How PS/2 and OS/2 handed the industry to Bill Gates
In April of ’87, IBM announced both a new PC architecture, PS/2, and a new operating system to run on the boxes, OS/2. “Real” business computers would also run IBM networking software and IBM SQL database software, all bundled into an “Extended Edition” of the operating system. There was only one real way of computing, IBM declared, and it had a /2 on the end. This was signalled by the job titles, the PC business was called “Entry Systems”: bikes with training wheels.
While IBM itself has subsequently survived and prospered, it’s a very different company today. It subsequently divested its PCs, printers and microprocessor divisions (along with much else) and one wonders how much different it would be today if it hadn’t devoted a decade and thousands of staff to trying to bring the PC industry back under its control. Ironically, Lenovo is the biggest PC company again – but by following the rules set by Microsoft and Intel.
OS/2 is an oft-told story, not least here at the Reg, where we gave an account of the OS wars on the 25th anniversary of OS/2. Dominic Connor also shared lots of evocative detail about the IBM culture at the time here (Part One) and here (Part Two). So there’s no need to do so again.
But every time history is written, it’s from a different vantage point – a different context. It no longer seems a joke to suggest, as IBM CEO Thomas J Watson probably never did, that the world needs only five computers. It’s a quote that adorned many
.sig files in the early days of email. Hah Hah. How stupid could a tech CEO be?
So if IBM had fought a different war, how might the world look?
If you’re not familiar with the significance of OS/2 and PS/2 and don’t want to read thousands of words, here’s a capsule summary. The setter of standards for the past three decades, IBM had responded to the microprocessor revolution by allowing its PC to be easily “cloneable” and run a toy third-party operating system. It didn’t matter too much to senior IBM management at first: PCs weren’t really computers, so few people would buy them. Business computing would be done on real multiuser systems. That was the norm at the time. But the runaway sales of the PC clones worried IBM and in 1984, just three years after the launch of the IBM PC, Big Blue plotted how to regain control. The result was a nostalgic backward-looking vision that under-estimated that computing standards would increasingly be set by the open market, not by IBM.
The PS/2 series of PCs had world-beating futuristic industrial design, based around clip-together plastic parts that made the beige tins of the time look instantly dated. And PS/2 would have been fine if it hadn’t been for the notorious proprietary bus: Micro Channel Architecture.
This plug and play bus was much better than the ISA standards of its day, and good enough for IBM to be using in workstations and even mainframes. However, it was not compatible with the (fairly crude) expansion boards of the day required to do networking or graphics; MCA cards were twice the price of comparable cards; and hybrid PCs were difficult to build. But the killer was that IBM demanded a licence fee from OEMs and sent its lawyers after MCA clones. The result was years of uncertainty. Seven years later, Apple was still making fun of how hard it was to get expansion cards to work in PCs. The real bedevilment of IBM’s PCs was the tech titan’s cost structure, which made it more expensive than the competition, at least without a heavy corporate discount.
But people don’t get emotional about PS/2s in the way they got emotional about OS/2, which, even as a former user, is pretty strange, given how much grief it gave you. The tl;dr of the OS/2 story is that IBM announced a highly advanced operating system for PCs, but it was five years (1992) before it shipped a version that was demonstrably better for the average user. (In fact, it was almost a year before anything shipped at all.)
Since operating systems aren’t an end in themselves, but merely a means to an end, a means of running something that alleviates your grunt work (like dBase or Lotus 1-2-3 at the time), the advantages of OS/2 were pretty elusive.
And, even in 1992, “better” meant managing old apps and files better – and the action in that “space” was taking place on Windows.
Because the industry was so unused to believing it could actually set standards, for a long time it didn’t. This nervousness in the wake of PS/2 and OS/2 had caused a kind of winter. People just didn’t bother upgrading – well, why would you? You had to wait and see what the standards would be.
So Microsoft put next to no effort into updating DOS (or even Windows) for the next two years. Application vendors continued to update their applications, but these remained character mode and there the lack of a standard for addressing extended memory also added to the inertia. Remember that OS/2 had been written for the 80286 chip introduced in 1984, and while the 80386 chip added new modes for protecting memory and virtualizing sessions, without the software to take advantage of it, weak demand ensured 386 PCs remained very expensive.
