Master Builders of OS/2

The MS OS/2 videos exhibit has now been completed with the addition of two PDF documents. These are scans of two fat three-ring binders that were handed out to attendees of the Microsoft OS/2 Developer’s Conference in New York City on July 7-9, 1987.

These are 700 pages containing hardcopies of conference slides. More or less all the slides are shown in the videos, but obviously print has a bit better resolution than VHS tape. This should make it easier to follow the videotaped presentations.

Note that the binder covers included the phrase “Master Builders”, which does not appear to be repeated anywhere else in the text.

The hardcopies also included plenty of space for conference attendees to take notes. All extant handwritten notes were transcribed, either on the same page (if enough room was left) or a blank page was inserted.

Transcribing the notes was not trivial, since they were more scrawled than handwritten, and often did not consist of complete sentences. Luckily the context was usually clear, and only in one or two places the transcript is a “best guess”. For the most part, the transcripts should be accurate, even if the original notes might not be (largely as a result of having been written in a hurry).

The handwritten notes enhance what is already an interesting historical artefact, now more than quarter of a century old. At the time the conference was held, Microsoft had sold 1,800 OS/2 SDK copies, which at the time consisted only of a pre-release OS/2 build, beta tools, and the Windows 1.04 SDK. The Presentation Manager specification wasn’t ready yet, the Windows 2.0 SDK was’t available either. Microsoft OS/2 was the future and Windows was just a side show. Good times!

This entry was posted in Development, Microsoft, OS/2. Bookmark the permalink.

11 Responses to Master Builders of OS/2

  1. what an incredible wealth of information…! I guess this is part of how they justified the $3,000 SDK prices.. (or was it $2,600? or was that 2.0?).

    Such a shame IBM didn’t let MS port Windows on top of OS/2, I almost venture that Windows was almost written in mind with something like OS/2 …

    Thanks to Steve Wood’s original memory-allocation design, many of the changes involved bypassing real-mode code that served only to emulate the protected mode of the 286. ..

    But then OS/2 would have been more like NT.. The lack of OS/2 apps probably wouldn’t matter as much if you could recompile for OS/2+Windows .. But then again just ask anyone with NT MIPS (or WindowsRT) just how many devs are willing to hit the recompiler.

  2. michaln says:

    It was $3,000 for the original OS/2 SDK, and $2,600 for the OS/2 2.0 SDK. And yes, after seeing just how much was included in the OS/2 SDK (the OS, complete documentation, Microsoft C + MASM, development tools, etc. etc. — probably 100 pounds of stuff!), I don’t think the price was unfair. It was really all the software needed to develop for OS/2 and the developer literally only needed to supply a clean PC with a totally empty disk.

    The real problem was that it was simply too much, and not everyone needed all of it. It wasn’t until early 1989 or so that Microsoft repackaged everything into bite-sized chunks.

  3. michaln says:

    Actually… if you watch the videos, Microsoft did port Windows on top of OS/2, and the result was called Windows Presentation Manager. The reason why the API had been changed was that the developers weren’t happy with some of the choices made in Windows. It’s anyone’s guess whether the API divergence was Microsoft’s inertia/foolishness or a secret plan (I’d assume the former). The incompatible API arguably hurt both OS/2 and Windows. Microsoft managed things far, far better with Windows 3.1/NT/Win32s/Win95.

    What was needed in the late 1980s was “PM Lite” (Presentation Manager running on top of DOS, with a DOS extender), which IBM was about to do but reportedly Gates + Ballmer talked them out of it because they were (rightly?) afraid it’d destroy Windows.

  4. lessons were certainly learned in the whole debacle. Or so I thought Windows 8 is basically ignoring everything that was learned from this & the NT 3.1 era.

    Like it or Loathe it, 80386 code will be with us for a long long time ..

  5. michaln says:

    Surely most Microsofties nowadays either don’t remember or never even knew what went down around 1990, but at least Steve Ballmer ought to remember as he was right in the middle of it…

    I agree that x86/AMD64 will be around for a long time, but at the same time its dominance looks shakier now than perhaps at any point in the last 25 years or so. ARM in particular is getting bigger all the time.

  6. 32bit arm is everywhere, but the possibility of 64bit arm is far more exciting.. And fitting that a darkhorse processor is really going to wind up changing the world.

    I’ve heard too many rumors about apple shifting to an all arm lineup.. if they can get something like rosetta again it’ll be amazing..

  7. michaln says:

    I don’t think Apple is stupid enough to do a full switch to ARM, unless they want to completely kill Macs (which I wouldn’t entirely discount!). An ARM-(under)powered Mac Pro to run your Aperture or Final Cut… yeah, that sounds appealing. All those folks running VMware Fusion or Vagrant on their MacBook Pros would no doubt love it too.

    But an ARM-based MacBook Air, that could be interesting. Apple has long experience with supporting multiple CPU architectures in their OS.

  8. Richard Wells says:

    I was impressed by the 1,800 sales of OS/2 SDK. Only 9 million copies of DOS had been sold which means there was one OS/2 developer for every 5,000 DOS users. Someone in the IBM corporate sales department was amazing.

