Who was using RISC iX back in the day?

[paulb]

Now that we have a dedicated RISC iX forum, I thought it might be appropriate to drop a load of questions in the hope that we might learn a bit more about RISC iX and how it was used back in the day. Many of us will have been aware of its introduction with the R140 in 1989, and some of us may have read coverage of the R-series models and the software available for it, with A&B Computing probably providing most of that coverage through Jon Vogler who had been a Unix enthusiast and Torch Triple X user prior to the release of the R-series.

So, first of all, who bought the RISC iX machines originally? Did individuals buy them? And what kind of organisations bought them? What was their motivation for buying them? Was it to acquire a cheap workstation, or was the idea to offer some kind of Unix exposure to existing Beeb and Archimedes environments? Did they put them on local area networks or even on the Internet itself?

For those buyers, why did they choose these machines instead of those from other vendors? Was it perceived value for money, familiarity with Acorn, or was it the possibly intimidating prospect of dealing with big commercial vendors like DEC, HP, IBM and Sun? Were they happy with their purchase or did they experience some buyer's remorse? If they bought the R140, did they upgrade to the R260? If they hadn't been Unix users before, did they stick with Unix afterwards?

So many questions! One recollection from a slightly later time from me, then. In my university department, one of the system administators had an Acorn tote bag from some academia-focused computing conference or exhibition, which intrigued me and some of my fellow students who were Acorn enthusiasts. This was probably 1994 or so, and the sysadmin told us that Acorn had been showing off their Unix machines a few years earlier, handing out those bags. Acorn's machines didn't seem to have left an impression on him, leaving him with little to say about them, however.

[steve-h]

From the bloke I got mine off (with some minor editing):

Little bit of history on the machine - it was owned originally by a university - I forget which - and the hostname 'prospect2' refers to its original purpose running a careers advice application. I bought it from a professor in south london

[Boydie]

Probably not the usage you were looking for, but I was running the nsict version on an A310 (4Meg, ARM3, 240 Meg Quantum Fireball SCSI) somewhere around 1994. Didn’t do anything particularly serious or useful with it.

Also had an account on nsict’s R140 (and later A540 when Eidos cleared out their machines) around the same time.

[tom_seddon]

Newcastle university had a set of RISC iX machines in the mid 90s, i think? A couple of my friends had used them, and they mentioned them because they knew i was into Acorn stuff. They were replaced with Linux PCs, probably a more sensible choice.

--Tom

[arg]

I suspect the typical customer was someone who wanted a cheap Unix machine, particularly if they liked BSD, and intended it for running their own software or generic open-source stuff (the open source movement hadn't really taken off at that point, but there was still some around).

We (SJ Research) were definitely not a typical customer, but probably our motivations were similar to those that were.

We'd started with a Sun 3/50 which we acquired for a song at a bankruptcy auction sometime around 1985/86 (I think my colleague had gone looking for something prosaic like filing cabinets and came back with the Sun - "is this any use?"). So we used it as a nice big screen workstation for writing unrelated code, and started using it as a Sun-NFS fileserver.

The Sun then became the heart of the Interspan email system - which had modems each driven by a cheap PC with an ethernet card and the Sun as fileserver and controlling things with Sun-RPC to the slave PCs (Sun providing a software toolkit for IP + NFS + RPC on PCs under MS/DOS). Customer systems at the other end of the modem connections were BBC micros, as was the very earliest central system: this was the scale-up system design.

But that wasn't sustainable as a production system - if the Sun died, we'd be in the soup. We had experience of Unix on 386 PC hardware from the aborted SJ Application Server project, but that was unattractive - SVR3 was rubbish in comparison to BSD (I continue to hold that opinion about SysV vs BSD, but it was particularly true in the case of SVR3), the licences were expensive, and sufficiently powerful 386 PC hardware not cheap. So when the R140 came out it was ideal for us: it could run all the same software we'd built for the Sun, and the facility to trade parts with other Archimedes machines we already had offered much more resilience than any service contract on a Sun.

The R140 was useless as an X-windows workstation machine - not enough RAM, not helped by the large page size, and further not helped by the fact that if you were forced to page to disc, the hard disc interface performance was horrible. However, if you just wanted it as a headless server it would serve NFS and run a few other programs very nicely indeed.

The R260 was a much better machine - now it had enough RAM to work properly, and the SCSI disc interface had much better performance.

I'm not quite sure what happened at SJ; I moved on at the same time as the Interspan service shut down, but I thought the R140 remained in use as a useful server machine, and I thought that when SJ went bust a year or so later I picked it up (not illicitly - we bought the residual assets off the Receiver and each grabbed what we thought was worth having of what was left). But the machine I actually have in my possession is an R260, and the contents of the disc show that it had been set up again as a fileserver but primarily as a mail server for early internet email (UUCP I think). So again it had been bought as a cheap way of getting a machine that would run BSD Unix and the tools that went with it. A year or two later and FreeBSD was on the scene on PCs and the niche that Acorn was selling to disappeared.

[paulb]

But that wasn't sustainable as a production system - if the Sun died, we'd be in the soup. We had experience of Unix on 386 PC hardware from the aborted SJ Application Server project, but that was unattractive - SVR3 was rubbish in comparison to BSD (I continue to hold that opinion about SysV vs BSD, but it was particularly true in the case of SVR3), the licences were expensive, and sufficiently powerful 386 PC hardware not cheap.

I guess that SCO Unix was the principal SVR3 product on x86 at that point in time. In 1989, it seems that a lot of the effort that went into merging SysV and BSD hadn't been completed, and the x86 platform was not as well served as it was only three or four years later, even when considering only the proprietary Unix products. I see that UnixWare only arrived in 1992, for instance.

So when the R140 came out it was ideal for us: it could run all the same software we'd built for the Sun, and the facility to trade parts with other Archimedes machines we already had offered much more resilience than any service contract on a Sun.

So, I imagine that dealing with the big Unix vendors was rather unattractive. Given my early exposure to DECstations and Ultrix, which was a BSD variant, I have to ask whether you ever considered DEC, or was that thought to be a gateway to potentially spending a lot of money?

The R140 was useless as an X-windows workstation machine - not enough RAM, not helped by the large page size, and further not helped by the fact that if you were forced to page to disc, the hard disc interface performance was horrible. However, if you just wanted it as a headless server it would serve NFS and run a few other programs very nicely indeed.

It rather seems that this was the niche for the R140: as a fairly cheap server that was going to be a lot more reliable as a server than a lot of PC-compatible solutions, especially where Unix technologies were involved.

The R260 was a much better machine - now it had enough RAM to work properly, and the SCSI disc interface had much better performance.

I get the impression from reviews from the era of this being the case. That the R140, while an exciting for Acorn, wasn't really a viable product as an actual workstation, and the excitement (or hype) was a bit more justified with the R260. In terms of performance, it does lag behind the big Unix vendor products of the era, and I don't think the pricing was that competitive - the £5000 price point was readily being matched by Sun and DEC - but perhaps Acorn could have carved out a niche, if nothing more, had they persisted.

Of course, they were rather dependent on ARM's roadmap, at least if they had decided to stick wholeheartedly with ARM, but in another universe there might have been another division spreading their bets on other architectures, maybe joining in with the Advanced Computing Environment adventure, eventually getting spun out from Acorn. It could have been quite exciting.

So again it had been bought as a cheap way of getting a machine that would run BSD Unix and the tools that went with it. A year or two later and FreeBSD was on the scene on PCs and the niche that Acorn was selling to disappeared.

Indeed, the free BSDs and Linux changed everything. Acorn could have coasted that wave as well, especially with the surge in interest in embedded Linux on ARM as well as the way network computers migrated to BSD and other Unix implementations, even when Acorn was still doing business.

[arg]

I guess that SCO Unix was the principal SVR3 product on x86 at that point in time. In 1989, it seems that a lot of the effort that went into merging SysV and BSD hadn't been completed, and the x86 platform was not as well served as it was only three or four years later, even when considering only the proprietary Unix products. I see that UnixWare only arrived in 1992, for instance.

Interactive Unix 386/iX was the SVR3-on-PC we were using. At the time, SCO Xenix was perceived as not-quite-real Unix, I think because the main product was a non-VM Unix running on 286 hardware. Later SCO obviously was more mainstream SVR3 with VM on 386 hardware; I'm not sure if Interactive got there first or were just marketing it better.

Certainly a few years later SCO were dominating the PC Unix space and Interactive had more or less disappeared (actually I now read they were acquired by Sun, but Sun also had Solaris/386).

So, I imagine that dealing with the big Unix vendors was rather unattractive. Given my early exposure to DECstations and Ultrix, which was a BSD variant, I have to ask whether you ever considered DEC, or was that thought to be a gateway to potentially spending a lot of money?

I don't think it was on the radar with DEC being perceived as big-money systems. I did later have exposure to DEC stuff on Alpha (a few years after the time we are talking about I was writing software that had to run on just about all the Unix systems going - RS6000/MIPS/Sparc/DEC Alpha/SCO - but even then my memory is of the DEC being more up-market than the others. Quite likely my perceptions were wrong, but that's what they were.

[paulb]

Interactive Unix 386/iX was the SVR3-on-PC we were using. At the time, SCO Xenix was perceived as not-quite-real Unix, I think because the main product was a non-VM Unix running on 286 hardware. Later SCO obviously was more mainstream SVR3 with VM on 386 hardware; I'm not sure if Interactive got there first or were just marketing it better.

I had forgotten about Interactive Unix, and I imagine that there are other Unix variants I am also forgetting, ignoring Unix-like products like Coherent, of course. Looking at Wikipedia's list of products, I found Venix, although there were a few other early products that didn't really stay the course.

Certainly a few years later SCO were dominating the PC Unix space and Interactive had more or less disappeared (actually I now read they were acquired by Sun, but Sun also had Solaris/386).

Yes, how could I forget Solaris and SunOS on x86?! Here are some Byte articles of potential interest from the era:

"Sun's New Workstation: the Sun386i", Byte, July 1988.
"Two Powerful Systems from Sun", Byte, May 1989, featuring the Sun-3/80 and SPARCstation 1.
"Some Flavors of Unix", Byte, May 1989.

So, I imagine that dealing with the big Unix vendors was rather unattractive. Given my early exposure to DECstations and Ultrix, which was a BSD variant, I have to ask whether you ever considered DEC, or was that thought to be a gateway to potentially spending a lot of money?

I don't think it was on the radar with DEC being perceived as big-money systems. I did later have exposure to DEC stuff on Alpha (a few years after the time we are talking about I was writing software that had to run on just about all the Unix systems going - RS6000/MIPS/Sparc/DEC Alpha/SCO - but even then my memory is of the DEC being more up-market than the others. Quite likely my perceptions were wrong, but that's what they were.

I think old habits, once established, died hard at DEC. Even though the company was built on competing with established players from below, the VAX era entrenched a certain strategic outlook and framed their product lines, which is why the DECstation range ended up being a "skunkworks" effort that went outside the usual operating procedures to be able to deliver something that customers were actually asking for but not getting. Even then, DEC's various factions seemed intent on shutting that product line down and introducing something wholly DEC-designed and presumably gold-plated into the bargain.

Since my own DEC experiences were entirely as a student using machines that other people had paid for, I had no real perception of DEC as a company. However, it was enlightening to learn more recently that the DECstation machines were quite reasonably priced compared to, say, Sun's workstations, and I imagine that their introduction could have been very worrying for Sun. DEC proceeded to fumble their strategy by discontinuing the MIPS-based DECstations, to be replaced by machines based on the wholly DEC-designed Alpha, and to apparently tell their customers to upgrade to those. So you could say that DEC certainly thought their customers had ample amounts of money to spend.

I have to say that considering the competitive situation at the time, choosing BSD as the basis of RISC iX was a shrewd move by Acorn, and there does seem to have been an opportunity at the bottom of the market. There were others also looking to get into that market, too, like Commodore, but they really didn't stand a chance with 680x0-based machines in the early 1990s. (I noticed someone's comment on a YouTube video recently about a supposed potential deal between Sun and Commodore for Sun to offer Commodore's Amiga 3000UX as a low-end workstation, but the whole thing sounds completely deluded.)

[paulb]

I have to say that considering the competitive situation at the time, choosing BSD as the basis of RISC iX was a shrewd move by Acorn, and there does seem to have been an opportunity at the bottom of the market.

In another thread, I mentioned the A&B Computing article from March 1989 where David Slight of Acorn is interviewed about RISC iX and the R140. In that interview, the justifications for choosing BSD included the "academic environment" from which BSD came being appropriate to the target market, along with better "integration of networking and other connectivity tools".

Despite the perception that System V was more "commercial", it was also pointed out that products from Sun and NeXT were in both respects "modern high-performance workstations that use BSD". Sun was arguably driving the workstation market at that point, so I think that concerns about commercial adoption based on a choice of BSD were rather spurious.

[jgharston]

But the machine I actually have in my possession is an R260, and the contents of the disc show that it had been set up again as a fileserver but primarily as a mail server for early internet email (UUCP I think). So again it had been bought as a cheap way of getting a machine that would run BSD Unix and the tools that went with it. A year or two later and FreeBSD was on the scene on PCs and the niche that Acorn was selling to disappeared.

That's essentially what we used our R260 for at AFE in Hong Kong. We ran it as the mail server, and a printer server for the (then) expensive laser printer, and NFS fileserver. It was accessible from the RISC OS machines and the PCs, either by Econet or Ethernet.

I did some bugfixing on the mailer program, which may have filtered back to the Cambridge mothership.

[paulb]

That's essentially what we used our R260 for at AFE in Hong Kong.

Did AFE sell any of the R-series workstations to anyone? I seem to recall that the Hong Kong Stock Exchange was a customer of AFE (see "Taking stock", Acorn User, September 1994), and since the London finance sector adopted products like the NS32016-based Whitechapel Computer Works MG-1 Unix workstation rather early on, with Acorn not delivering Unix on the 32016, I wondered if finally getting a Unix workstation out of the door brought about any opportunities or whether the company missed its chance.

[TimoHartong]

From the R260 I own I know it belonged to the NLR ( Nederlands Lucht en Ruimtevaart laboratorium ) , the Dutch institute for Air and space it was used for documentation and software development. I try to get more information about it. Based on the documentation with it I would say it was bought by the lab in 1991. From the R140 I don't have information available.

[paulb]

From the R260 I own I know it belonged to the NLR ( Nederlands Lucht en Ruimtevaart laboratorium ) , the Dutch institute for Air and space it was used for documentation and software development. I try to get more information about it. Based on the documentation with it I would say it was bought by the lab in 1991. From the R140 I don't have information available.

Very interesting. Although the reasons why something got bought can be fairly mundane, it can still give us an insight into the way the customers and users were thinking and whether the suppliers were reaching their target audience. The early 1990s were a time of considerable change for some organisations, with mainframes and minicomputers being replaced by workstations, leaving the door open to companies like Acorn.

Yesterday, I stumbled once again across some materials from IXI Limited, whose software was featured on the R-series machines, and one of the brochures describes the adoption of X. desktop at the Norwegian state telecoms company, Televerket, subsequently partly privatised and now known as Telenor:

"X. desktop™ User Stories"

In that case, it was Digital, Sun and IBM whose workstations were being used, though. Acorn didn't even seem to get the nod in the user story from the University of Cambridge.

[philpem]

Interactive Unix 386/iX was the SVR3-on-PC we were using.

Do you have any of the Interactive Unix stuff left?

I've been repairing a cable TV headend PC which runs on Interactive. Sadly I'm missing some of the install and patch disks.

As for the Sun/Interactive deal - from what I could gather, Sun bought Interactive (specifically their Lathman Associates division) to do the i386 port of Solaris. The last version of Interactive (v4.1.1) apparently included a coupon for a Solaris licence!

[arg]

Do you have any of the Interactive Unix stuff left?

None of what you want, I'm afraid. I was surprised to discover recently the source of the BBC end of the SJ Application server, and a floppy purporting to be the server end drivers, but it would need those missing Interactive install discs to bring it to life.

[sweh]

Do you have any of the Interactive Unix stuff left?

I've been repairing a cable TV headend PC which runs on Interactive. Sadly I'm missing some of the install and patch disks.

As for the Sun/Interactive deal - from what I could gather, Sun bought Interactive (specifically their Lathman Associates division) to do the i386 port of Solaris. The last version of Interactive (v4.1.1) apparently included a coupon for a Solaris licence!

My version is a wee bit older than that... version 1.0.6 !

I'm not sure I could even install this any more; the only 5.25" drives I've got are for the Beeb, and who knows if these disks have rotted...

There appears to be an archive of versions at http://tenox.pdp-11.ru/os/interactiveunix/ but I have no idea of the provenance of those images.

[paulb]

I have to say that considering the competitive situation at the time, choosing BSD as the basis of RISC iX was a shrewd move by Acorn, and there does seem to have been an opportunity at the bottom of the market. There were others also looking to get into that market, too, like Commodore, but they really didn't stand a chance with 680x0-based machines in the early 1990s.

I went down the rabbit hole on the whole Amiga Unix thing recently, but reviewing comments about BSD being far more desirable than SVR3 (or in arg's own words, "SVR3 was rubbish in comparison to BSD"), this might partly explain why both Atari and Commodore held off on releasing Unix workstations until SVR4 came out. But in doing so, they completely missed any kind of window for adoption in the marketplace.

Commodore had SVR3 running on the 68020-based Amiga 2000 version of their workstation effort, and Atari had UniSoft's port of SVR3 on the TT030. Both of them presumably thought that SVR4 was so desirable that bringing it to market would result in lots of sales interest. Commodore must also have realised that it needed a 68030-based machine, too. While both companies took their time, Sun, Apollo and HP all brought 68030-based machines to market during 1989, which may well have been Sun's final 680x0-based machines as well.

Here, it is worth comparing this with Acorn's strategy. The R140 was pretty underpowered - broadly similar in performance to the 68030-based machines - but Acorn was able to bring the R260 to market within a year or so, making it at least somewhat competitive. RISC iX was already deployed and being improved from 1989 onwards, and it seems that by choosing BSD, Acorn had selected a target market and knew what it wanted. In contrast, Commodore's strategists who were also targeting education, or who at least had one favoured institution who had adopted the Amiga 3000UX (having mandated A/UX for its students previously), seem to have been making guesses about what potential customers wanted.

By 1990, when Commodore's product came out, a 68030-based system would have been quite dated, even at the low end. There are a few stories about putting a 68040 in the Amiga 3000UX as soon as Motorola were ready with the chip, but Commodore never followed through. Given that HP were keeping their customers going (and Apollo's) with 68040-based machines during 1991, that might have kept Commodore somewhat in the running. Meanwhile, Atari seem to have been rather out of their depth with this whole Unix thing and it isn't even clear whether regular customers even had a chance of buying their product as it was belatedly finished off in 1992. Both companies withdrew from the Unix market shortly after.

Going beyond the basic specifications, I think that all of these new entrants failed to appreciate the depth of the existing offerings. The Amiga was supposedly this wonderful graphics machine, but Commodore had to bring in an independently developed graphics card to make X11 usable on the machine. The TT was enhanced over the ST variants, presumably to give it a shot at running X11, but it isn't clear how successful this was. Maybe Acorn's machines - at least the R260 - were more usable without the need for additional enhancements.

Well, I guess I went down the rabbit hole here, too!

[sweh]

Here, it is worth comparing this with Acorn's strategy. The R140 was pretty underpowered - broadly similar in performance to the 68030-based machines - but Acorn was able to bring the R260 to market within a year or so, making it at least somewhat competitive. RISC iX was already deployed and being improved from 1989 onwards, and it seems that by choosing BSD, Acorn had selected a target market and knew what it wanted.

I think it's also worth noting that Sun had sewn up a lot of the Unix workstation market with their BSD based systems (using 68020 in the Sun/3 series until 1989 when they released 68030 models). BSD was definitely well known and understood for workstations; SVR3 was seen more on servers (in my experience, anyway!) or as part of Microsoft Xenix.

[philpem]

Picking on the "SVR3 was rubbish in comparison to BSD" quote a little -- what did BSD of that era offer which was better than SVR3?

Asking as I've got an AT&T UNIX PC (which runs SysV, possibly R3 but I'm not sure), and a copy of ISC Interactive Unix (which is definitely SVR3) and I'm curious!

[paulb]

Picking on the "SVR3 was rubbish in comparison to BSD" quote a little -- what did BSD of that era offer which was better than SVR3?

Well, the Wikipedia summary gives some hints about what SVR3 was missing, particularly the BSD and SunOS notes.

[arg]

Picking on the "SVR3 was rubbish in comparison to BSD" quote a little -- what did BSD of that era offer which was better than SVR3?

A host of things, since this was the era when Berkeley innovation was at its height, while SVR3 was mostly AT&T trying to make money out of what they already had.

The first and most glaring one is the Berkeley Fast Filesystem. Interactive was straight SVR3, but they shipped Berkeley FFS as an add-on on an extra floppy that you were strongly encouraged to use as it was just so much better than the original filesystem that in SVR3 was much the same as in V7.

Networking and the socket interface originated in BSD; SVR3 had Streams which was an interesting piece of work but ultimately a solution to the wrong problem. IIRC the ethernet/TCP-IP in Interactive was Streams based but maybe had a library on top to give you sockets.

BSD had mmap() while SVR3 had SysV shared memory; maybe mmap() was immature at that point, but it was ultimately the approach that won out.

Sun NFS was obviously originated by Sun and not native to either SVR3 or BSD, but SunOS at that point was BSD4.3 with Sun add-ons, so it more immediately appeared in BSD.

All the tools were just better/more numerous in BSD (vi among many others) - BSD was originally just a distribution of such tools, with the kernel pieces and ultimately a complete OS coming later.

Maybe my memories are biassed/inaccurate, but certainly my perception at the time was that SVR3 was primitive and out-of-date.

SVR4 then appeared to be driven mainly by Sun who jumped ship from BSD to SVR but wanted to keep all the good stuff they were used to; my perception was that it was then no longer a technical laggard as SVR3 had been, but on the other hand it was now a 'kitchen sink' with loads of overlapping facilities while BSD remained cleaner.