back to article It's 30 years ago: IBM's final battle with reality

Thirty years ago this month, IBM declared war on reality – and lost. For the 30 years prior to that April day in 1987, nobody had dared to challenge IBM’s dominance of the computer industry. It would take almost a decade for the defeat to be complete, but when it came, it was emphatic. In April of '87, IBM announced both a new …

  1. Professor Clifton Shallot

    Interesting times

    And interesting speculation.

    I wonder if a smaller Microsoft making developer tools might even have been 'cool' for a while.

    I loved OS/2 but at that point I didn't have any work to do, I was free to just play so the fact that it could be spectacularly flaky just added to its charm.

    Doom in a window? Yes. Slowly. And briefly. But achieved. Well worth the entire UI hanging if you sneezed near it.

    1. Anonymous South African Coward

      Re: Interesting times

      On the other hand, my experience with OS/2 on a 33MHz 386SX with 4Mb RAM was excellent. First 2.1 then Warp, then Merlin... haven't had any crashes or funny things. Just wished the dang PC would go faster.

      DooM played better than under DOS and some games just loved the extra memory I was able to give to them with the DOS settings under OS/2...

      Good days, good memories.

      1. Anonymous Coward
        Anonymous Coward

        Re: Interesting times

        Agreed, I had it on a 386sx 25 (2mb ram) and then a 486dx2 66 (8mb ram), ran videos, games and windows apps faster than in windows and dos. Was a good OS, shame it didn't do better than it did. Have been using it up until a couple of years ago, some older machines where I work used it. But that wasn't a pleasant task, keeping it integrated with the current environment and making work around to keep it 'compliant'.

        1. Professor Clifton Shallot

          Re: Interesting times

          It was better at DOS and Windows apps than it was at native.

          One native GUI app could hang the whole interface making the system unusable.

          The multi-tasking was a bit iffy too - nothing stopped a high-priority process from hogging all resources forever.

          But it was still the best thing around at the time in many, many ways.

          1. Anonymous Coward
            Anonymous Coward

            Re: Interesting times

            OS/2 had a single applications message queue, or something alike, and a non responding application could block the whole queue. 3.0 run DOS applications very well, most Windows application worked well, but some had issues - i.e. Delphi 1.0. But despite IBM asserting it could also support Win32 applications, it never did, and from 1995 onwards the world was quickly migrating to 32 bit applications - and native OS/2 applications were too few and sparse.

            But OS/2 was just a part of the PS/2 equation - PS/2 and MCA boards were really too expensive compared to clones - and once Compaq and others started delivering good enough machines, it was too late to close the stable. Nor DRI/GEM nor Apple could have turned the tide - it was a matter of customers money.

            Back then PCs and their software were still expensive (much more expensive than today), and for many it was a quite demanding investment - asking even more like IBM did, was just cutting out many, many customers who could only afford clones - all of them running MS-DOS and then Windows.

            1. Richard Plinston

              Re: Interesting times

              > OS/2 had a single applications message queue,

              It was Windows 1 through 3.11 that "had a single applications message queue".

              """Preemptive multitasking has always been supported by Windows NT (all versions), OS/2 (native applications), Unix and Unix-like systems (such as Linux, BSD and macOS), VMS, OS/360, and many other operating systems designed for use in the academic and medium-to-large business markets."""

              > But despite IBM asserting it could also support Win32 applications, it never did,

              OS/2 Windows 3.x did support Win32s applications by loading in the win32s module, just like Windows 3.1 could do. However, Microsoft added a completely spurious access to virtual memory beyond the 2Gbyte limit of OS/2 (Windows supported 4Gbyte accesses) just to stop OS/2 using that beyond a particular version. Microsoft then required the new version in their software.

              Exactly what you would expect from Microsoft, and still should do.

              > But OS/2 was just a part of the PS/2 equation - PS/2 and MCA boards were really too expensive compared to clones

              OS/2 ran fine on PC clones and ISA boards. The link between PS/2 and OS/2 was entirely spurious.

              1. AndrueC Silver badge

                Re: Interesting times

                I think you might be confusing two things there. Having pre-emptive multi-tasking doesn't preclude having a single message queue. The two things are only tangentially related.

                Multi-tasking is the ability to run multiple processes (or multiple threads) by switching between them. Pre-emptive multitasking means that the OS can force a task switch, cooperative multitasking means that each process has to yield control back to the OS. OS/2 was an indeed one of the earliest (possibly the earliest) PC OS that was preemptive. Windows was only cooperative until 9x and NT.

                But nothing about being multi-tasking requires that the OS even support message queues. Early versions of Unix offered pre-emptive multitasking but in the absence of X-Windows probably didn't have any message queues. In fact arguably an OS probably never would. Message queues are usually a higher-level construct typically implemented in the GUI framework.

                And, sadly, it is indeed true that the early versions of Work Place Shell (the OS/2 default GUI) had a single message queue. IBM's recommendation was that developers implement their own queue sinks. The idea being that every application would have a dedicated thread that did nothing but accept messages and store them in a queue. The main application thread(s) would then empty this 'personal' queue at their leisure. I'm not sure why they wanted this design - maybe because then it was the application's responsibility to manage message storage? Sadly (and not surprisingly) most developers couldn't be arsed. As a result the WPS could often lock up. Now the OS itself wasn't locked - other processes would keep running just fine. If you were lucky enough to have a full screen VDM open you wouldn't even notice until you tried to go back to the WPS. When it happened to us my colleague and I used to ask the other one to Telnet in to our boxes and kill the main WPS thread to get things going again.

                One of the big features that OS/2 Warp finally brought was multiple message queues. Sadly by then it was too late. It's a shame because I did like the WPS. It's object oriented nature was great. Windows has never offered that. In the WPS an icon is an object that knows where it should appear. In Windows it's just an icon that Explorer choose to render in a particular location. Right click it and Explorer tries to work out what to put on the menu. Do the same in WPS and the icon will create its own menu.

                OOP GUIs are powerful things.

                1. Mine's a Large One
                  Thumb Up

                  Re: Interesting times

                  "Now the OS itself wasn't locked - other processes would keep running just fine. If you were lucky enough to have a full screen VDM open you wouldn't even notice until you tried to go back to the WPS. When it happened to us my colleague and I used to ask the other one to Telnet in to our boxes and kill the main WPS thread to get things going again."

                  We had a couple of apps which used to do this occasionally on our servers and the first we'd know about it would be when we tried to do anything at the server, something we didn't often do - everything had been merrily working away in the background but with a single app having frozen the WPS.

                2. Steve the Cynic

                  Re: Interesting times

                  "Windows was only cooperative until 9x and NT."

                  Pedantry alert!

                  Win3.X in *enhanced* mode would coop among the running Windows apps, and preempt among DOS VMs and between the VMs and the collection of Windows apps.

                  Well, unless multiple preempted things tried to do anything that called down into DOS, which was mutexed.

                  1. Richard Plinston

                    Re: Interesting times

                    > Well, unless multiple preempted things tried to do anything that called down into DOS, which was mutexed.

                    The major limitation to all MS-DOS based system, including Win95/98, was that they did not recover disk access time. When a disk access was actioned by a process it waited for it to complete: the seek time and the latency left the CPU idle, as you say by being mutexed.

                    Real multi-tasking and multi-user systems, such as Unix and DRI's MP/M/Concrrent/etc, started the disk request, put that process on the waiting queue, and got on with the next process timeslice. When the disk transfer completed it caused an interrupt and a reschedule.

              2. Fred Down

                Re: Interesting times

                I worked on OS/2 at IBM Hursley, and at least in the versions I worked on OS/2 hand a single message queue to Presentation Manager. I did a talk on comparisons between Presentation Manager and X Windows and this was discussed,

                OS/2 had for the time excellent multi-threading, the theory was that any program would grab a message from PM's queue, spin off a thread to handle it and then release the message queue, easier said than done.

                My views on why OS/2 died :-

                1 very poor developer support the development kit was expensive whist Microsoft was practically giving theirs away.

                2 Poor peripheral support, for the longest time it did not support Sound Blaster the de-facto sound device at the time.

                3 Required much greater hardware than was common at the time. The minimum stated hardware requirement was way to small to allow it to run properly.

                I came from running Xenix on 286 hardware which managed pre-emptive multi tasking properly and had for years before OS/2.

            2. cmaurand

              Re: Interesting times

              Actually, it did support 32 bit applications. Microsoft kept changed something in windows in the way it handled 32 bit applications, IBM adjusted, then Microsoft came out with Win32s, IBM adjusted, Microsoft changed it again (something about the way windows does things in memory) and IBM gave up on trying to keep up with Microsoft's changes.

              1. AndrueC Silver badge
                Thumb Up

                Re: Interesting times

                Actually, it did support 32 bit applications.

                That Windows integration piece in OS/2 was cool, I thought. Pretty much seamless and you could choose how seamless it was - full screen or pretend to be an application on the desktop. The bit where it could link into an existing installation was just brilliant. Licensing payments to Microsoft for Windows? Not today, thanks :D

                But then the entire VDM subsystem was cool. A true 'Virtual DOS machine' rather than the Windows poor cousin. You could even boot up different versions of DOS. I seem to recall one of their bug fixes was to the audio subsystem to support a Golf simulator that triggered tens of thousands of interrupts per second. From what I vaguely recall they said they removed the need on the card side and faked the results inside the VDM. The VDM team must've gone to extraordinary lengths in order create what we'd probably now call a Virtual Machine.

              2. Anonymous Coward
                Anonymous Coward

                "Actually, it did support 32 bit applications. "

                Yes, OS/2 supported OS/2 native 32 bit applications, and Win16/32s applications. It never supported Windows *full* Win32 applications - those written for Windows 95/NT.

                I run OS/2 on a Pentium starting in 1994 until it was clear the PC world was moving towards Win32 applications, and I couldn't develop them on OS/2. So I had to move to Win95 (which, under many aspects, looked inferior, in my experience, especially FAT vs HPFS), thus I quickly moved to NT4, as soon I got a Pentium II and enough RAM to run it.

      2. AndrueC Silver badge
        Thumb Up

        Re: Interesting times

        I remember Sunday afternoons playing Geoff Crammond's Formula One Grandprix in a VDM while downloading messages from CompuServe using the multi-threaded Golden Compass. My first experience of real multi-tasking on a PC.

        I also used it to develop DOS applications as the crash protection meant that I only had to reopen the VDM, not reboot the machine.

      3. ChrisC Silver badge

        Re: Interesting times

        I have fond memories of Warp too - back then I was doing some research work on robotic equations of motion, which had eventually evolved into a hideously complex Matlab script to do all the hard work for me. I'd just define the system geometry parameters at the start, click Go, twiddle my thumbs for an hour or so, and then get a complete set of optimised motion equations out the other end.

        Unfortunately this was all being done in the Win3.1 version of Matlab, and as bad as the co-operative multitasking was in 3.1 generally, it was a shining beacon of excellence compared to how it behaved once Matlab started up - I'm pretty sure the Matlab devteam must have misread the Windows documentation and thought it featured "un-cooperative multitasking", because once you let Matlab loose on a script it was game over as far as being able to do anything else on that PC was concerned.

        As a hardcore Amiga user at the time, I knew that multitasking didn't have to be this godawful, and I was convinced that the PC I had in front of me, which at the time had roughly twice the raw processing power of the fastest Amiga in my collection, really ought to be able to multitask at least as well as the slowest Amiga in my collection...

