This material is protected under international copyright. If you wish to reprint this article in part or in full, please contact firstname.lastname@example.org for permission. You may also visit http://www.stardock.com to see this article and related articles.
I work for an OS/2 software company called Stardock Systems, Inc. We develop and publish 32bit OS/2 software for both corporate OS/2 sites as well as individual OS/2 users. Our products range from complete desktop environments for securing and standardizing corporate desktops (Object Desktop Professional), to utilities to allow corporations and "power users" to recover from system hangs (Process Commander) to consumer entertainment products such as Galactic Civilizations, Trials of Battle, and most recently Links OS/2.
Of course, it's easy to say that now. Who would have thought that back then. Windows 3.0 was barely being preloaded and it was too unstable as a corporate client to worry too much about.
But it was largely the workplace shell that delayed OS/2 and in hindsight it would have been better to release it for an OS/2 version 3 in 1993 and had kept the cruddy old Program manager type setup that was in OS/2 1.3. Because had IBM done that, OS/2 would have looked and felt like Windows 3.0 did except it would multitask, not crash, and run true 32bit software, and run existing Windows programs and it would likely have been out in 1991, nearly a year before Windows 3.1 was out.
Picture that. OS/2 2.0 could have come out as a 32bit, multithreaded OS that ran DOS (better than DOS), Windows and new OS/2 software. Was stable, worked great on networks, and had good performance. The only competition was the buggy Windows 3.0 which ran on top of DOS. The only thing different in this scenario than the OS/2 2.0 we actually had is this didn't have the WPS and came out 9 months earlier – before Windows 3.1 vapor could cloud the scene.
But as history wrote, OS/2 2.0 was theoretically released on March 31, 1992 (though most OS/2 buffs know that it really wasn't generally available until June).
By then, Windows 3.1 was out and was fairly fast and much more stable. Still not as stable as OS/2 was but any early user of OS/2 2.0 could tell you that the workplace shell of OS/2 2.0 was pretty flaky. Black icons, trap errors (i.e. the register dump kind), and of course everyone was using VGA and no sound whereas Windows 3.1 looked pretty, had SVGA support, and had some sound at least.
In the Fall of 1992, IBM made available the Service Pack for OS/2 2.0 and for some, a beta of Windows 3.1 (WinOS2 3.1) which made life much better for OS/2 2.0 users. By June of 1993, OS/2 2.1 was released which as its point upgrade name would imply, largely addressed the problems of OS/2 2.0.
It was with the release of OS/2 2.1 and the availability of MMPM/2 (the multimedia stuff for OS/2) that I began writing OS/2's first major commercial game (I say major because there were commercial black jack programs and such that came before Galactic Civilizations). The amount of attention that Galactic Civilizations received as a beta surprised most people including IBM. Suddenly, lights went on at IBM and the belief grew that OS/2 could be made into a general consumer platform.
In hindsight, this was probably a mistake since IBM wasn't prepared to do what it really took to be a good consumer product not to mention the consumer market isn't nearly as profitable per capita as the corporate product. A typical example of this is the Usenet Newsgroups. Lots of end user OS/2 users hang out there but most OS/2 ISVs make their money from corporate site licensing and service agreements. If every OS/2 user in the Usenet suddenly changed OS's tomorrow, it probably wouldn't affect OS/2 ISVs noticably (except most of their tech support calls would disappear). That's not to say that end users on Usenet aren't important, it's just that end users, as a general rule of thumb don't generate as much profit as a single IS manager who buys $50,000 in units and might make a couple tech support calls per quarter. Obviously end users matter to shareware authors and companies that produce entertainment software (nearly 15% of Stardock's revenue comes from its entertainment division). But back in 1993 or so, OS/2 as a mainstream consumer platform looked like a good idea. I sure thought it was but who knew how difficult it would be for IBM to try to create the infrastructure necessary to deal with massive numbers of end users.
But IBM dove into the consumer market head first and the result was OS/2 Warp 3. As with every release, IBM made a deadline and stuck to it regardless of the consequences. IBM's not the only one's to run into that sort of problem – giving into public pressure to meet a specific date and then releasing something that they thought was ready but probably needed just a couple more weeks of testing. The lack of Winbios support, and a config.bak problem tarnished an otherwise wonderful product.
Warp 3 went out and get nailed because of installation problems. Coincidentally, of course, in October of 1994 (when Warp 3 was officially released) Microsoft released the big "Chicago" beta which became Windows 95 later on. The press, which seems to generally prefer to talk about vapor than substance essentially reprinted Microsoft's Chicago reviewer's guide (while OS/2 users complained that the guide was in Word 6 format which had just come out too). OS/2 Warp 3 ended up being compared to the unreleased Chicago (i.e. what Chicago would be when it came out which changed from day to day).
By embracing the end user market, IBM created the biggest technical support nightmare that the IBM company may have ever seen. The #1 tech support report for OS/2 Warp 3 was not how to get TCP/IP stacks going or how to link Novell up with OS/2 clients, or how to make sure DB/2 would work on the new version. No, it was "How do you get DOOM 2 to run on OS/2 with sound?" IBM has a highly paid, highly trained technical support staff that was meant to deal with Fortune 500 companies who had paid millions of dollars for software and hardware from IBM. They weren't prepared to have to deal with a bunch of people trying to run video games and the support costs from this really hurt PSP at the time. This probably has a lot to do with why PSP today goes out of its way to discourage "kitchentop" users because they don't want to support every new user that wants to play some video game on OS/2.
