Operating System/2 announced 25 years ago

On April 2nd of 1987 (not April 1st, that wouldn’t do!), IBM and Microsoft jointly announced Operating System/2, the long-awaited protected-mode version of DOS.

However, OS/2 was not the only product announced on that day. OS/2 was merely one part (albeit very important one) of a massive product rollout blitz on both IBM and Microsoft’s part, purportedly the beginning of a new era of personal computing.

Perhaps coincidentally, this is also the 20th anniversary of the release of OS/2 2.0, the first mass-market 32-bit operating system for PCs. Not to the day (the official release date was March 31st, 1992; still the first quarter of 1992) but close enough to be significant.

In April 1987, IBM most importantly rolled out the Personal System/2 line of computers, including (finally) a 386-based Model 80. High-density 1.44MB 3½” media were introduced, along with PC DOS 3.30 update to support the new hardware; VGA and the 8514/A high-resolution graphics accelerator were announced; and a slew of updated networking hardware and software was rolled out. Among the more interesting IBM announcements was AIX for PS/2, IBM’s BSD-derived UNIX variant originally used on the RT PC. OS/2 was one of the biggest items, and the announcement included Presentation Manager and OS/2 Extended Edition with database and communication capabilities.

On Microsoft’s side the announcements were no less important. OS/2 (including the Presentation Manager) was again the big ticket item, but there was also MS-DOS 3.3, Windows 2.0, XENIX System V/386, and LAN Manager.

The announcements were interesting in that many of them concerned products still a year or more from completion. IBM’s PS/2 systems, DOS 3.30, and networking hardware and software were mostly available immediately or within weeks. For AIX on the other hand IBM merely announced that an announcement regarding future availability would follow in Q4’87. IBM OS/2 1.0 SE was announced to be available in Q1’88, and OS/2 1.1 with Presentation Manager towards the end of 1988.

On Microsoft’s part the announcements were almost exclusively vaporware. Only 3½” media support (existing products shipped on the new format) and PS/2 mouse drivers were scheduled to arrive within weeks.

XENIX for the 386 was supposed to be available for licensing by OEMs in Q2’87, but that turned out to be overly optimistic. At any rate, this was probably Microsoft’s last major XENIX-related announcement. UNIX in general turned into an enemy, even though Microsoft owned a significant chunk of SCO.

MS-DOS 3.3 was expected to be shipped by OEMs sometime in 1987, with the timing obviously not entirely under Microsoft’s control. Windows 2.0 was scheduled for 3Q’87 which did in fact happen with Windows/386 for COMPAQ.

Microsoft expected OS/2 1.0 to be available to OEMs sometime before the end of 1987 (which did happen, barely) and OS/2 1.1 with Presentation Manager sometime in 1988 (again barely happened).  For LAN Manager, Microsoft didn’t even provide any estimates; the product was clearly very far from completion.

The most fleshed out product on Microsoft’s side was the OS/2 SDK, planned to ship on August 1st, 1987; the kit would set back prospective OS/2 developers by $3,000. In the product Q&A, to the question “Why is the SDK so expensive?” Microsoft provided the humorous answer “The SDK is not expensive” and went on to explain that their costs were high and really, developers should be dying to develop for the next generation OS. We know how that turned out.

It’s interesting to consider how complete the products were at the time of the April 1987 announcements. The OS/2 “kernel” (OS/2 without Presentation Manager) was likely quite usable, judging from the first OS/2 SDK (completed between April and June 1987). The Presentation Manager on the other hand was probably very incomplete (it’s doubtful Microsoft had anything working at the time), although the specification was probably complete. LAN Manager was most likely merely a glint in Microsoft’s eye, and even the specification was far from being completed.

The big OS/2 announcement was the first not-yet-tangible product of the IBM-Microsoft Joint Development Agreement (JDA). It was meant to usher a new era of PC computing, but in retrospect it was more like the end of the IBM PC (with emphasis on IBM). The PS/2 systems, while technically excellent, turned out to be a dead end. Rather than helping IBM regain control of the PC market, they led to IBM’s irrelevance. For IBM’s PC business, it was the beginning of the end, for Microsoft it was merely the end of the beginning.

Five Years Later

Almost exactly five years after OS/2 had been first announced, IBM shipped OS/2 2.0. A 386 version of OS/2 was anticipated in the April 1987 announcements, but the project was seriously delayed. Microsoft shipped the first 32-bit OS/2 SDK at the very end of 1989. IBM picked up the pieces in 1990, but that delayed the release by at least a year.

The PC industry landscape in April 1992 was very, very different from April 1987. The 800-pound gorilla was now Microsoft, not least thanks to the huge success of Windows 3.0. IBM was still a major OEM, but no longer a driving force of the industry. The PS/2 line, though not unsuccessful, ended up isolating IBM instead of helping it regain its leading role.

Microsoft had given up on OS/2, and by the time OS/2 2.0 was released, more or less pretended that it had never been involved. Microsoft may have been impatient (later, it took NT quite a long time to become a major force; and the situation was hardly different with 16-bit Windows), but the relationship with IBM was poisoned and probably couldn’t work.

It’s clear that Microsoft wanted to be in the driving seat. The JDA was merely a hedge, an insurance against some other company being anointed by IBM. To make matters worse, there was a major culture clash between Microsoft and IBM. While IBM was a very structured, process driven company, like most corporations that have been around for a few decades, Microsoft had much more of a startup mentality where only results matter and the process be damned (including things like writing formal specifications and test cases). By 1990, Microsoft viewed IBM as a major drag and decided to go it alone, with fairly spectacular results.

In retrospect, it’s difficult to say whether the future had already been decided in April 1992. Perhaps OS/2 would have had a chance had Microsoft not engaged in numerous underhanded anti-competitive practice. Then again, perhaps IBM never stood a chance.

25 years later, OS/2 is more or less completely gone while Windows remains the number one PC operating system with no serious competition. Ironically, the Microsoft of 2012 is probably much closer to IBM of 1987 than it is to its former self. The future of Windows also doesn’t look nearly as bright as it used to, and in the so-called post-PC era Windows may well go the way of horse and buggy. Twenty-five years from now, we shall be much wiser.

This entry was posted in IBM, Microsoft, OS/2, PC history. Bookmark the permalink.

27 Responses to Operating System/2 announced 25 years ago

  1. Yuhong Bao says:

    “(later, it took NT quite a long time to become a major force; and the situation was hardly different with 16-bit Windows)”
    Don’t forget Chicago and the delays.

  2. Yuhong Bao says:

    And the DR-DOS-related lawsuits.

  3. wow 20 years of OS/2 2.0 … I still remember buying a copy for university so I could be ‘legal’ on one of the 486’s that I used for college work.. they seemed ok with bring your own OS if you donated it…. 😐

    Although some time later they let me have it back, as they didn’t like the idea of OS/2 contaminating some machine. LOL

  4. michaln says:

    Sorry, you misunderstood. Chicago was delayed in development, but it was a smashing success immediately upon release. Very different story. In contrast, Windows 1.0 was over a year late and more or less a flop. It took about five years (until version 3.0) before Windows started selling well. The NT story was similar; about 2 years late in development and then it took another 3-4 years (until version 4.0) before the product became a hit. (With NT it of course took much longer until it became the primary OS).

  5. Yuhong Bao says:

    Still, it is horrible that it took 10 years after Intel released the 386 before 32-bit programming to become popular.

  6. michaln says:

    That I can agree with.

  7. michaln says:

    Yep, seen it. Pretty decent.

  8. Paul Jankura says:

    Do we have any idea why there wasn’t interest in developing a protected mode successor to DOS before 1987? I mean, when IBM was developing the PC AT, did they see it as a DOS machine or was the intention that a successor OS would be developed? Wouldn’t it have made more sense for them to think about a protected mode OS strategy in 1983 when they were building the AT rather than 1986-1987? Or was the problem that Microsoft was responsible for OS development and MS was devoted to DOS, and their real mode multitasking DOS until sometime in 1986?

  9. michaln says:

    There are two somewhat separate issues… protected-mode OSes for the 286 and protected-mode versions of (or successors to) DOS. There was XENIX from MS, there was (supposedly) PC/IX from IBM, there were 3rd party UNIX variants and clones (QNX). None of those made much of a dent, probably to a large extent because a system capable of running UNIX decently was just too expensive back then. I don’t think IBM developed the PC/AT as a DOS machine, but I also don’t think they had any idea how hard DOS would be to dislodge. Based on their past experience of how hard it had been to displace CP/M, they would have had every reason to think that replacing DOS would be just as easy. Not that different from what happened to Intel with the Itanium, really.

    The problem with a protected-mode successor to DOS was that the 286 was awful at mixing a protected-mode OS with DOS. Intel didn’t realize anyone would want to do such a crazy thing back when they designed the chip. OS/2 went through a lot of hoops to make real and protected mode coexist, and people still talked about a penalty box or DOS coffin, and it was for a reason. I think it just took a long time to develop a workable solution to mix DOS and 286 protected mode at all. All those bimodal device drivers and segments accessible from both real and protected mode were a kind of black magic. Some people at Microsoft thought the 286 wasn’t worth targeting and they should have gone straight for the 386; the problem of course was that in 1987, everyone talked about 386s but hardly anybody had one.

    The JDA was signed in mid-1985, so the interest clearly was there at that point. I don’t think when they signed the deal they thought it’d be pretty much 1988 before anything shipped at all, and even then only half the product. It’s true that Microsoft was also working on the real-mode multitasking DOS 4 at the same time. And on Windows. (And on XENIX.) All those projects intersected in interesting ways, with the NE module format being one clear example. I don’t know how many people worked on each project at Microsoft, but I suspect that they had a lot more people working on DOS 5 (OS/2) than on DOS 4 or DOS 3 back in 1986-1987. If you look at the differences between DOS in 1985 (3.1) and 1987 (3.3), it was really not a lot.

    I believe 286 XENIX could run most executables written for the 8086 versions in protected mode, but in comparison to DOS, it was a highly structured and advanced OS. DOS (really 86-DOS) was written with backwards compatibility with CP/M in mind, rather than forward compatibility with protected-mode processors… and it showed. A 386 was needed to make the transition from DOS to a protected-mode OS. The technology was there in 1987, but the market wasn’t.

  10. Yuhong Bao says:

    “I believe 286 XENIX could run most executables written for the 8086 versions in protected mode”
    And Windows 3.0 could too if the program was properly written.
    “the problem of course was that in 1987, everyone talked about 386s but hardly anybody had one.”
    Yea, only in 1988 did Intel finally release the 386SX in an attempt to make 386s affordable. Then MS began developing what became OS/2 2.0. Unfortunately the success of Windows 3.0 led Microsoft to screw up yet again and by the time Win95 was finally released, the (by then 10 years old) 386 was close to being obsolete.

  11. michaln says:

    How is Windows 3.0 relevant to 1987??

  12. Yuhong Bao says:

    “Still, it is horrible that it took 10 years after Intel released the 386 before 32-bit programming to become popular.”
    And Chicago was delayed so badly it was released after OS/2 *Warp 3* (which was much better than OS/2 1.x), which is why I mentioned it. And it is during that time that MS was playing dirty against it.

  13. Yuhong Bao says:

    “Perhaps OS/2 would have had a chance had Microsoft not engaged in numerous underhanded anti-competitive practice. Then again, perhaps IBM never stood a chance.”
    Maybe, but my point is that even the JDA with it’s problems would be better than this alternative.

  14. Rugxulo says:

    Here’s an old post titled “What’s happening to OS/2” by Gordon Letwin from 1995 that explains a lot. (Though I realize that he’s biased in favor of Win95 and MS since he worked there. Nevertheless, he was an important part of all this.)

    http://groups.google.com/group/comp.os.ms-windows.misc/msg/d710490b745d5e5e?&hl=en

    P.S. What good is a 32-bit OS if nobody has the RAM for it? NT used too much, and even Win95 would only very barely run on a 386. I think that alone is why DOS support lasted so long. Though I do think it’s ridiculous that common OSes nowadays like Vista and Win7 need at least 1 GB to run well. You can dislike DOS all you like, but at least it runs on pretty much anything. (And DOS extenders overcame any memory barriers long before Windows, though 3.0’s introduction of DPMI was quite successful too.)

  15. michaln says:

    The funny thing is that some people were saying back in 1987-1988 that OS/2 should have been 32-bit from the beginning. Which tells you that there is no way to please everybody…

    DOS is crap as a server, that’s the problem. A multitasking OS which doesn’t keel over with the slightest breeze is needed to get something decent out of a server, and it doesn’t matter much whether that OS is OS/2, NT, NetWare, UNIX, or something else. People who need performance and stability can afford the extra memory.

    Also, saying that DOS can run 32-bit stuff but a 32-bit OS needs too much memory is a bit self-contradictory. Back in the early 1990s, a 32-bit game like DOOM already required 4MB, the same minimum as many 32-bit operating systems. Many, many 32-bit apps had similar requirements (AutoCAD, DTP, whatever). 32-bit was simply used for tasks that needed a lot of memory and which 16-bit code could not handle at all or only with difficulty. Clearly not everybody needed that, but many folks did.

    And yes, Windows 95 barely ran on a 386, but then again in 1995 a 386 barely ran anything, period. I know because I upgraded from a 25MHz 386 to a 90MHz Pentium in 1995 🙂

  16. Yuhong Bao says:

    “32-bit was simply used for tasks that needed a lot of memory and which 16-bit code could not handle at all or only with difficulty. Clearly not everybody needed that, but many folks did.”
    And with a full 32-bit OS would also take some of the 4MB, at least paging large applications to disk would have been possible.

  17. Yuhong Bao says:

    “(Though I realize that he’s biased in favor of Win95 and MS since he worked there. Nevertheless, he was an important part of all this.)”
    I do agree that the JDA was not very good. My point is that the alternative MS took was worse. By the time the worst of MS’s attacks to OS/2 happened such as “Microsoft Munchkins”, I think Letwin already left MS.

  18. Yuhong Bao says:

    And as another note, Win95 and OS/2 2.0 and later had an advanced shell based on objects (Explorer or Workplace Shell) which took a lot of memory. Early MS OS/2 2.0 SDKs had the old OS/2 1.2 shell which took a lot less RAM.

  19. michaln says:

    Or as another example, take Windows NT… even with an outdated shell, it managed to gobble up memory like there was no tomorrow 😉

  20. Yuhong Bao says:

    “the problem of course was that in 1987, everyone talked about 386s but hardly anybody had one.”
    And remember OS/2 required new apps, while Windows/386 did not have that problem. Why do you think MS created the Family API?

  21. Yuhong Bao says:

    “but the relationship with IBM was poisoned and probably couldn’t work.”
    Probably couldn’t work? What makes you think that?

  22. michaln says:

    Hindsight. We know the relationship didn’t work, and there were many reasons for it.

  23. Yuhong Bao says:

    I know it didn’t work well, but my point was that the alternative MS took was worse.

  24. Yuhong Bao says:

    As mentioned in the article later.

  25. michaln says:

    It was perhaps worse – for everyone except Microsoft…

  26. Yuhong Bao says:

    Funny that Joel who happens to wrote this is a former MS employee:
    http://www.joelonsoftware.com/articles/fog0000000069.html

Leave a Reply

Your email address will not be published. Required fields are marked *