I described IBM’s vision of computing as nostalgic and backward-looking, with PCs limited to office paperwork while real computing took place on (IBM) servers. But what if IBM had dared skip a generation and tried something really daring?
At the time, there just that weren’t many options.
A week before IBM’s PS/2 and OS/2 roadmap was unveiled, Digital Research Inc shipped a multitasking DOS for 286 computers – Concurrent DOS 286 – to OEMs. This emulated the original IBM PC chip in software, and offered a huge leap in multitasking over Microsoft DOS. DRI also had a graphical shell, GEM. Why not dump Microsoft for DRI, which ran the industry standard OS when IBM set about making its first PC? It was really about control. IBM had excellent engineers and many PhDs; it could make it itself. IBM had little appetite to give Gary Kildall, who understood the microcomputer business much better than IBM, another sniff.
As it turned out, DRI struggled, along with everyone else, to make Concurrent DOS work reliably on the 80286 chip, particularly when it came to networking. The PC was a wild west, and reliable compatibility really needed the hardware virtualisation of the 80386 chip.
The obvious alternative was rapidly maturing: Unix. The trade press had busily hyping “open systems” for years. In fact, it’s little remembered now the PS/2 came with a version of IBM’s Unix: the Advanced Interactive Executive or AIX for the PS/2. “The multiuser, multitasking virtual memory operating system will be a subset of the Unix-like AIX operating system for the RT PC”, as InfoWorld reported.
(Note that a Unix would be “Unix-like” forever after.)
But IBM didn’t like Unix, and didn’t get serious about selling it until Sun, with the rest of the industry in hot pursuit, was eating into its business. IBM didn’t like distributed architectures that were too distributed, and for IBM, Sun’s emphasis on “the network is the computer” put the emphasis in completely the wrong place. The “Open Systems” hype was that the CIO user would mix and match their IT – and that was the last thing IBM wanted.
And there were more immediate, practical difficulties with leaping a generation of technology and yoking the IBM PC to Unix long term. It wasn’t ease of use. Later, usability was the stick used to beat Unix vendors with, but at the time DOS and Unix were equally hard to use. Backward-compatibility was the main issue: they wouldn’t run the PC applications of the day. It seemed far more plausible that IBM could persuade the big PC application vendors of the day – like Lotus and Ashton Tate – to migrate to OS/2 than bring their Unix ports to IBM’s version of Unix. Far less risky for IBM too.
With the benefit of hindsight, the third option would have been far more attractive: why didn’t IBM just buy Apple? It wasn’t IBM-compatible, but most of IBM’s kit wasn’t IBM-compatible, either. (Hence one of those grand IBM unifying strategies of the time: “System Application Architecture”.)
As it turned out, IBM and Apple would work closely together, but only out of desperation, once it was clear that Windows was a runaway success, and both had lost the developers to Microsoft. Much of the first half of the 1990s saw several thousand IBM and Apple developers working on ambitious joint projects that never bore fruit: Taligent, a WorkPlace OS, and much else.
Apple had a working Intel port of MacOS… in 1994. IBM and Apple had even agreed in principle to merge in 1995, only for Apple CEO Michael Spindler to get cold feet at the signing. He wanted more money from the buyer (the Apple Way).
Still, it’s fascinating to speculate what a popular, consumer-friendly OS might have done if IBM had been prepared to license Apple’s Macintosh OS aggressively, so it supplanted DOS as the industry standard. Many vendors had a foot in both camps at the time, so it really would have been a case of investing in existing initiatives, rather than starting from scratch. Without big IBM’s support, Microsoft would have been relegated to a tools company with a nifty spreadsheet, possibly eventually divesting both. I wonder who would have bought Excel?
Alas, nothing was further from the thoughts of IBM’s management at the time, obsessed with taming the industry and bringing it back under central control. What they’d always done had always worked. Why change?
Thomas J Watson may never have said the world will have five computers – it is almost certainly a myth. But it’s striking as the world moves to three (or four) computing platforms, IBM isn’t one of them. ®