So IBM ends up having tens of thousands of phone calls pouring in over trivial issues while the product sells at the local CompUSA for $89.95 with a good $30 of it going to third parties (mostly Microsoft) in royalties and another $10 in manufacturing costs. With Warp 3, IBM did the good fight and basically went at it by not trying to make a profit on it but to get market share. Critics of IBM always say that IBM didn't try hard enough but with OS/2 Warp 3. They did and they were having good success for awhile. IBM's marketing program was ineptly done to quite an extent (and indeed in a Winter 1996 meeting with IBM's ad agency I questioned them about what the heck they were thinking in those terrible TV ads). But OS/2 was indeed taking off and certainly making Microsoft nervous. In the Winter of 1995, Bill Gates is said to have remarked in frustration, "How can we compete with something that seems to have unlimited funds thrown at it?" (This was reprinted in the trade magazines from the time). It was IBM's next move though that got PSP and OS/2 into serious trouble: OS/2 for the PowerPC.
Around this time IBM was getting pretty giddy, while Windows 95 wasn't out yet, IBM was spending huge amounts of money on marketing and courting partners. An IBM business partner can gauge how important they are to IBM by the number of duffle bags they get in a given year. 1994/1995 was a big duffle bag year if you were an OS/2 ISV or partner. It really looked, despite a few bumps in its initial release, that OS/2 Warp could capture a good 15% to 20% of the general OS market. IBM was courting hardware vendors as well as software vendors to work with IBM. We'll talk about how many of these "partners" blatantly took advantage of IBM's good faith later.
The PowerPC version was also born of this vision of seeing OS/2 finally taking off and trying to expand on that success. Confident that any technical issue could be surmounted, they believed that Workplace OS (as Power OS/2 was called in those days) would run several OS's at the same time. Many people, even at the time, thought IBM might be jumping the gun a bit. OS/2's success on Intel wasn't exactly cemented yet even though things were looking up. Nevertheless, IBM worked on the PowerPC version of OS/2 for a couple of years with high amounts of resources really starting to pour in right after the Warp launch so that Fall Comdex 1994 you could already see OS/2 apps such as Desktop Observatory and Sundial's Relish running on the PowerPC version of OS/2 (as long as you didn't move the mouse). Unfortunately, popular rumor doesn't make fact and the rumor – that still exists today that the x86 chipset is at the end of its design limits was and continues to be simply not true (any computer engineer can tell you that any chipset can essentially be extended forever if you have enough money which Intel does). Someone with power at IBM convinced the powers that be that Intel couldn't really boost the speed of the x86 process line much more than where the 486 was already. The mediocre performance of the Pentium 60 made the case stronger. So IBM believed that they could create a processor for PC's that was several times faster than the Intel chips and do so very cheaply.
When this failed to happen, not only did he PowerPC not take off as a new platform (other than Macs which weren't competing with 680x0 chips at this point), it made OS/2 for the PowerPC useless. Years of work and energy had poured and subsequently wasted on OS/2 for the PowePC. The final version of it (which does exist) doesn't even have networking. This sort of thing tends to really take the wind out of one's sails.
Not only did it take the wind out of many in PSP, but it angered the other parts of IBM which lost their faith in PSP to deliver product in a timely fashion. It is my belief that the failure of OS/2 for the PowerPC to be delivered on time helped spell the doom of PSP getting funding from Lou Gerstner. In his mind, I believe, PSP had blown their chance. PSP got isolated from the rest of IBM to a degree because of this (in my opinion of course).
Even worse, projects are usually comprised of a handful of truly critical people. For example, OS/2 SMP was largely done by a single person who later left which is (so I hear) one of the reasons why it took so long for OS/2 SMP to get updated (this is rumor keep in mind). So imagine yourself in their shoes, you just spent 1 maybe 2 years working on the biggest greatest project of all time to have it be for nothing. This was a scenario that would repeat itself later with OpenDoc. What ends up happening is that many truly talented people end up leaving. IBM lost a lot of key developers because of this kind of frustration. The mega-team that delivered miracles in the form of OS/2 2.0 to OS/2 3.0 began to dwindle.
OS/2 for the PowerPC's doom was half PSP's fault for not getting it out on time and half the PowerPC chip's group's fault for not delivering on the promise of next generation performance.
When Windows 95 came out, it came up against OS/2 3.0 that was pretty much unchanged. Much of the developer resources at IBM that could have been putting new features into OS/2 had been working on PowerPC for OS/2. Instead of using money to get more third party support on OS/2 Intel, monetary resources for third parties were spent encouraging ISVs to write to the PowerPC, trade shows, etc.
Again, all this is in hindsight and at the time, most of the decisions made some sense. But when Windows 95 came out, contrary to popular belief in the OS/2 camp, it was not a bug-ridden, unstable piece of junk. It did the job and did it decently well for lots of people. Not nearly as good as OS/2 but Microsoft, the master of putting the carrot in front of the mule said "Win95 not stable enough? Not enough multitasking? No problem, just wait for NT 4.0 which will have that and run Win95 software…"
The answer from IBM was OS/2 Warp 4 which we'll talk about in part 2: