former is usually all that the rest of us have to go on.

Then again, corporate messaging has a role to play in making sure that the public is correctly informed, and if that fails to occur, who is really to blame for that? If nothing else, though, the selection of references for that page and the framing of Galileo in those articles is rather informative in itself. And sometimes, the perception of a topic, however inaccurate, is a phenomenon in its own right.

[piersw]

It was Chris Berry who, as Chrisbtoo, made some remarks on the Talk:Acorn Computers page.

He was a Symbian fan, proving my point about anti-RISC OS perfectly...

The Galileo (operating system) page perhaps reflects the public perception of the effort, as opposed to the corporate reality, but then the former is usually all that the rest of us have to go on.

I think that article's a pretty good reflection of the project. It was intended to replace RISC OS on all Acorn's embedded devices (STBs, NCs, etc), but a pipe dream for anything with a desktop - unfunded and no intention of resourcing it, as far as I'm aware. The "Fate" section of the article effectively says that, if you read "Acorn's customers" as meaning Oracle, Funai, etc, not end-users of desktop machines.

I'd never heard of the Acorn/Psion/Symbian rivalry before. I remember them talking to Acorn to consider licensing our Java, though it didn't go anywhere as Acorn's knowledge and IP was all in the RISC OS port.

[flibble]

digging through archives I found an email about using Cadence on Sparc machines for PCB design in 1992

From: ABienek (Alex Bienek)
To: Hardware
Subject: SPARC station requirements
Message-Id: <2AFA962F@ABienek>



With the installation of the Cadence design tools we will be moving into a
new method of working on the CAD system.

Previously the Valid tools were used almost exclusively for schematic entry
for the sole purpose of making PCBs. This meant that use of the terminals
would generally be limited to a week or two intensive schematic capture
followed by sporadic updates during the lifetime of the project. This meant
that providing three terminals, located away from the engineers normal
design environment, although not ideal, was workable.

Having said that, I always maintained that much more use could have been
made of the system had the terminals been distributed among the engineers,
providing terminals on desks as required. This has been borne out by the
experience within the Victoria project where Bernard has a terminal on his
desk on which to enter the Victoria schematic and now considers such direct
access to a terminal "absolutely essential".

With the new tools a great deal more of the design process can be carried
out on the CAD system, including block level system design, high level
textual modelling for simulation, low level textual design for simulation
and synthesis, and schematic capture for simulation and PCB layout.
This will involve a great deal more of the engineers time spent using
the system, and conversely the greater the access to the system the
more it will get used, for sketching and testing ideas, developing
system models, and logging changes.

When the changeover to SPARC IPX stations is complete there will be three
available. However, predicted requirements are:-

Ian Nicholas ASIC development Dec '92 - June '93
Alex Bienek ASIC development Jan '93 - July '93
Bernard Siddle Victoria PCBs Nov '92 - June '93
A.N. Other OPC/Stork/???? Q2 '93
Bruce/Vic et al Misc.

For these reasons I believe this department should push for at least
one more IPX station, possibly two to give good coverage.

Comments ?, agreement ?, disagreement ?

Alex.

[BigEd]

Nice find - thanks!

(In my own world, outside of Acorn, I do recall the transitions from one computer per building, to one per team, to one per office, to one per desk. At each point it seems like a terribly expensive thing to do and difficult to justify.)

[flibble]

(At each point it seems like a terribly expensive thing to do and difficult to justify.)

You're not kidding! here's another mail where they break down the costs. The software costs are terrifying.

Date: Thu, 19 Nov 92 08:15:09 GMT
From: pdawson (Peter Dawson)
To: mhill, hardware
Cc: gmacfarlane
Subject: Latest on CAD/Cadence costs
Auto-Acknowledge-To: pdawson (Peter Dawson)
Message-Id: <2B0B4D0D@pdawson>

Graham has been talking to Victoria team members, and we have had a
preliminary meeting with Mike. We are still struggling to get an idea of
both Victoria's needs and the longer term estimated workload.

However, we felt that you would be interested in more detail of the
cost implications. All of the relevant Cadence software will be available
under floating licences from the beginning of 1993.

Costs (current estimates, including discount) :

Sun IPX - £9.5K

Composer Schematic Entry - £10.1K (floating licence)

Composer Verilog-XL option - £24.7K " " "

Composer PLD option - £10.4K " " "

Library (Logic Automation) licence " " "

So a basic Schematic Entry node (IPX/Composer) costs £19.6K plus approx
£2.5K maintenance.

So, for a realistic set-up:-

* 5 extra basic nodes
* 2 extra Verilog licences
* 1 extra PLD licence
* Library licence

the costs are : £98K + £49.4K + £10.4K + £10K = £167.8K
plus approx £24K maintenance

These figures assume that we only need one Synergy licence and do not
cover the question of extra simulation models.

Regards,

Peter Dawson

[paulb]

However, predicted requirements are:-

Ian Nicholas ASIC development Dec '92 - June '93
Alex Bienek ASIC development Jan '93 - July '93
Bernard Siddle Victoria PCBs Nov '92 - June '93
A.N. Other OPC/Stork/???? Q2 '93
Bruce/Vic et al Misc.

Victoria is the Risc PC development effort, is it not? I guess we can see another facet of the above timeline in the following:

"From Archimedes to Risc PC"

VIDC20 would have been available first, I guess, although it seems like the ARM6 family might have been available by the end of 1992. Here's the ARM page of a rather useful reference summarising available processor products of the era:

"EDN's 19th Annual µP/µC Chip Directory", EDN, 26 November 1992.

The Apple Newton which used the ARM610 was eventually shipped in August 1993, apparently. The 3DO Interactive Multiplayer - another notable ARM6-based product - launched in October 1993.

[BigEd]

Thanks for the info - I'd searched 4corn for "Victoria" but now I see why that was fruitless - "Vicky"!

[flibble]

Here's a document I found that explains Acorn's econet and ethernet network in 1990, over 400 machines networked!

[paulb]

Here's a document I found that explains Acorn's econet and ethernet network in 1990, over 400 machines networked!

Just the statistics go some way to satisfying our curiosity! I see that the Apollo machines feature in the Ethernet network on the final page. Cadence solutions were mentioned a few posts ago, and I wonder if the Apollo workstations had been refreshed or needed replacing by this time. At around this time, Apollo had been acquired by HP and there were aspirations to encourage Mentor Graphics, whose products Acorn could have been using, to extend their commitment to HP's own workstations:

"The Workstation War Heats Up", HP Professional, July 1989.

There's also an IBM System/38 which looks like a classic office automation system, but perhaps it could have been used for more exotic purposes. We also see that, as someone claimed previously in a discussion here, Acorn's A680 prototypes were being used internally alongside their R140 products. There is also confirmation that the VAX machines were 11/750 models running Unix.

[BigEd]

Excellent document flibble! Only read half so far... then need to follow paulb's links.

[flibble]

Well, we've reached 1997, and the era of Acorn 'eating its own dog food' is well and truly over. Each employee to get a PC and Office.

[piersw]

Well, we've reached 1997, and the era of Acorn 'eating its own dog food' is well and truly over. Each employee to get a PC and Office.

Not exactly. Only new staff. Given they'd stopped recruiting, it didn't really change much.

I've forgotten the guy's name, but around that date they brought in a new head of CSG who announced the plan. He wasn't universally popular because he appeared to be quite anti-RISC OS. Given the direction of the company, it was probably an unfair opinion as he was doing what he was asked, but he didn't stay long.

[jgharston]

What cannot be done with the same ease is for the machine itself to spot that an update is available - a holy grail that the CSG has wanted to implement for RISC OS machines.

Odd. I implemented that fairly simply on the RISC OS machines at AFE.

[BigEd]

Do I detect from the various memos that the IT group (3 people in '97, supporting 245 users) was initially(!) called ISG and then later called CSG?

[piersw]

What cannot be done with the same ease is for the machine itself to spot that an update is available - a holy grail that the CSG has wanted to implement for RISC OS machines.

Odd. I implemented that fairly simply on the RISC OS machines at AFE.

The proposal (Jeremy sounds plausible, but I can't remember) that caused most upset was that engineers were told their computers were to be locked down and we were unable to change anything. No software could be installed without permission, and all updates were to be pushed by the IT department (I don't recall its name so I'll invent a new one).

There's a huge amount of logic behind this, and was the norm in major corporate companies, but engineers don't like change. RISC OS was incapable of being locked down (no security model), and engineers wanted complete control of their machines (how else can you develop modules?).

But it was unpopular with the engineers. The policy was enforced on Windows (which hardly any engineer had). I was working on Java and got a new source and binary drop every week or so. For a while, I needed to open a support ticket for every binary that I downloaded so someone could come and install it for me. Thankfully poor Phill gave in and gave me the password after a bit.

Despite what it proposes, I don't recall many engineers receiving a Windows NT-powered PC - I had one because my requirements were unique - Java, and that I had to produce gantt charts regularly. Certainly most new recruits only had one computer, and it was either RISC OS (for 'downstairs', ART) or a PC running x86-Solaris (for 'upstairs', Galileo/FirePath). Even existing upstairs engineers kept using RISC OS until after the company split (I'm pretty sure Sophie and Tim only made the switch after we moved to The Quorum).