        I can't recall how I stumbled upon OS/2 as the solution, all I do remember is that having learned of its existence and its claimed abilities to do stuff that Windows could only dream of doing, I dashed into town and bought my own copy of Warp, and once I got over the hurdle of getting it installed as a multi-boot setup with my existing fine-tuned DOS/Win3.1 setup (having expended god knows how many hours tweaking it to run all my games nicely - yes, even those that expected to have almost all of the base memory available, but still also needed to have CDROM *and* mouse drivers shoe-horned in there somewhere too - I didn't want to mess that up) I fired it up, installed Matlab, and tentatively clicked Go... Umm, is it running? This can't be right, the OS is still perfectly responsive, I can launch other Win3.1 applications without any signs of hesitation, and yet my Matlab script really does claim to be churning its way through its calculations about as quickly as it did hogging Win3.1 all to itself.

        From that day on, Warp became my go-to OS for anything work-related until the day I finally ditched Win3.1 and made the switch to 95.

        So yes, count me in as another one of those people who, despite the problems OS/2 had (I'll readily admit that it could be a bit flakey or just a bit obtuse when trying to get it to do what you wanted it to do) will still quite happily wax lyrical about just how bloody amazing it was in comparison to a DOS/Win16 based setup for anyone wanting to unlock the true potential of the hardware in front of them. Even today I still don't think the Windows dev team *really* understand how multitasking ought to behave, and I do wonder just how much productivity is lost globally due to those annoying random slowdowns and temporary hangs which remain part and parcel of everyday life as a Windows user, despite the underlying hardware being orders of magnitude more powerful than anything we could dream of having sat on our desks back in the 90's.

      4. CrazyOldCatMan Silver badge

        Re: Interesting times

        my experience with OS/2 on a 33MHz 386SX with 4Mb RAM was excellent

        It was (even more so than DOS/Windows) sensitive to the hardware - if you had anything slightly out then OS/2 would do lots of mysterious things..

        I used it, right up to the end when IBM dropped it. Along with linux (my first linux machine was a 386sx25 with a 330MB ESDI drive - was originally OS/2 with an IDE 80MB drive but OS/2 simply wouldn't load when the ESDI drive interface card was inserted. So I used it for linux instead).

        After all, OS/2 did TCP/IP long before Windows had a stable IP stack supported by the manufacturer (yes, yes, there was Trumpet WinSock and Chameleon TCP/IP but they were not from Microsoft and MS refused to support them).

        1. martinot

          Re: Interesting times

          "After all, OS/2 did TCP/IP long before Windows had a stable IP stack supported by the manufacturer (yes, yes, there was Trumpet WinSock and Chameleon TCP/IP but they were not from Microsoft and MS refused to support them)."

          Not true.

          For OS/2 2.11, and also later on with Warp 3, I had to get a separate TCP/IP product, but right from the start I had it in build in with first version of Windows NT (3.1 and forward).

          It was not until I got OS/2 Warp Connect that was the first OS/2 product to ship with TCP/IP it build in from start in the OS (even that a bit more separate install and not as fully integrated as on Windows NT).

          Windows NT 3.1 was released in July, 1993. OS/2 Warp Connect was released in May, 1995.

          1. Richard Plinston

            Re: Interesting times

            >> "After all, OS/2 did TCP/IP long before Windows had a stable IP stack supported by the manufacturer ..."

            > For OS/2 2.11, and also later on with Warp 3, I had to get a separate TCP/IP product,

            Yes, it was a separate product, but that does not contradict the person you are replying to. It was available and supported by the manufacturer of the OS.

            """The Extended Edition of 1.2 introduced TCP/IP and Ethernet support."""


            OS/2 1.2 was released in 1989 - 4 years before NT.

    2. Planty Bronze badge

      Re: Interesting times

      Interesting times again. Microsoft are now in IBM shoes, facing irevellance as nobody really cares about their products anymore. Microsoft have been running around throwing out all sorts of random things, hoping something will stick. Nothing stuck, everything sucked.

    3. Steve Channell

      Re: Interesting times

      Lets try not to forget that Microsoft only got the PC-DOS gig after pitching Microsoft port of AT&T Unix (Xenix) for IBM's Personal Computer. It is also worth mentioning that Windows was initially developed on Xenix and ported to DOS after initial testing.

      Had it not been for IBM, the world would be.... pretty much as it is now

      1. Richard Plinston

        Re: Interesting times

        > pitching Microsoft port of AT&T Unix (Xenix) for IBM's Personal Computer.

        That is extremely unlikely, and most likely false. IBM wanted their PC to compete with Apple II plus Z80 softcard running BASIC and CP/M, not with mini-computers*. Xenix required at least 512Kb, a hard disk, and was multiuser, the PC was to be sold with as little as 16Kb running ROM BASIC and using the cassette port - just like the minimum Apple II. In fact the first 5150 model A could not be fitted with more than 256Kb (if anyone could afford that). The model B (which I have here) as indicated by a blue letter B in a circle on the back panel was revised to cater for the full 640Kb. There was no hard disk interface until the IBM PC XT (though Xibec did develop one later).

        Also IBM went to Microsoft for the BASIC ROM (and other language compilers). The knew that they wanted CP/M which Microsoft also sold with their Z80 softcard for the Apple II.

        * completely the wrong division within IBM.

        > It is also worth mentioning that Windows was initially developed on Xenix and ported to DOS after initial testing.

        In the 1980s almost all software for CP/M and MS-DOS (including CP/M and MS-DOS) was developed on mini-computers running Unix/Xenix or DEC systems. Micro-computers (at 4.77KHz) were completely inadequate to be used by a development team (small disks, no network, slow compiling, edlin?). Yes, Windows would have been written and compiled on a Xenix multi-user system with green-screen terminals, probably using vi. This does not mean that it would have been run on Xenix.

        Even as late as when the started work on Windows NT the Intel PC was not suitable for developing it. It was done on MIPS workstations and later ported to x86.

        1. Anonymous Coward
          Anonymous Coward

          "In the 1980s almost all software for CP/M and MS-DOS was developed on mini-computers"

          Are you sure about that? A lot of initial DOS software was often written directly in assembly (although C compiler became available quickly). Edlin was the DOS "default" editor, but not the only editor available for DOS (WordStar itself could be used as a text editor). Network was available, IBM introduced its PC LAN quite early - and Netware servers as well. But i don't know how many really needed networking, since often in the early days software was written by a single developer. Disks weren't large, but applications were small as well.

          1. Richard Plinston

            Re: "In the 1980s almost all software for CP/M and MS-DOS was developed on mini-computers"

            > A lot of initial DOS software was often written directly in assembly (although C compiler became available quickly).

            Perhaps I should have specified 'non-trivial software'.

            Interesting that you start with: "A lot of initial DOS software", including MS-DOS itself, which obviously wouldn't have been written using DOS - because it wasn't available yet.

            > IBM introduced its PC LAN quite early

            It was initially released in 1984, 3 years after the IBM PC.

            Certainly a team could use a network server, but then the PC is acting as a terminal.

            > since often in the early days software was written by a single developer.

            All the software you mentioned (except edlin) took more than a single developer.

            > Disks weren't large, but applications were small as well.

            The IBM PC initially catered for two 160Kb SSSD disks and only later went to DS and 10 sector to give 360Kb. It wasn't until MS-DOS 2.x and IBM PC-XT that hard drive were available.

            On release the IBM PC catalogue listed PC-DOS, CP/M-86, UCSD, Wordstar, Visicalc, Peachtree, compilers for: BASIC, COBOL, Pascal, FORTRAN; Forth, a 'program generator', none of which were capable of being compiled using 160Kb disks, and the source code of each probably was too large for the 10Mbyte disk of the XT.

            I was developing software on micros since the very late 70s, with way more resources than an IBM PC later had, using Wordstar and other, but it was only relatively small stuff, mainly in COBOL. The compiles would run all night

        2. John Styles

          Re: Interesting times

          Nah. I started work in 1985 and only really ever worked on PCs (apart from some weird odds and sods), all the development was done on PCs - these were all expensive vertical market applications going for 20K or thereabouts a copy, with teams of 10 to 20 people in a couple of cases. Mostly FORTRAN 77 and C. (Lahey Fortran, Lattice C then the Microsoft ones).

        3. martinot

          Re: Interesting times

          "Even as late as when the started work on Windows NT the Intel PC was not suitable for developing it. It was done on MIPS workstations and later ported to x86."

          According to Cutler and other NT managers this (development on machines running Windows NT on MIPS) was a conscious decision by them to ensure that NT was portable and platform neutral, and not locked into x86 code.

          1. Richard Plinston

            Re: Interesting times

            > this (development on machines running Windows NT on MIPS) was a conscious decision by them to ensure that NT was portable and platform neutral, and not locked into x86 code.

            Originally Cutler started to use the Intel i860XR workstations but this failed. In 1988, when development started, the best x86 was the 80386 at 25MHz*. This was completely inadequate so they moved to MIPS workstations.

            * 33MHz in 1989.

    4. Faux Science Slayer

      Rockefeller front IBM replaced by Rockefeller front MicroSoft

      Edison did not invent electric light, Alex Bell stole his telephone patent, Marconi STOLE Tesla radio patents and Glenn Curtiss stole a dozen Wright Brothers patents. All were set up fronts for Rockefeller monopolies.

      "JFK to 9/11, Everything is a Rich Man's Trick" on YouTube....end feudalism in 2017

  2. schifreen

    Too much credit

    You seem to conclude that the problem lay with IBM's poor strategy. You give the company way too much credit. They didn't really have a strategy at all and, if they did, no one knew what it was. Or where they would be working once the latest departmental reorg was complete.

    The problem with OS/2 is that it was a solution to a problem that hardly anyone actually had. IBM knew this. Ask senior IBM people at the time (as I did) exactly what problem OS/2 was designed to solve, and why the world needed to ditch the free OS that came with their PC in order to install something that, frankly, never installed properly anyway, and they singularly failed to come up with a sensible answer. Or any answer at all.

    1. Little Mouse Silver badge

      Re: Too much credit

      I'd almost forgotten how pervasive the image of IBM was back in those days. I joined the PC party in the days of 286, and IIRC, PC's only really fell into two camps - either IBM or IBM clones.

      Amazing that they could get from that position of dominance to this.

    2. I am the liquor

      Re: Too much credit

      I don't think it's true that OS/2 was a solution to a problem that never existed. Perhaps it was too early. But by the mid 90s when customers did understand the problem, and Microsoft was answering it with Windows NT, surely that should have meant that OS/2 was mature and ready to take advantage.

      OS/2 Warp was a better design in many ways than NT 3.51, but IBM's head start didn't help them win the race, and now we all run Windows NT on our PCs.

      1. Anonymous Coward
        Anonymous Coward

        Re: Too much credit

        Many people started to have problems with the single applications model DOS had - it's no surprise one of the first big hits of Borland was Sidekick - which made TSRs (Terminate and Stay Resident) applications common.

        Still, everything had to work in the 1MB of memory available in real mode. Extended/Expanded memory couldn't be used to run code from. DOS Extenders later could run bigger applications, but still a single one.

        Windows was well accepted not only because it was a GUI, but because it allowed to run more than one application, switch among them easily, and move data across applications. The limited multithreading was not an issue on desktops with only a single core CPU.

        If you were using the PC just to play games, it really didn't matter, but for business users it was a great productivity boost.

      2. CrazyOldCatMan Silver badge

        Re: Too much credit

        OS/2 Warp was a better design in many ways than NT 3.51, but IBM's head start didn't help them win the race, and now we all run Windows NT on our PCs.

        Windows winning has often been called "the triumph of marketing over excellence".

      3. martinot

        Re: Too much credit

        "OS/2 Warp was a better design in many ways than NT 3.51, but IBM's head start didn't help them win the race, and now we all run Windows NT on our PCs."

        I used both OS/2 2.11/Warp3 and NT 3.5/3.51/4 at the time (and FreeBSD). From a technical point I thin NT was way better designed (more stable, more future proof, scalable, etc.) than OS/2, but what I think OS/2 shined in was running DOS (NT was designed more for future tech than past tech) and WPS really much better than NT (at least before the Win95 GUI come to NT with NT4).