    I doubt a PM-Lite for DOS would have been successful. It is difficult to convince developers to move code to a more limited subset of functionality. PM-Lite would require programmers to remove functionality from their existing PM programs in order to run on the less capable system. Family API and Win32s had issues getting much developer interest and those had been defined before the bigger versions were released. I don’t know if the improvements to PM-16 were worth the loss of 5 years of hard won Windows GUI knowledge. I think IBM’s strategy would have put MS out of business so the split between the two companies was inevitable.

    Windows 8 is the same development technique used with Win95. Run almost all your existing applications plus new ones still written for older operating systems while getting used to the new interface style. Whether long-term Win 8 has similar impact to Win95 will be answered in a few years but a lot of complaints about Win 8 are similar to articles that projected failure for Win 95 because other systems got it right and MS shouldn’t force a new interface.

  9. michaln says:

    Just to clarify… The OS/2 SDK was a purely Microsoft enterprise; no IBM involvement whatsoever was apparent. It would be interesting to know how many copies were bought by companies like Lotus or Borland, but clearly 1,800 copies sold didn’t translate into 1,800 new apps. Incidentally, in 1988 Microsoft claimed to have 4,000 OS/2 SDK users.

    A PM Lite would have been meaningful in 1988-1990, when OS/2’s hardware requirements were considered too extreme; certainly not in 1994 or so. Just like Windows 95 was meaningful in 1995 but not in 2001, and Win32s was meaningful in 1993 but not in 1996.

    The improvements to PM did not cost 5 years of hard-won Windows GUI knowledge – not even close. When PM was defined, Windows had been out for less than two years. Even when PM was released in late 1988, most PC developers had zero Windows knowledge. At the same time, Windows and PM were architecturally so similar that Microsoft shipped the Windows 1.x/2.0 SDKs as part of the OS/2 SDK as training material before there was any PM code.

    And if Windows 8 is like Windows 95, what’s the Windows NT? 🙂 I would also disagree that Windows 95 was all about a new interface. It was primarily a new (for some definition of “new”) 32-bit OS technology, long filenames, Plug an Play, etc. etc., with the interface being fairly secondary.

  10. Richard Wells says:

    SDKs didn’t go to consumer software vendors as much as to internal IT development groups at big companies. I had worked at number of firms that had OS/2 SDKs in back room storage. All were IBM mainframe shops and there were remnants of abortive 3270 to OS/2 conversion efforts. This was not territory that mid-80s Microsoft had any sales presence; any sales of SDKs had to be generated by IBMers declaring OS/2 to be the future. Fulfillment of orders was done by MS.

    PM lite needed to be defined in 1987 before PM itself was released. IBM’s efforts for PM-Lite started after the first batch of PM programs were written and were using advanced OS/2 features. I doubt anyone would enjoy stripping out threads from their code just to run on a hoped for DOS version. Of course, IBM changed direction on the merits of DOS several times as internal battles between software sales volume versus the benefits of maintaining an all-IBM hardware stack were fought. IBM memos from that time must be very interesting reads.

    As a general rule, subset of systems don’t get developer interest. MS has tried several times and mostly failed. Other companies tried it as well with similar lack of success. I really doubt that IBM could have pulled it off with PM for DOS, especially by 1990 in order to meet the 10 million sales per month result Windows 3 obtained.

    Five years includes prerelease Windows where outside developers could point to code that just plain didn’t work. It took Hursley lab a long time to stabilize and speed up PM-16 to a state worth developing for. I think getting Excel and Pagemaker running on top of Windows 2+ on OS/2 in 1988 would have done more for OS/2 sales than having an arguably better PM with very few applications for several more years. I call it Windows 2+ to suggest that extra functions would have been added to the Windows base to support at least some underlying OS/2 functionality.

    If you extend the analogy, Windows NT would have equated to Windows Phone 7 where many new features were tried out but no one cared because no one bought the product. Really most of the new features in Windows 95 mattered in 1996 or 1997 but at the time of release it was old software and old hardware that worked alongside a new UI.

  11. michaln says:

    If you have some numbers on who the OS/2 SDK buyers were, it’d be great if you could share… I have to assume that companies like Lotus or Borland must have bought a few, and there were many smaller companies which also developed OS/2 software. I don’t know how many they bought; I do know that all the Microsoft training and demo materials were targeted at PC ISVs, not all-IBM mainframe shops.

    Windows 95 (and Win32s) is a counterargument to “subset of systems don’t get developer interest”. Windows 9x was clearly a stopgap measure to get Win32 apps developed at a time when Windows NT could not be unleashed on the general public. I’d say the strategy worked out quite well.

    I do get the argument that moving more or less unmodified Windows on top of OS/2 might have been a better strategy. I question how suitable it was for running on top of a protected-mode operating system. If substantial changes were already required, it might have made sense to bite the bullet and change the API some more. That admittedly doesn’t answer the question whether it was a good strategic move to replace GDI with GPI and make the GUI considerably more complex (if cleaner).

Leave a Reply

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