    3. Warm Braw

      Re: A solution to a problem that hardly anyone actually had

      It could have been a solution to a problem that a lot of people had. If you've ever had the misfortune to write software for IBM mainframes and been stuck with TSO or SPF (or even VM/CMS which is only better my comparison with the alternatives) you'd have given your eye teeth for a less developer-hostile environment. If they'd put decent mainframe developer tools onto a personal workstation they'd have had a core market to build on.

      But this wasn't the IBM way: accounts were structured along product lines and no mainframe systems salesman was going to let a PC salesman or a System/34 salesman onto his territory if the thought it might cannablise his commission. No product manager was going to risk the ongoing investment in mainframe software by allowing personal computers to connect other than as the dumbest of terminals. It's ironic that IBM used the word "System" so much: it never really understood systems, just product families..

    4. Vic

      Re: Too much credit

      You give the company way too much credit. They didn't really have a strategy at all


      I was working at an IBM Systems Centre when the PS/2 launched. We had to go to customers to tell them how wonderful this MCA thing was - without having any knowledge of whether or not it was any better than ISA.

      And that was a shame, really - MCA *was* better. But no-one found out until it was way too late. And the Model/30 didn't have MCA anyway...


    5. a_yank_lurker

      Re: Too much credit

      @schifreen - Itsy Bitsy Morons always had a schizophrenic attitude towards PCs and to a lesser extent minis at the time. They worshipped big iron and could not understand why people would want a "toy" or "crippled iron". What they failed to grasp is many computing activities are not very resource intensive on any computer (I wrote a thesis on an Apple IIe) even early PCs. These are activities that could be easily automated and put on a smaller computer. Others grasped the vacuum left and moved in with both feet.

      The other issue for them was selling PCs is very different than selling a mainframe. One buys a PC much like one buys any other appliance from a retailer. There is no formal bid process with tenders to opened and reviewed. At retail, the sales staff is less interested in the brand you bought but is very interested in selling you something.

      1. Richard Plinston

        Re: Too much credit

        > Itsy Bitsy Morons always had a schizophrenic attitude towards PCs and to a lesser extent minis at the time. They worshipped big iron and could not understand why people would want a "toy" or "crippled iron".

        You write as if IBM were one thing. It was divided up into several divisions, each with their own sales and marketing and each competing against the others. The mainframe division (360/370) wanted a small computer to counter the Apple IIs with Visicalc, Z80 softcard and CP/M software invading their sites. The IBM-PC was designed to be 20% better than the Apple II (160Kb floppies instead of 120Kb etc) and also act as a terminal (which is why the IBM PC has DTE serial ports while other micros had DCE) while running the same software. There were also 3740 (terminal) PCs and 360 emulating PCs (with Motorola 68x00 co-processor boards) for developers to use to write mainframe software.

        The mainframe division did look down on the Series One, System 3, System 36 and System 38 (AS400) and other divisions, but did not see the IBM PC as any threat at all. They did want to exclude other brands though.

    6. Richard Plinston

      Re: Too much credit

      > and why the world needed to ditch the free OS that came with their PC

      MS-DOS and Windows were never free*. When you bought a computer with Windows installed, and sometimes when Windows was _not_ installed, money went from the OEM to Microsoft. That cost was part of the price.

      * actually there was a 'free' version: 'Windows with Bing' that no one wanted.

    7. addinall

      Re: Too much credit

      MS-DOS was never free. Unless you stole it.

      1. Richard Plinston

        Re: Too much credit

        > MS-DOS was never free. Unless you stole it.

        When SCP sold 86-DOS to Microsoft for development into PC-DOS and MS-DOS (MS had previously licenced it from SCP) the agreement was that SCP would have as many copies of MS-DOS as they wanted for free as long as they were shipped with a computer (SCP built the Zebra range of S-100 based computers).

        After the fire in the SCP factory, which stopped them building computers, they started selling V20 chips (faster clone of the 8088* with 8085 emulation built in) and V30 chips (ditto 8086) with a free copy of MS-DOS. MS bought out the agreement for a reputed $1million.

        * swap this for the 8088 to get a 20% faster machine that could also run CP/M software (with suitable loader).

        1. Pompous Git Silver badge

          Re: Too much credit

          "swap this for the 8088 to get a 20% faster machine that could also run CP/M software (with suitable loader)."
          Even betterer, it could run Z80 code, though I never got a round tuit. The extra speed was welcome though :-)

        2. Tcat

          Re: Too much credit

          Give the man a cigar!

          Exactly, I was a next door neighbor (business wise) to SCP in Tukwilla WA on Industrial Blvd.

          I may add someday to the L/PM discussion of Mr. Gary not giving a crap about Intel CPUs

          and the transfer to SCP of L/PM and move to SCP which SCP used for an early version of AutoCAD

          @ 25K USD for a station,

      2. Richard Plinston

        Re: Too much credit

        > MS-DOS was never free. Unless you stole it.

        After per-box pricing* was declared illegal, MS came up with another scheme where MS-DOS and Windows were bundled together at the price of Windows alone. Effectively this was MS-DOS for free to stop DR-DOS being installed. At the time it was MS-DOS 4.01 versus DR-DOS 5 which was infinitely superior and it took MS 20 moths to nearly catch up with MS-DOS 5, at which point DR released DR-DOS 6 with task switching. MS took another year to almost catch up with MS-DOS 6.

        * OEMs were contracted to pay Microsoft for MS-DOS on every box sold regardless of whether it had MS or DR-DOS (or other) installed. This was to strangle DR-DOS sales. The alternative to accepting this contract was to never sell any MS products ever again.

        1. Anonymous Coward
          Anonymous Coward

          Re: Too much credit

          "At the time it was MS-DOS 4.01 versus DR-DOS 5 which was infinitely superior and it took MS 20 moths to nearly catch up with MS-DOS 5, at which point DR released DR-DOS 6 with task switching. MS took another year to almost catch up with MS-DOS 6"

          But MS had a trick up its sleeve to sideline DR-DOS ... with Win 3.0 they had a long (several months) "beta" period where peopel cold download the "beta" for free to try out so there was a lot of prelaunch publicity. Part of that publicity was that Win 3.0 didn't work on DR-DOS as there was an error durign start-up. In legal terms DR weren't allowed to access the beta so they couldn't counter all the "if you want win 3.0 you'll need MS-DOS and not DR-DOS" stories in the press. In reality the error was, I believe, due to MS deliberately using a result from a DOS call which wasn't precisely specified where Win3 worked with the value MSDOS returned but not with the one DRDOS returned ... trivial for DR to fix and seem to recall they did this as soon as Win3 launched but the damage was done.

          At the time I was using DR-DOS so I assumed that to get Win3 I'd need to factor in the cost of MSDOS as well and at which point the price of OS/2 (with a special launch offer) was similar so I went for OS/2

          1. Richard Plinston

            Re: Too much credit

            > In reality the error was, I believe, due to MS deliberately using a result from a DOS call ...

            It was called the AARD code if you want to search for accurate details.

          2. Anonymous Coward
            Anonymous Coward

            "beta" period where peopel cold download the "beta" for free

            In 1990 not many could download software, most PCs were without any kind of connectivity, and services like CompuServe and BBS not really for everyone...

            1. Pompous Git Silver badge

              Re: "beta" period where peopel cold download the "beta" for free

              "In 1990 not many could download software"
              Back then I had a 300 baud modem to connect to Pegasus Networks. A rough back of the envelope calculation tells me that the approximate cost of downloading Win3.0 would have been $400. I cannot recall it occurring to me in those far off days to DL software. It was log on, read incoming emails as they downloaded, and immediately log off.

      3. Pompous Git Silver badge

        Re: Too much credit

        MS-DOS was never free. Unless you stole it.
        Only two of the many computers I've owned came with MS-DOS, most were without an OS. Since MS refused to sell DOS at retail, most people did just that; they stole a copy of DOS. A much smaller number of us purchased DR-DOS and reaped the benefits of an arguably better DOS than DOS. Especially if you also ran the 4DOS command processor.

        1. Richard Plinston

          Re: Too much credit

          > Since MS refused to sell DOS at retail

          MS had a contractual moratorium on selling MS-DOS at retail for 10 years with IBM. This expired and MS released MS-DOS 5 for retail sales.

          > A much smaller number of us purchased DR-DOS and reaped the benefits of an arguably better DOS than DOS.

          It was significantly better that the contemporary MS-DOS 4.01 and had a 20 month lead on MS-DOS 5.

          Allegedly it reached a 20% market share until MS brought in illegal per-box pricing and bundled MS-DOS+Windows at Windows price.

          1. Pompous Git Silver badge

            Re: Too much credit

            "MS had a contractual moratorium on selling MS-DOS at retail for 10 years with IBM. This expired and MS released MS-DOS 5 for retail sales."
            B-b-b-b-ut that can't be true. Shirley MS are responsible for everything bad in computing... ;-)

        2. FuzzyWuzzys

          Re: Too much credit

          DR-DOS was superb, it had so much stuff bundled in and it worked like a dream and as proven MS took months to catch up to what DR-DOS did. DR also has one of the more tragic stories in PC history, well worth checking out.

  3. Martin 47

    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.

    Think someone needs to mention that to Microsoft

    1. stephanh

      The whole situation seems very similar to Microsoft today desperately getting a bolthole in the mobile market, against Android. Microsoft's "Windows Experience" Android strategy me a lot of OS/2 (replace a working, familiar OS by something nobody needs).

      "Hegel remarks somewhere that all great world-historic facts and personages appear, so to speak, twice. He forgot to add: the first time as tragedy, the second time as farce."

      -- Karl Marx, "The Eighteenth Brumaire of Louis Napoleon"

    2. Pompous Git Silver badge

      "Since operating systems aren’t an end in themselves, but merely a means to an end... the advantages of OS/2 were pretty elusive."
      Not really. It was far from unknown in publishing to have to leave the machine rendering a print job because it couldn't do anything else at the same time. Being able to continue working while printing would have been a blessing, but had to await WinNT.

  4. Admiral Grace Hopper

    The Man In The High Data Centre

    I enjoy a good counterfactual. I still wonder occasionally how the world would have looked if IBM and Apple had got Pink to the point where it was marketable.

    1. Anonymous Coward
      Anonymous Coward

      Re: The Man In The High Data Centre

      Bloke at our small village local being cagey about what he did at IBM. Eventually I twigged and said "Oh, you're working on Pink?" He seemed amazed that anyone would know about it, and eventually chilled, but it struck me that little episode was a good analogy for the problems with the ill fated IBM/Apple dalliance.

    2. jake Silver badge

      Re: The Man In The High Data Centre

      Somewhere, I have a T-shirt with the IBM logo of the time superimposed over the "full color" Apple logo of the time on the front. On the back, it reads "Your brain, on drugs.". The first time we wore them at work, we were told we'd be fired if we wore them again ...

  5. Colin Bull 1
    Thumb Up

    succesful standard

    The PS2 brought with it one of the longest used PC standards - VGA. Only in the last year or two has the VGA standard connector been superceded by HDMI.

    1. MrT

      Re: succesful standard

      And the PS/2 port for mice/keyboards.

      The Model M keyboard set a high standard that many modern items still struggle to beat...

      1. Peter Gathercole Silver badge

        Re: succesful standard

        Though to be truthful, the key action in the Model M keyboards had appeared in the earlier Model F keyboards, and the very first Model M Enhance PC keyboard appeared with an IBM PC 5 pin DIN connector on the 5170 PC/AT.

        We had some 6MHz original model PC/ATs where I worked in 1984, and even then I liked the feel of the keyboard. Unfortunately, the Computer Unit decided to let the departmental secretaries compare keyboards before the volume orders went in, and they said they liked the short-travel 'soft-touch' Cherry keyboards over all the others (including Model Ms).

        As this was an educational establishment, the keyboards got absolutely hammered, and these soft-touch keyboards ended up with a lifetime measured in months, whereas the small number of Model Ms never went wrong unless someone spilled something sticky into them.

        I wish I had known at the time that they were robust enough to be able to withstand total immersion in clean water, as long as they were dried properly.

      2. I am the liquor

        Re: succesful standard

        Yes you still get a lot of motherboards with both PS/2 and VGA ports. Astonishing when you think about how many other different kinds of ports and expansion interfaces have come and gone over that period.

        The Model M slightly predated the PS/2 I think, but for sure they're inextricably linked in the memories of those of us who used the PS/2. People may not get misty-eyed over the PS/2 itself, but the Model M, that's something else. I found my old buckling-spring keyboard the other day... I'd forgotten how nice it was to type on, and that was just a Gateway 2000 one, not an IBM. I think I might treat myself to one of the new Unicomp ones when they get the beige ones back in stock.

      3. Anonymous Coward
        Anonymous Coward

        Re: succesful standard

        Yes, and in typical IBM fashion it's no longer available. IBM's problem was (and still is) that if they can't control something, they're not interested in it, regardless the profit margin.

    2. Wade Burchette

      Re: succesful standard

      The PS/2 standard is for keyboards/mice and it is still an important connector. It just works.

      And VGA has not been supplanted by HDMI, but by DVI which was supplanted by DisplayPort. HDMI is designed for TV's and has a licensing cost.

      1. Dave 126 Silver badge

        Re: succesful standard

        > The PS/2 standard is for keyboards/mice and it is still an important connector. It just works

        I remember PCs that wouldn't recognise a PS/2 keyboard if it was only plugged in after boot up.

      2. /dev/null

        Re: succesful standard

        Aaaand... 1.44MB 3.5in floppy disks, which were around for a while too.

        1. Danny 14

          Re: succesful standard

          I bought a second hand denford milling machine. The license comes on a 3.5in floppy disk - I needed to buy a USB drive as I didn't have any working ones left (not even sure I had a motherboard with a connection any more either).

        2. PickledAardvark

          Re: succesful standard

          Ah, those proprietary floppy disk drives which couldn't tell the difference between a 720Kb and 1.4Mb disk. I hate to think about the hours spent recovering data from misformatted floppies.

          And whilst the industrial design of the PS/2 was attractive and internals were easily accessible, quality of manufacture (the ones made in Scotland, at least) was lousy. I still carry the scars from fitting lids back on Model 55s.

          1. Dave 32

            Re: succesful standard

            Ah, you forgot about the 2.88MB floppy disks.

            I happen to have an IBM PS/2-model 9595 sitting here under the desk with one of those drives in it. ;-)

            Ah, yes, the model 9595; those had that little 8 character LED "Information Panel" display on the front of the case. It only took a tiny bit of programming to write an OS/2 device driver to turn that into a time-of-day clock. For years, that was the only thing that I ran on that 9595 (I called it my "600 Watt clock".).

            Hmm, why was there suddenly a price spike for old IBM PS/2-model 9595 machines? ;-)


            P.S. I'll get my coat; it's the one with the copy of Warp in it.

      3. Richard Plinston

        Re: succesful standard

        > HDMI is designed for TV's and has a licensing cost.

        But it is trivial, as evidenced by the Raspberry Pi Zero having HDMI and they sell retail for $5.00 (if you can find them).

  6. GlenP Silver badge

    OS/2 and PS/2 Memories

    I found one, and only one, use for OS/2 "back in the day". There was an electronics design package that needed it, but in the end for the hassle I persuaded the engineer he didn't really need it.

    We had one of the first PS/2 networks at the college where I was working, Model 30 workstations and a Model 60 server. Lord knows why they went that way instead of the RM networks that were standard (I suspect to try and keep it out of the control of the IT department, they failed as they still had to get us to get it working). The server and MCA boards were so rare when the network card failed IBM Greenock had to build a new one for us!

    I later did remote support for insurance agents with PS/2 Model 50s, the ones that if you turned them off then back on again within a minute would kill the PSU. The usual IT Fix, "Turn it off and back on" became, "Turn it off, DON'T turn it back on yet!"

    1. P. Lee

      Re: OS/2 and PS/2 Memories

      >the usual IT Fix, "Turn it off and back on" became, "Turn it off, DON'T turn it back on yet!"

      Weirdly I have this problem with an x79 ASRock board. I have to cut the power to it for 5-10 minutes before it will power on again. I've resorted to turning it off at the wall when I'm done.

      1. Anonymous Coward
        Anonymous Coward

        Re: OS/2 and PS/2 Memories

        "Weirdly I have this problem with an x79 ASRock board. I have to cut the power to it for 5-10 minutes before it will power on again. I've resorted to turning it off at the wall when I'm done."

        Sounds like your PSU is crap - not losing charge in the capacitors and leaving residual power on the output lines, holding the board in standby mode.

        My suggestion - check the output voltage on the PSU cables (12V, 5V, + & -) after the PSU is switched off and watch it over time.

        If it's not zero when the power goes off, you may have a duff/cheap PSU.

        1. Danny 14

          Re: OS/2 and PS/2 Memories

          or the board regulator caps are crap. They are simulating a power spike so the board is in failsafe "off".

      2. Robert Carnegie Silver badge

        Re: OS/2 and PS/2 Memories

        Re off means off. Someone the other day mentioned a problem of using an externally powered USB hub peripheral that sends power back to the PC motherboard. If the motherboard also doesn't deal well with that, that could account for it not "resetting". Educated guess.

      3. Updraft102

        Re: OS/2 and PS/2 Memories

        Check your PSU for bulged caps.

    2. bombastic bob Silver badge

      Re: OS/2 and PS/2 Memories

      Back in the 90's, a few months before the release of Windows 3.0, I took an OS/2 presentation manager programming class at the local (night school) city college. Got an 'A'. Only 6 students survived to complete the class, and the room was packed on day 1 when I thankfully had my add slip signed by the prof... [and we had "the PS/2 machines" in the lab to ourselves, since they were the only ones that could run OS/2].

      And I really _LIKED_ OS/2 PM. I was able to format a diskette while doing OTHER THINGS, kinda cool because DOS could _NEVER_ do that! Version 1.2 was nice looking, too, 3D SKEUOMORPHIC just like Windows 3.0 would soon become!

      But when i tried to BUY it, I ran into NOTHING but brick walls. It was like "get a PS/2, or wait forever for OEMs to get it 'ported' to THEIR machines". Bleah.

      THAT is what killed it. NOT making it available for CLONES. When 'Warp' finally released, it was too little, too late.

      But the best part of OS/2 was it's API naming, which follows the MORE SENSIBLE object-verb naming, rather than verb-object. So in Windows, it's "CreateWindow". In OS/2, it's "WindowCreate". And wouldn't you know it, when you read the DOCUMENTATION all of the things that work with WINDOWS are in the SAME PART OF THE MANUAL!

      Damn, that was nice! OK I have hard-copy manuals for OS/2 1.2 still laying about somewhere... and corresponding hard-copy Windows 3.0 manuals that I had to thumb back-forth with all the time. Old school, yeah. HARDCOPY manuals. And actual message loops (not toolkits nor ".Not" garbage).

      1. Pompous Git Silver badge

        Re: OS/2 and PS/2 Memories

        "THAT is what killed it. NOT making it available for CLONES. When 'Warp' finally released, it was too little, too late."
        A friend who was a developer at the time says the main thing that killed OS/2 was the cost of the SDK: over a $AU1,000. BillG was giving away the SDK for Windows at computer developer events.

  7. Lord Elpuss Silver badge

    "But it's striking as the world moves to three (or four) computing platforms, IBM isn't one of them."

    IBM is very much one of the top players. In Cloud, I would say AWS, Azure and IBM are the top 3. Business Intelligence? Oracle, IBM, Microsoft. AI/Cognitive? IBM Watson, Google DeepMind are the big two, Microsoft Cognitive coming a far third (along with highly innovative smaller vendors with a great solution but lacking scale - these will be swallowed by the big 2 (or 3) in short order).

    Don't discount Big Blue too early.

    1. Danny 14

      plus they can make a killing selling IPv4 addresses if they want to.

  8. wolfetone Silver badge

    The IBM PS/2 Model 70 was my first PC, bought for me on my 10th birthday in 1997. Knew nothing about the computers, so it got thrown out a year later for an upgraded Acer 486. It had a huge impact on me, probably the one thing that got my love of computers and computing in general going.

    In 2012, after many a weekend looking at eBay, I found a Model 70 for sale in London for £50. I got in the car, drove down and picked it up. I've other more interesting pieces of technology in my collection, but the Model 70 is the jewel as far as I'm concerned.

    1. Peter Gathercole Silver badge

      When the IBM AIX Systems Support Centre in the UK was set up in in 1989/1990, the standard system that was on the desks of the support specialists was a PS/2 Model 80 running AIX 1.2. (I don't recall if they were ever upgraded to 1.2.1, and 1.3 was never marketed in the UK).

      386DX at 25MHz with 4MB of memory as standard, upgraded to 8MB of memory and an MCA 8514 1024x768 XGA graphics adapter and Token Ring card. IIRC, the cost of each unit excluding the monitor ran to over £4500.

      Mine was called Foghorn (the specialists were asked to name them, using cartoon character names).

      These systems were pretty robust, and most were still working when they were replaced with IBM Xstation 130s (named after Native American tribes), and later RS/6000 43Ps (named after job professions - I named mine Magician, but I was in charge of them by then so could bend the rules).

      I nursed a small fleet of these PS/2 re-installed with OS/2 Warp (and memory canalized from the others to give them 16MB) for the Call-AIX handlers while they were in Havant. I guess they were scrapped after that. One user who had a particular need for processing power had an IBM Blue Lightning 486 processor (made by AMD) bought and fitted.

      1. Danny 14

        I remember the day the boss signed a huge contract to strip out the IBMs and put Gateways in instead. 200 machines, and in 1990 it was brand new 486 33s (not the x2 ones), it was a fortune. Out with IBM and in with DOS, Windows for workgroups and good old novell netware logon scripting. Good days and it all pretty much worked. we even had Pegasus mail back in the day and a JANET connection.

  9. kmac499

    Back in them days the old "No-one got fired for buying IBM" was still rampant, and the site I was on, or more accurately the tech support crew, backed PS/2 OS/2 for that very reason.

    The wheels started to come off quite quickly with the cost and availability of MCA cards.

    The other thing that really killed it isn't mentioned in the article. That was Token Ring networking, anyone else remember Madge cards?. I'm no network guy but IIRC it was all very proprietary, and an expensive pig to modify once installed (MAU's ??). Unlike the XEROX Ethernet system with simpler cabling and connectors,

    Of course some PS/2 architecture does survive to this day. The keyboard and mouse connectors; all that work and all that's left is a colour coded plug .....

    1. Danny 14

      terminator monsters. Much easier to cable and you knew that just about any network issue was someone putting a mouse driver on the startup disc (not enough conventional memory to load the network driver) or someone had nicked the terminator.

    2. 2Nick3

      I still have one of those devices to charge the ports in a CAU (or was it MAU - it's been a while. One was unpowered, the other plugged into mains) in my "Interesting things I've used in IT" box. Took a 9V battery and charged up the capacitors so when you plugged in a cable it would be detected and the client system would connect. If a port on the CAU wasn't active for too long the capacitors would lose their charge and you'd have to dig one of these out to get it back online.

    3. Bandikoto

      I wrote fixed the Madge Tolkien Ring network card driver when I worked at Telebit, long, long ago, in a Valley far, far away. I'm pretty sure that was part of the work to get IPX working properly in Fred, as well. Token Ring was rated at 60% faster than thin Ethernet, and actually much faster at Load, but a real pain in the behind to work with.

    4. Franco

      They never really learnt the lesson of backing the wrong standards, or at least kept trying to promote their own.

      I remember trying to get a game working on my cousin's PC, which was an IBM Aptiva and it must have been in the mid 90s or so as it ran Win95 and had a Pentium chip in it. Certainly it was long after sound cards had settled on either Sound Blaster or AdLib compatibility, but IBM were persisting with MWave, their combined modem/sound cards that were compatible with feck all.

      1. Sandtitz Silver badge

        "IBM were persisting with MWave, their combined modem/sound cards that were compatible with feck all."

        MWave modem/sound cards were still uses Thinkpads in very early 2000s.

    5. ecofeco Silver badge

      I DO NOT miss Token Ring.

    6. aqk

      Token-Ring? Ugh! Sorry, Madge! IRMA is here

      Se my comments elsewhere here about the shitty token ring.

      I thing THIS -Token Ring - is when IBM began to fall apart, in spite of all their FUD about Ethernet..

  10. Peter2 Silver badge

    I remember the pictured computer in the article, and this was typed on the (pictured) model M keyboard that came with that computer.

    They don't build equipment like that any more.

    Which is probably just as well, given that the keyboard weighs practically as much as a modern laptop notebook.

    1. Anonymous Coward
      Anonymous Coward


      I still have a model M on my home PC. Only real issue with it is that it doesn't have a Windows key. I also wonder how long motherboards will come with a PS2 keyboard port.

      Nothing I have ever used siince comes close to it.

  11. BarryProsser

    Perspective of the PC developers?

    As a retired IBM UK employee, I have followed The Register’s articles for years. I like this article more than most of the ones written about IBM. I would love to hear the perspective of the IBMers who were directly involved in the PC developments 30 years ago. I doubt many of them read or participate here. Where best to engage them? Yahoo, Linkedin, Facebook, others?

    1. Bandikoto

      Re: Perspective of the PC developers?

      Do you participate in your local IBM Club? I know that at least one still exists in San Jose, California. There's an IBM Club in Boca, which seems like a good place to start, given that was the home of the IBM Personal Computer personal computer.

      1. BarryProsser

        Re: Perspective of the PC developers?

        Bandikoto, Thanks for the reply. No, I do not participate in any local IBM clubs. I am not aware of any IBM Clubs in the UK. I tried your link to the IBM Club in Boca. I do not meet the membership requirements to join it.

        1. W.S.Gosset
          Thumb Up

          Re: Perspective of the PC developers?

          Barry, not sure how old you are but if connecting with UK people with a similar big-metal history is at least _some_ substitute for purely IBM experience, perhaps check out the ICL crew. Old-skool bigmetal competitors with IBM, in a friendly way -- the companies started about the same time, just on different continents.

          They basically keep in touch and share reminiscences & stories through their pension newsletter, sponsored by the new owners of ICL, Fujitsu.

          Check out the 2 collections of old big-metal anecdotes here:

          These are absolute classic stories for anyone really, not just bigmetal types. How often nowadays does the installation process involve digging a vast hole with bulldozers, dumping loads of scrapped cars into it all cabled together, covering it all back up again with dirt, and watering it? I can't _remember_ the last time I did that.

          Social site: -- esp.see the List of Friends link.

          And here's a newsletter, with an organiser's email/phone prominent whom you could contact:

          Those newsletters have brilliant stories embedded in them, by the way. eg: "Bert Morton and Jim Woodhead were members of Doc Keene's development team during the 1939-45 war, and to us new starters, they were characters in their ownright. Bert never used an Avo to time camshafts on his machines. 110v (the general machine voltage) was his accepted safe voltage and he generally used his right hand with his two middle fingers withdrawn to his palm, and index and little fingers extended as probes, to "feel" the open and closed voltage conditions at the cam contact points. He warned his team against trying to emulate his success in this measuring technique."

    2. niksgarage

      Re: Perspective of the PC developers?

      I'd like to say thank you for the memories as well. I retired from IBM in 2005, but I am still working (divorces do that to you), and I was in the thick of things in Boca in the 80s. El Reg is one of my morning go-to sites along with Slashdot and Hackaday, so as far as I am concerned, if it's here I will read it!

    3. pyrasanth

      Re: Perspective of the PC developers?

      May get lost here but there are lots if IBM clubs in the U.K. Google IBM Hursley Club for example. You'll find them wherever there was a large IBM presence. North Harbour, Warwick, Basingstoke, Southbank and Greenock are likely.

      Ex-Greenock employees worked on PC support as well as manufacturing a load of stuff.

      Also you may find someone on the watchingibm page on Facebook. It's mainly a follow in from the endicott alliance but a lot of IBM and ex-IBM read it.

      Hope that helps!

  12. John Smith 19 Gold badge
    IT Angle

    Although technically OS/2 is still around

    As EComStation

    In hindsight (always 20/20) IBM's mistake was to seek to start making money off of MCA ASAP. Had they had been much more reasonable they would have had a little bite of every board (through licensing) made and MCA would have dominated.

    BTW let's not forget what a PoS Windows 2.0 was or how the 286 was so retarded that the MS tech who worked out how to switch it from real to virtual mode and back was hailed a f**king genius.

    1. Anonymous Coward
      Anonymous Coward

      Re: Although technically OS/2 is still around

      The 286 designers never thought that someone would have wanted to return to the much more limited real mode once into the much more advanced protected (not virtual) mode. Entering protected mode was easy - just set the proper bit. Getting out was "impossible" - but through a reset.

      After all, back then, backward compatibility wasn't still a perceived issue. Probably Intel believed everyone would have rewritten the software to work in the new, better, protected world. It turned out it was wrong (just to make the same mistake years later with Itanium, though....).

      The problem was not the 286 itself (and IMHO we'll see the segmented memory model back again one day because it implements a far better security model...), it was most DOS advanced application were written to bypass DOS itself and access the BIOS and HW (RAM and I/O ports) directly, something that was hard to emulate on a 286, and made porting to other OS more difficult.

      Probably CP/M applications were better behaved, and a "protected" CP/M would have been easier to develop.

    2. Paul Crawford Silver badge

      Re: 286

      The article has one significant mistake - the 286 did support protected operation, even the option for "no execute" on memory segments. But as it was designed to be either 'real mode' 8086 compatible OR protected mode you had the fsck-up of having to use a keyboard controller interrupt to bring it out of halt state back to 'real' mode.

      The major advances for the 386 were:

      1) 32-bit registers

      2) The "flat" memory model and virtual memory support (not the 16-bit segments of 286, OK still segments but way big enough for a long time)

      3) The option to easily change protection modes.

      1. Anonymous Coward
        Anonymous Coward

        Re: 286

        The segmented mode has more than the "no execute" bit. It has an "execute only" mode - the CPU can execute the code in the segment, but the application can't nor read nor modify if (goodbye, ROP!). Still, managing segments adds complexity, security checks slow down execution, and applications are less free (which, from a security point of view, is a good thing).

        The "flat memory model" was not an inherent new feature - it means just to use large segments, large enough the application never need to load another one (which usually meant the whole application address space). Also, usually, code and data segments overlap to make everything "easier" (and far less secure).

        286 16-bit segments were too small to allow for that. 386 32-bit segments allowed that, while the new pagination feature allowed for "virtual memory" with a page (4k) granularity. 286 virtual memory worked at the segment level - with 64k segment it could work, but with flat 4GB segments it couldn't work, obviously.

        But what made the 386 able to run DOS applications was the Virtual 86 mode. In that mode, the hardware itself trapped direct accesses to memory and I/O ports, and allowed the OS to handle them, without requiring complex, fragile hacks.

        This mode is no longer available in 64 bit mode, and that's why Windows 64 bit can't run DOS applications any longer (Windows supports console applications which are native Windows applications, not DOS ones).

    3. Richard Plinston

      Re: Although technically OS/2 is still around

      > BTW let's not forget what a PoS Windows 2.0 was or how the 286 was so retarded that the MS tech who worked out how to switch it from real to virtual mode and back was hailed a f**king genius.

      It was an IBM tech that worked out how to switch the 80286 back to 8086 mode using the keyboard IF chip, there was a standard instruction to switch it to protected mode. The mechanism was incorporated into OS/2 1.0, MS stole that for Windows/286.

  13. David Lawrence

    Ah yes I remember it well

    I was working for IBM at the time, at their UK Headquarters, so I got my free copy of OS/2 Warp.

    There was a big demo with loads of PCs all running Windows-based games and software. I hit problems installing it on my PC because I wasn't prepared to re-partition my hard drive and lose all my existing data. Rubbish.

    I watched OS/2 go down the drain, followed by several other doomed products. Then the bottom dropped out of the mainframe market and things took another slide. The 'partnership' with Lotus was a bit of a disaster too.

    IBM? They just do software and services right? And a lot (but not all) of the software was obtained through acquisitions. I remember when they actually made stuff - like a PC keyboard that cost £120.

    Shame really.

    1. Danny 14

      Re: Ah yes I remember it well

      lotus notes still makes me blink asymmetrically when I see pictures of it. <shudder>

  14. adam payne

    I loved the model M keyboard such a fine piece of design and engineering.

  15. Anonymous Coward

    IBM was its own worst enemy

    It's been a while but back in the days I was a serious OS/2 advocate. Look, if you even get other people to end up trying out OS/2 because they became sick and tired of Windows 3.11 often bodging up and not being able to network properly then yeah...

    But IBM more than often didn't even seem to care all that much. Looking back I think it was a bit the same as the stories we get to hear about Microsoft now: how divisions in the company do different things, don't always work together and in some rare cases even compete. Even at the expense of customers if they have to!

    But IBM... I enrolled in the OS/2 support program (I seriously don't remember how I pulled this off anymore, I think I asked (and got) permission from my work to look into all this and also use their name) which ended up with IBM sending me several beta versions of OS/2 products. Including several OS/2 server environments. It was awesome. OS/2 server (a green covered double CD, that much I remember) was basically OS/2 with additional user management and network configuration settings.

    Yet the funniest thing: IBM couldn't care less about your test results. At one time I got an invitation to go to IBM in the Netherlands for an OS/2 server demonstration which would also showcase some of their latest product (I recall being showed a very lightweight laptop). At arrival you had to search for the entrance and where it all was, because any announcements or directions were no where to be found on site.

    I bought OS/2 3.0 Warp and the 4.0 Merlin and it always worked like a charm. I seriously liked OS/2 much better than anything else. So when I had the opportunity to buy a PC through my work it was obvious what I would need to get, right? An IBM Aptiva. That would be an ultimate, the thing to get for OS/2. Because obviously an IBM OS will definitely run on IBM hardware, right?

    Context: this was at the prime of my OS/2 endeavors. I could optimize and write a config.sys file from mind if I had to, I knew what drivers to use, which to skip, what each command did. Memory optimization? Easy. Bootstrapping a *single* floppy disk to get an OS/2 commandline? Hard, yet not impossible (try it, you'd normally get multiple disks to boot with).

    It took me one whole weekend, dozens of phonecalls to the IBM support line, and the conclusion was simple: IBM did not care about OS/2 for their own hardware. And with that I mean not at all. It did not work, no matter what I tried. Even they told me that this wasn't going to work. Compaq out of all brands did care. Compaq, the brand which tried extremely hard to appeal to the general customer by making their hardware "easy" to use and also "easy" to customize (comparable to Dell a bit) didn't only target Microsoft and Windows. Noooo.... When I eventually ditched my IBM I got myself a Compaq and I also purchased an extra set of drivers and installation media (3 boxes of 3.5 floppy disks, approx. 37 in total) and guess what? Next to a full Windows 3.11 installation plus a different program manager and dozens of drivers it also included several disks with OS/2 drivers. I removed Windows and installed OS/2 that very same evening.

    Compaq... which often advertised that they made Windows easier. And also delivered OS/2 drivers for their harware...

    IBM, which made OS/2 also made hardware, never even bothered to provide OS/2 drivers for their own PC's. Not even if you asked them.

    Does that look like a company which cared?

    IBM was its own enemy sometimes.

    1. Uncle Ron

      Re: IBM was its own worst enemy

      IBM was and still is it's own enemy. So many comments above reflect this so well. Eg: "IBM's mistake was that it tried to make money right away from MCA." So true. IMHO, it is the MBA's permeating the entire company that are the enemy. They know nothing about the business IBM is actually in, only about cost recovery, expense containment, and fecking business models. For the last 30 years, the real heroes in IBM have been the ones who cut the most, or spend the least, or pound suppliers the worst.

      This virus is especially dangerous when a non-MBA contracts it. When they see who gets the most recognition, they can't wait to de-fund sales commissions or training programs or development staffs. They think they are "doing good." It is not only true in IBM. Companies all over the West are infected with the idea that reducing costs (and innovation) towards zero, and increasing revenue towards infinity, is all we should be working on. So, fewer Cheerios in the box, fewer ounces of pepper in the same size can, cut the sales-force, reduce the cost (and quality) of support, and on and on and on.

      If there is one word that summarizes this disease, and a word I cannot stand to hear in -any- context, it is the word, "Monetize." It encapsulates all the evils of what I feel is the "Too Smart by Half" mentality. I cannot say how many times I have heard the phrase, "how much money are we leaving on the table?" or, "how many more will we sell if we..." and the room goes silent and a good idea is dropped.

      I am sorry I am rambling, I am sad. Never be another System/360 or Boeing 747. Incremental from here on out. Elon Musk doesn't seem to be infected...

  16. Anonymous South African Coward

    Single Input Queue

    Gotta love the silly old SIQ... one app chokes the SIQ and you can do absolutely nothing, except hard reboot :)

    Fun times indeed. I had a couple of SIQ incidents as well. All but forgotten, but recalled to memory now :) Sometimes a CTRL-ALT-DEL would work, sometimes not.

    And who remember the CTRL-ALT-NUMLOCK-NUMLOCK sequence?

  17. Kingstonian

    Memories - Nostalgia isn't what it used to be.

    It all worked well and was of its time. PS/2 and OS/2 made sense in an IBM mainframe using corporate environment (which I worked in) with some specialised workgroup LANs too. OS/2 EE had mainframe connectivity built in and multitasking that worked. Token Ring was much better for deterministic performance too where near real time applications were concerned and more resilient than ethernet at the time - ethernet would die at high usage (CSMA/CD on a bus system) whereas token ring would still work if loaded to 100% and just degrade performance gracefully. Ethernet only gained the upper hand in many large corporate environments when 10 base T took off. Token ring would connect to the mainfame too so no more IRMA boards for PCs

    There was OS/2 software available to have a central build server where each workstation could be defined on the server and then set up via the network by booting from floppy disk - useful in the corporate world. DB/2 was available for OS/2 so a complete family of useful tools was available. And IBM published its standards

    IBM was used to the big corporate world and moving down to individuals via its PCs whereas Microsoft at that time was more individual standalone PCs and moving up to corporate connectivity. The heritage still shows to some extent. Novell was still the LAN server of choice for us for some time though.

    The PS/2 was easy to take apart - our supplier showed us a PS/2 50 when it first came out. He had to leave the room briefly and we had the lid of the machine and had taken it apart (no tools needed) before he returned. He was very worried but it was very easy just to slide the parts back together and they just clipped into place - not something you could do with other PCs then. I came across an old price list recently - the IBM model M keyboard for PS/2 was around £200 (without a cable which came with the base unit- short for the model 50 and 70 desktops and long for the 60 and 80 towers! Memory was very expensive too and OS/2 needed more than DOS. In fact EVERYTHING was expensive.

    OS/2 service packs (patches) came on floppy disks in the post. You had to copy them and then return them!

    Starting in computing just after the original IBM PC was announced this all brings back fond memories and a huge reminder of the industry changes.

    1. addinall

      Fake News. Re: Memories - Nostalgia isn't what it used to be.

      Your memory is shot. OS/2 was developed my Microsoft AND IBM, first released jointly in December 1987. Bill Gates was at Wembley flogging the OS during 1988.

      Way before then Microsoft had UNIX System 7 running and released as XENIX for the 8086/80286 architecture.

      The development of OS/2 began when IBM and Microsoft signed the "Joint Development Agreement" in August 1985. It was code-named "CP/DOS" and it took two years for the first product to be delivered.

      OS/2 1.0 was announced in April 1987 and released in December. The original release is textmode-only, and a GUI was introduced with OS/2 1.1 about a year later. OS/2 features an API for controlling the video display (VIO) and handling keyboard and mouse events so that programmers writing for protected-mode need not call the BIOS or access hardware directly. In addition, development tools include a subset of the video and keyboard APIs as linkable libraries so that family mode programs are able to run under MS-DOS. A task-switcher named Program Selector is available through the Ctrl-Esc hotkey combination, allowing the user to select among multitasked text-mode sessions (or screen groups; each can run multiple programs).

      Communications and database-oriented extensions were delivered in 1988, as part of OS/2 1.0 Extended Edition: SNA, X.25/APPC/LU 6.2, LAN Manager, Query Manager, SQL.

      The promised graphical user interface (GUI), Presentation Manager, was introduced with OS/2 1.1 in October, 1988.[9] It had a similar user interface to Windows 2.1, which was released in May of that year. (The interface was replaced in versions 1.2 and 1.3 by a tweaked GUI closer in appearance to Windows 3.1).

      OS/2 occupied the "Intellegent Workstation" part of SAA (Systems Application Architecture) and made use of the APPC PU2.1 LU6.2 SNA network stack.

      1. Anonymous Coward
        Anonymous Coward

        "The development of OS/2 began..." etc. etc.

        That mostly looks copy & paste from Wikipedia. A link would have been enough

      2. Anonymous Coward
        Anonymous Coward

        Re: Fake News. Memories - Nostalgia isn't what it used to be.

        OS/2 was developed my Microsoft AND IBM

        And that's why an OS/2 subsystem lingered in Windows for some time.

        (They finally killed it, right? And the POSIX subsystem?)

        1. W.S.Gosset

          Re: Fake News. Memories - Nostalgia isn't what it used to be.

          >And the POSIX subsystem?

          Dunno about "subsystem". NT kernel has always been posix compliant. That's why things like (godsend) CygWin work -- they just recompiled the unix shell (+ core tools) on NT. And all current Windows postWin95 are NT.

          I am midst salvaging a major project with 0 resources courtesy of NT's posix compliance -- aka CygWin. So, no, MS didn't kill posix compliance. TBH, I doubt anyone in senior mgt is even aware what it means, let alone have the mojo to lead a rewrite of the kernel to remove it.

          One quiet, but quietly major, triumph for the engineers.

  18. Primus Secundus Tertius Silver badge

    No Stop button in window

    The big defect in OS/2 that I met was the lack of a stop button in any window. Yes, you could close the window but that did not stop the task, just left it sitting in the background.

    It was Windows 95 that brought us a proper stop button.

    We had an OS/2 system returned to us as not working. The user had been closing the window, and then later starting up another instance. The system was clogged with dormant tasks. Once I removed them, everything worked again; what we had to do then was to update our User Guide.

    1. BinkyTheMagicPaperclip Silver badge

      Re: No Stop button in window

      No it did not. Closing the main application window killed the app in all the instances I can think of. Minimising is a different matter, of course.

      The only default I can think of that was very unhelpful was the automatic restarting of all applications and folders that were running when shutdown was requested. Easily fixed in config.sys, but should never have been a default.

    2. Anonymous Coward
      Anonymous Coward

      Re: No Stop button in window

      "Stop" button? You mean the "Close" icon in the window titlebar? Before Windows got a specific icon, it was achieved by double clicking the "system" icon - that that looked like [-] in Windows before 95, and the used the application icon. IIRC OS/2 worked the same way.

      It still works in some application which hide the system menu icon.

  19. Frish

    IBM what could have been

    At one point, OS/2 could run multiple PC-DOS Windows "Better than Microsoft can" since memory was partitioned, and a crash in one window wouldn't affect the whole machine. Microsoft wrote a stub to detect whether OS/2 was installed, and killed the attempt to load PC-DOS...

    Where IBM Boca missed the boat was thinking that this is a "personal" computer, and therefore, not a "commercial" one. IBM should have OWNED that market, entirely, and the competing product within the corporation that lost to Boca's product recognized that but got shoved aside by the sexiness of the PC, and all the departures from Big Blue tradition, etc.

    Also, the way IBM execs got paid meant they were shortsighted about 'solutions' that included IBM Products that they didn't get paid for.

    As Product Marketing Manager for IBM CICS 0S/2, (announced at Comdex '89, a Sr. VP from MIcrosoft shared with me on the show floor "That's the most important announcement in this entire show" as I handed him a press release THAT WAS NOT included in the show's press kit, since the PC division was in charge, and they kept other Division's products from being included...) I tried to get the then President of the PC division to just whisper CICS OS/2 to the management of a very large Insurance firm. He would have left with a 40,000 PC order, but, instead, chose to say nothing...IDIOTIC but true.

    1. Primus Secundus Tertius Silver badge

      Re: IBM what could have been

      "...Microsoft wrote a stub to detect whether OS/2 was installed..."

      Windows 95 was distributed to end users as an "update CD". It would not run unless it detected an installed Windows 3.x or was presnted with the first W3 floppy disk. It would also accept the first OS/2 Warp 3 floppy disk.

      1. Pompous Git Silver badge

        Re: IBM what could have been

        "Windows 95 was distributed to end users as an "update CD". It would not run unless it detected an installed Windows 3.x or was presnted with the first W3 floppy disk. It would also accept the first OS/2 Warp 3 floppy disk."
        That's because Warp included a full Win 3.x license.

  20. Jim 59


    I remember seeing a demonstration of GEM at the PCW show in 1986 or 1987, at - Olympia I think it was. Very impressive it was too. Didn't it also come bundled on some Amstrad PCs ?

    1. Richard Plinston

      Re: GEM

      > Didn't it also come bundled on some Amstrad PCs ?

      Yes. They came with both DRI's DOS+ and GEM and MS-DOS.

      It was in all Atari 512 (and derivatives) running on TOS, which was written by DRI. It also came with BBC Master 512 which ran DRI's DOS+ and GEM on an 80186 (or was it 80188?) processor.

  21. Justin Clift

    OS/2 resurrection

    Just to point out, OS/2 is being resurrected as "ArcaOS":

    They're been working on it for a while, and (though I've not used it), apparently it's targeted for release in under two weeks:

    Modern Qt 5.x has been ported to it, and many Qt based apps are apparently working properly.

    Saying this as DB Browser for SQLite is one of them. One of our Community members has been keeping us informed. ;)

  22. Jim-234

    OS/2 Desktop virtualization before it was cool

    I used to run OS/2 very well on a 386 based PC

    I could have several windows open each with different virtual OSes, and you could move around the virtual OS image files etc.

    So for example if you had some odd networking hardware (like LanTastic) that only wanted a specific DOS version, well spin up a new little dedicated Virtual OS image for it.

    While people think all this virtualization stuff is so new.. it was around 25 years ago and worked rather well considering the hardware you had available at the time.

    It's a shame Microsoft was able to blackmail IBM into discontinuing OS/2, with a different vision OS/2 might have become a serious contender for workplace use.

    1. Handlebar

      Re: OS/2 Desktop virtualization before it was cool

      Er, IBM were doing virtual machines in the 1960s ;-)

    2. Pompous Git Silver badge

      Re: OS/2 Desktop virtualization before it was cool

      "It's a shame Microsoft was able to blackmail IBM into discontinuing OS/2"
      Say what? Are you crazy?

  23. VanguardG

    The quote from Mr Watson was supposedly in 1943 - when computers were larger than many houses and weighed as much as several buses. To say nothing of being extremely pricey (since they were essentially built on-site by hand) and expensive to operate, since they needed to be staffed by a crew trained in just that one machine, plus the power needs were enormous, and the heat produced prodigious. And, at that time, you really had fairly limited tasks that needed doing that required that kind of number-crunching. Did he say it? Maybe not, but given the time frame, it really doesn't seem as boneheaded as it would have sounded 15 years later.

    Microchannel *would* have been sweet, were it not for the expense. A basic sound card of the time was $70 for ISA - the same card in MCA was $150.

    As for plugging in a keyboard after boot...I still find it amusing that someone actually wrote an error code of "Keyboard not detected. Preset F1 to continue." If there's no keyboard, there's no F1 to press.

    1. Primus Secundus Tertius Silver badge

      There are other variations of that quote about five computers: e.g. the UK would only need that kind of number.

      For the publicly known computations of that time - gunnery trajectories etc., that number is perhaps right. But I believe over a dozen instances of Colossus were built for code cracking. So even then the estimate of five was way out.

      However the real expansion of computing came with data processing, where record keeping outweighed the relatively small amount of computation. IBM should have known that, given their existing business in accounting machines fed with punched cards.

      1. Richard Plinston

        > So even then the estimate of five was way out.

        Actually he was speculating on the sales of the particular model that they were currently building. He was counting the number of government agencies that would be able to afford them and find them useful.

        It was only much later that anyone tried to use computers for commercial purposes that would find them a place in businesses: LEO - Lyons Electronic Office was developed for payroll, stock distribution and manufacturing (of cakes).

        In the 1950s the British government were deciding where _the_ computer would go. They chose a town that was a major railway junction because then the train loads of punch cards could easily be shipped to it.

    2. Updraft102

      Well, yeah... there's no F1 to press, so if you want the PC to boot and you see that message, you will have to plug a keyboard in and press F1. That and make sure the keyboard lock is in the off/unlocked position, which is probably more what the keyboard check/stop on fail configuration was about than anything else.

      The keyboard lock was an actual physical lock on the front of the PC that used a round key (like on vending machines) which, if set to the locked position, would prevent the PC from detecting the keyboard or responding to keypresses.

      Enabling the "Press F1 to resume on keyboard error" BIOS setting made the keylock into a rudimentary system protection device. It wasn't all that effective against anyone who could get into the computer case, as bypassing the lock was as easy as pulling the cord for the keylock off the motherboard, but PC cases also typically had provisions for a small padlock to keep the cover on the case back then too. It wasn't great protection, but it probably provided some relief from pranksters, who probably would not be determined enough to cause physical damage to someone's PC for a joke.

      1. VanguardG

        I remember - and people were perpetually misplacing the key to their padlocks. The round keylocks require a lot of expertise to crack open, but the padlocks...well...they were nearly always bought cheap. Basic (legal) hand tools could pop them open in seconds, without any damage to the case and, at most, a scratch or two on the padlock. Some people who were kinda serious had monitor stands that had a built-in, locking keyboard drawer. But those were usually employed by people who had a "favorite" keyboard they were afraid would get stolen by a jealous co-worker rather than because of any actual security concerns.

  24. Tim Brown 1

    The UK had the best tech for personal computers at the time

    For PCs during that period, in pure tech terms , Acorn's ARM machines running RISC-OS were way ahead of offerings from anyone else and prior to that the BBC micro (built by Acorn).

    It's just such a shame that Acorn lacked any international marketing savvy then.

    1. Richard Plinston

      Re: The UK had the best tech for personal computers at the time

      > It's just such a shame that Acorn lacked any international marketing savvy then.

      And yet Acorn became a worldwide powerhouse chip design expert that currently sells licences for billions of chips every year. Even before phones started using them ARM were selling tens of millions of licences for chips to power embedded equipment (modems, routers, PABX, ...).

      ARM = Acorn RISC Machines

    2. jake Silver badge

      Re: The UK had the best tech for personal computers at the time

      Not really. I had a 16 bit Heath H11A personal computer in 1978. Acorn didn't ship a 16 bit computer until 1985 ... The old girl still runs. Loudly.

  25. Anonymous Coward
    Anonymous Coward

    "Thomas J Watson may never have said the world will have five computers"

    It is funny that people take this statement to be a sign of IBM not understanding the potential of the computing market, if it was ever even said. It actually makes a lot of sense though. TJW, correctly, didn't think it would make sense for every business out there to go build their own little data center and a "best we can do" computing infrastructure. Better to let a giant time share (now called cloud provider) handle all of that complexity and just rent the resources as needed. It is kind of like saying there is only room for a handful of electrical utilities in the world. Even if everyone, at MSFT's urging, went out and bought a gas powered generator for their house... it still makes sense that there is only room for a handful of utilities.

  26. JohnCr

    A tiny bit more

    A few comments.

    1) IBM was STRONGLY urged to skip the 286 and create a true 32 bit, 386 version of OS/2. Even Microsoft was strongly pushing IBM in this direction. IBM's resistance to do so was a major contributor to the split between IBM and Microsoft.

    2) The MicroChannel was better than the ISA bus. The problem was it was not good enough for the future. The PC evolution was moving towards faster graphics and network (100 mbps) connectivity. The MicroChannel, even 3 generations into the future did not have the bandwidth to meet these needs. The industry evolved to the PCI interface. As an interesting historical coincidence, PCI uses the same type of connectors as did the MicroChannel. And the PCI interface found its way into other IBM systems.

    3) The IBM PC was inferior to Apple's products. The PC became more successful because a new industry and IBM's customers worked together to make it successful. (Steve Job - The Lost Interview) In 1987 IBM turned a deaf ear to the industry and its customers. When IBM stopped listening its fortunes turned. This culture took over the whole company and was a major factor in the company almost going out of business in the 1990's.

    1. Richard Plinston

      Re: A tiny bit more

      > Even Microsoft was strongly pushing IBM in this direction.

      Microsoft had developed its own 286 versions of MS-DOS: 4.0 and 4.1 (not to be confused with the much later 4.01). These was also known as European DOS because the were used by Siemans, ICL (where I worked) and Wang. These versions supported a limited multitasking of background tasks and one foreground program. I have a manual here on how to write 'family' applications that would run on 8086 MS-DOS or in protected mode on 80286 MS-DOS 4.x.

      It was dumped when they switched to writing OS/2 with IBM.

    2. niksgarage

      Re: A tiny bit more

      MCA was an architecture that worked fine in workstations, just not PCs. I met Fred Strietelmeyer - the architect of MCA (fairly sure that's his name) in Austin, TX in the 80s who told me that MCA was a fine architecture, but not all implementations were. RS/6000 had multi bus-mastering working with MCA, mostly because the address on the channel was a logical address, which went through the memory manager, so AIX could easily control and protect the physical memory space. PS/2 used physical addresses, which meant that either bus mastering was turned off, or the bus mastering cards needed to have a copy of the memory manager on board as well. If you were running AIX, MCA was not a problem or even a question deserving of five minutes of thought.

      The PC industry hated MCA, the connector, the architecture and its licencing. They came out with EISA - a backward-compatible connector to extend the AT bus. I always found it a huge irony that PCI used the same physical connector as MCA years later.

    3. niksgarage

      Re: A tiny bit more

      AND you are exactly right about the 286/386 wars. I was in Boca when the AIX guys from Austin came to see the CPDOS developers (as OS/2 was known in Dev), and showed them true multi-tasking on 386. They were baffled why IBM was writing another OS when we already had a virtualising, 32-bit ready pre-emptive multi-tasker that ran on multiple hardware platforms. It's only issue was it couldn't run on 286. And for that reason alone, IBM spent enough money on OS/2 development that they could have paid for the Hubble telescope AND had it repaired.

      I also saw the numerous prototype machines in Boca (one called 'Nova' in dev as I recall) which had a 16MHz 386, lots of memory on board and an AT expansion bus. (Also had the 1.44MB diskette drive) Nice machine, could have sold really well. Only the Model 30 was allowed to be shipped, in case the AT bus continued to outshine MCA.

  27. Marshalltown

    What grief?

    It always puzzled me what grief OS/2 was supposed to create. I used it as a substitute for Windows, running windows s/w into the early years of the century. I gathered that IBM might have still been attempting to extract its pound of flesh from developers, but as far as I was concerned, it worked fine. I built my own machines and it ran on them without problems. I also liked Rexx as a scritping language. It was immensely more useful than does and much less of a pain (to me) than MS BASIC and all its little dialects and subspecialties. The only real grief I encountered was developers who "simply couldn't" do a version for OS/2 - and, of course, MS doing their best to see that their software was less compatible than need be.

    History seems to forget how thoroughly MS would break things with each "improvement." Many of the useful "improvements" in Windows were first present in OS/2 and some really nice features vanished when IBM decided the effort wasn't worth the candle. The browser with its tree-structured browsing history was remarkable. No browser since has and anything to match. Even now, relicts of the OS/2 interface are still present in KDE and GNU. Microsoft has finally moved "on" with the horrible looking and acting interface of Windows 10.

  28. DJSpuddyLizard

    Ah... Concurrent DOS...

    IBM did actually use DR Concurrent DOS 286 - but in their 4680 Point-of-sale (described often as a real P.O.S. by those of us who used it) OS.

    1. Richard Plinston

      Re: Ah... Concurrent DOS...

      > IBM did actually use DR Concurrent DOS 286 - but in their 4680 Point-of-sale (described often as a real P.O.S. by those of us who used it) OS.

      Yes, that was a DRI product but it was not Concurrent-DOS it was FlexOS. This shared code with MP/M-86, as did Concurrent-DOS (neither of which had an 80286 product) but was 80286 based. The main difference was that FlexOS would not run MS/PC-DOS programs and Concurrent-CP/M-86 / Concurrent-DOS would run several of them at once (as well as CP/M-86 programs).

      DRI had pre-emptive multi-user, multi-tasking systems since 1978 with MP/M which ran on 8085 and Z80 micros with bank switched memory (I have a couple of RAIR Blackbox/ICL PC1s here and an ICL PC2 8085AH2 with 512Kbyte). MPM2 and MP/M-86 (for the 8086) were released around 1980. Concurrent-CP/M-86 with multiple virtual screens ran on an IBM-PC (and other machines - I have a stack of 8086 ICL PC2) and could used EEMS memory cards such as AST RamPage to get several Mbytes of memory and do context switching with just a handfull of register moves.

      Concurrent-CP/M-86 was demonstrated the same month as MS-DOS 2 was released. It had pre-emptive multi-tasking (and multiuser with serial terminals). The virtual screens were just a keystroke away so one could run SuperCalc, Wordstar, and other programs at the same time and just flick between them - even on the serial terminals.

      Later, this was developed for 386 into DR-Multiuser-DOS from which DR-DOS 5 and 6 were derived.

      There was a FlexOS-386 which had an enhanced GEM-X but it was dropped to concentrate on the Concurrent range.

  29. Randall Shimizu

    OS/2 failed because IBM has had this aversion to marketing over years. It is only recently that IBM appointed VP of marketing. Prior to the OS/2 Warp launch IBM had announced that it would launch a big advertising campaign But about a week prior to the IBM pulled the resources. There was lot of people and companies that were considering switching to OS/2 Warp. But MS launch a huge campaign for windows 95 and rest was history. IBM should have pressed it's advantage with client server computing. OS/2 was years ahead of Windows at the time.

  30. John Brown (no body) Silver badge

    PS/2 and OS/2

    The naming convention always struck me as odd. In my mind, the /2 meant divide by 2, ie half an OS and half a PS :-)

    1. AlanC

      Re: PS/2 and OS/2

      Ah - "OS divided by 2" - I'd never thought of it that way before.

      Interestingly, for the short time that MS were promoting it, they spelt it OS-2, which could I suppose be read as "OS minus 2".

      I'm not sure what if anything we can deduce from that!

    2. Uncle Ron

      Re: PS/2 and OS/2

      Nah. IBM (and many others) used the "/" for many, many products. Like System/360. It was a naming convention that seemed to lend weight, not division, to a product family.

  31. Ian Joyner Bronze badge

    Big Blue: IBM's Use and Abuse of Power

    To understand the lead up to PCs people should read Richard DeLamarter's Big Blue: IBM's Use and Abuse of Power.

    It goes back to the 1890s and shows how IBM became dominant through not such ethical practices. The antitrust suit against IBM was dropped by Ronald Reagan, which prompted DeLamarter to write the book.

  32. Mikel

    Partnering with their competitor on OS/2

    That was not a smart move.

  33. Pompous Git Silver badge

    Thinking back...

    As I mentioned in an earlier post to this thread, being able to continue using the computer while a print job was rendering would have been a very great boon in publishing. When we heard that you could do just that very thing under OS/2 Warp, The Git decided to suck it and see.

    No matter what I tried, OS/2 flatly refused to install on the testbed machine. IBM refused to provide support because the testbed machine wasn't an IBM. So I never got around to purchasing CorelDRAW! and PageMaker for OS/2. Many years later, I discovered that the reason OS/2 Warp refused to install was because the hard disk was "too big". "Big" hard disks were de rigeur in publishing as document file sizes tended to be large. Itty Bitty Minds presumably thought itty bitty hard disks were good enough for those too stupid to purchase their hardware.

  34. anthonywharris

    Make sure your driver strategy is in place if you launch a new O/S

    Good article Andrew. I was there in Miami when IBM launched the PS/2. They closed off the streets six blocks around the exhibition centre and as far as I remember the Beach Boys played. They should probably have saved their money. One thing you missed is that not only was OS/2 late but the driver support in the operating system was very poor. This meant that as well as blocking all the plug in cards through the new bus architecture, they also bricked all of the add on peripherals for the PS/2 that had worked with the IBM PC. Add to that the fact that the OS/2 driver team started competing with driver developers for driver business and also blocked them from developing for the architecture (until the OS/2 DDK made a brief appearance and then disappeared) and the factors that contributed to IBM's demise was complete. I recall that when the driver team saw the source code of one of our drivers some years later they threatened to sue us. That was until I pointed out that the code was based on the OS/2 DDK and they went quiet but couldn't quite believe that we had managed to obtain a copy in the few weeks that it had popped its head above the surface. Microsoft worked out early on that driver support is a key element to the success of an Operating System. Something that they seem to have lost sight of a bit with Windows Vista onwards although I suppose the switch to 64bit has made backwards compatibility more difficult. Keep the nostalgia coming Andrew, it's not like it used to be! Tony Harris

    1. John Styles

      Re: Make sure your driver strategy is in place if you launch a new O/S

      Yes, this a thousand times. Most of my experience of OS/2 was in the 1.2 / 1.3 days. The closest I have ever got to throwing a PC out the windows was trying to install the sodding thing on some random bit of hardware. Every time I wiped it and tried something else it somehow got less far. Microsoft were much better at running on random bits of hardware. The IBM approach was equivalent to the self driving car that works fine on clean roads with no pedestrians, fine in theory but useless in practice.

      The one message queue was a big mistake though - and IBM were a nightmare to deal with. To be fair I suspect that unless you are the CEO of a Fortune 500 company they are a nightmare to deal with. And I suspect if you are outside the top 100 companies they probably aren't THAT great.

  35. BinkyTheMagicPaperclip Silver badge

    Such a missed opportunity

    I didn't use OS/2 1.x at the time (only later), but beyond 1.0 it was fine for server based apps and a good, solid platform. Not so much for desktop apps - insufficient driver support, high memory requirements, and limited app support put paid to that.

    OS/2 2.x and beyond was a much improved proposition, but suffered in competition with the large number of Windows apps. The userbase were not, in general, prepared to pay more for a smaller amount of higher quality features - the reality of running a minority platform.

    OS/2 might have got further if IBM concentrated on Intel, but instead they wasted vast amounts of effort on OS/2 PPC. Much though I loved OS/2, the succession plan was flawed. Windows NT is simply better architected, they spent the time maintaining compatibility with 16 bit apps, and had much improved security, and multi user support. OS/2 was effectively dead before it really caused a problem, but it would have caused issues later on.

    System <n>/Mac OS were also flawed, and the early versions of OS X sucked, but Apple are much better at retaining compatibility whilst updating the OS (at least for a few years, until they drop old kit like a brick).

    I've still got an OS/2 system, and a lot of apps, and will be assembling an OS/2 1.3 system (because I'm a masochist and like trying OS). Haven't bothered with eComstation, but might give Arca 5.0 a go if it's any good, and not ludicrously priced. There aren't too many OS/2 apps I really want to run these days, though.

    One final note : it's *synchronous* input queue, not single. If messages are not taken off the input queue it hangs the interface, but does not stop apps running. There was a workaround implemented in Warp 3 fixpack 16, but until then a badly behaved app was a real pain. However, Win32 successfully moved away from the synchronous input queues in Win16, to asynchronous in Win32, without breaking too many apps. IBM should have put in the engineering effort to do the same.

    There are also some substantial differences between OS/2's architecture, and Windows (or indeed anything else). For instance the co-ordinate origin in Windows is at the top left of the screen, but in OS/2 it's the bottom left (OS/2 uses the mathematically correct option here)

  36. Carlitos911

    The "fun" takeaway from this is that while "industry analysts" have been consistently wrong for at least the last 30 years, we're still listening to what they say...

  37. FuzzyWuzzys

    Concurrent DOS, great stuff!

    One of the first multiuser systems I really learned as a spotty teenager back in the late '80s. Working with my Dad on getting shared DataEase databases working at his workplace. We had C/DOS on the "master" PC system and a couple Wyse terminals hanging off the serial ports, 3 systems independently using a shared, albeit cutdown, PC based RDBMS system. I loved it so much as a kid that I ended up with a career working in database systems.

  38. Paul Lahaie

    Better than OS/2 / Win9x / NT

    Around the time OS/2 was making its way onto the scene and most people use DESQview to multitask on their 286/386 PC, Quantum Software Systems (now QNX Software Systems) had a real-time multiuser, multitasking , networked and distributed OS available for 8086/8088 & 80286 processors.

    On PC/XT hardware in ran without any memory protection, but the same binaries would run on the real mode and protected mode kernel.

    Something about being able to do

    $ [2] unzip [3]/tmp/

    Would run the unzip program on node 2, read as a source archive the /tmp/ file on node 3 and extract the files into the current working directory.

    We accessed a local BBS that ran a 4-node QNX network (6 incoming phone lines + X.25 (Datapac) )

    Even supported diskless client booting and the sharing of any device over the network. Though at around $1k for a license, it wasn't "mainstream".

    It's too bad the few times Quantum tried to take it mainstream, the plans failed. Both the Unisys ICON and a new Amiga had chosen QNX as the base for their OS.

  39. W.S.Gosset

    OS/2 from the Big App developers' POV

    WordPerfect's then-CEO wrote this memoire, eventually put it on web for free. It's actually a good read, especially if you've never experienced the insane self-meme-absorption that is american corporate insiders. Which is why it went from biggest in world to *pop* so suddenly.

    OS/2 first mentioned near end of Ch.8 and then passim. It shows quite a different view of OS/2.

  40. Korev Silver badge

    Productive OS/2?

    We had a productive OS/2 machine at one of our sites up until very recently. I think the only reason it was got rid of was the site being axed.

    It was running a protein separation machine and had to dual into Win98 if you wanted to copy your data to the NAS. It was impressive that it lasted so long baring in mind it lived in a cold room at <5c.

  41. ImpureScience

    Still Sort Of Miss It

    I really liked OS/2, and for a while I thought it would take the place that Windows now owns. But IBM had no idea how to sell to single end users, and get developers on board. Despite having a superior product their financial policies guaranteed only customers ended up being banks and insurance companies.

    I'm a musician, and I remember going on for over a year with the guy they had put in charge of MIDI on OS/2. It never happened, because which bank, or what insurance company, would be interested?

  42. Anonymous Coward
    Anonymous Coward


    No mention of Ambra, a very spiffy IBM compatible that competed with itself

  43. Anonymous Coward
    Anonymous Coward

    Model M

    Peter Gathercole and Model M fans everywhere,

    I'm typing this on my IBM Model M keyboard, dated 9-30-92 per the label on the bottom. I've moved it from one machine to another over the years and it has happily kept on clicking away. Like a lot of old IBM gear it was built (overbuilt?) like the proverbial tank - DisplayWriters anyone? I'm convinced it will outlive me.


    1. aqk

      Re: Model M

      Yep. Indeed, chappies! IBM made really good keyboards! And typewriters too!

      I loved their sturdy 029 keypunch machines. The holes in those 80-column cards were crisp and clean! IBM sure made great stuff!

      Their disks? Not so good as the competition. But when an IBM salesman took your boss on a 3 day junket, and you were then faced with maintaining your operating system(s) and data on a dozen or so shiny new 3340 disk-drives, well all I can offer is that it was a least some sort of guaranteed job security. But not necessarily for my boss.

  44. W.S.Gosset

    Model M rebuilt, USB

    Chaps and Chapesses everywhere,

    The Model M was and remains and IS the best keyboard ever made.

    The second best by a very narrow margin was the old Apple Extended Keyboard II.

    . I've used both and it is a very narrow margin, almost neck & neck

    Well, the latter is available, brand new, and USB, from a German crew who decided to bring it back.

    Essentially, they got the rights from Apple to re-build it. And they're fussy. They did it. Exactly.

    I've got one. And it's awesome. Flashback-city. The good ol' days, Part II.


  45. SME Integrator

    The point

    From someone who used to install this stuff 30 years ago.......the point of this story is that just because a vendor rules the current market (like Novell and IBM did,) it's difficult to see past that and realise that Microsoft is slipping along with, believe it or not Cisco.

    Watch this space, change is the only constant

  46. aqk

    TOKEN RING! It's WAAAY better than Ethernet!

    And faster too! Ethernet only has a 10 mbs speed, but token ring will soon offer 16 mbs!

    And our routers are something to behold! Listen carefully- you may actually hear the clicks of their relays switches!

    Yes I remember the dorky IBM salesman trying to sell us this crap 30 years ago...

    We laughed at it, and stayed with our ethernet.

    Then 5 years later we were outsourced, and I suddenly, beside our own ethernet system, I had to maintain a horrible token-ring network used by a major international company (much larger than ours) whose IT dept had also been outsourced... Oh, the humanity!....

  47. wizdude

    eComStation by Serenity Systems and Arca Noae

    the king is dead. long live the king.

    everyone seems to be talking about OS/2 as though it is dead. it certainly isn't and many people still use it today.

    Serenity Systems created eComStation which runs on modern hardware and was originally based on OS/2 Warp Version 4. It's still in full production today.

    and then we also have Arca Noae which is due for release soon.

    Warpstock still runs events annually

    many people still run this platform and some developers still release applications for it as well. there is a massive archive on Hobbes and also OS/2Site

    software such as Weasel which is a fully featured mail server is still being actively developed as of right now :-)

    cheers, wizdude

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like