Twenty-five years ago this month, on April 2, IBM announced its new PS/2 computers and a new multitasking operating system to run on (most of) them–OS/2. They even lured a bunch of the actors from M*A*S*H to do an ad campaign for them.

It didn’t seem like it at the time, but that was the beginning of the end of IBM’s PC business.

The PS/2 line initially sold well, but their high price and proprietary Microchannel architecture eventually did them in. The machines were durable and well-built. Some of the preproduction IBM PS/2 Model 80s made their way to the University of Missouri, and I was still nursing them along in 1998. The idea of a server lasting 10+ years was ridiculous, and they were hopelessly obsolete at that point, but they worked better than they sounded. In more recent years, I’ve had 5-year-old servers give me more trouble than those deathbed-ridden Model 80s did.

OS/2, meanwhile, remains one of the most underrated technologies of all time. I’m extremely familiar with it, and I owe my career to my knowledge of OS/2. The university hired me on as a part-time desktop support technician largely because I knew OS/2, and a few years later hired me on as a server administrator because the HR department couldn’t find anyone in central Missouri who knew OS/2.

And although OS/2 failed as a consumer operating system, it thrived in certain niches. When a bank president who attended my church in 1998 heard I knew OS/2, he offered me a job right on the spot. Unfortunately for both of us, I’d accepted a job in St. Louis that very week. I’ve often wondered what-if over that decision.

You see, OS/2 was huge in banking. Any time you saw an ATM with a plain text display, it was a good bet the machine was running OS/2. It’s been a few years now since I’ve seen the old OS/2-based ATMs on a regular basis, but I still find one occasionally. It runs reliably on extremely modest hardware, and banks are reluctant to scrap something that works.

As it turns out, OS/2 plays a vital role in running New York City’s subway system too, as well as cash registers for the Safeway supermarket chain.

I know OS/2 use is dwindling. I told a recruiter many, many years ago that if he ever had a need for someone with a lot of OS/2 experience, to call me. He never has. And for years I used to search job boards every so often for OS/2 just to see what would come up. Not much ever did, and I guess I gave that up probably in 2008 or 2009.

But I guess I shouldn’t say OS/2 is completely dead. You see, OS/2 was initially a joint venture between IBM and Microsoft. That alliance fell apart in 1990 or thereabouts. Depending on who you believe, Microsoft betrayed IBM, or IBM betrayed Microsoft. What matters more is that IBM was working on OS/2 2.0 and Microsoft was working on OS/2 3.0. IBM’s line of the family led to the official OS/2 2.0, 2.1, 3.0 and 4.0 releases from IBM.

Microsoft OS/2 3.0 became Windows NT, which led to Windows NT 3.1, 3.50, 3.51, 4.0, Windows 2000, Windows XP, Windows Vista, and Windows 7, in addition to Windows Server 2003, Windows Server 2008, and all future Windows products.

I denied it longer than most–I’m pretty sure that I spent a few hundred dollars on OS/2 software as late as 1997 or 1998, at a time when a few hundred dollars was a huge amount of money to me. But OS/2 never really stood much of a chance.

Microsoft deserves some of the blame, because Microsoft did do everything they could to try to snuff out OS/2 and its 10% market share. I don’t think OS/2’s continued existence in those tiny niches bothers Microsoft anymore, but I also think that’s partly because so many of the Microsoft executives who remember it are all retired now.

The other problem is that in the 1990s, the public at large generally mistrusted IBM. The public didn’t understand Microchannel, but they did know that it was something that IBM did that was expensive and proprietary and they didn’t like it. It was extremely difficult to sell IBM PS/1 and Aptiva PCs at retail, even though there was little to no difference between them and the Compaq computers sitting next to them. I couldn’t sell PC DOS 6.1 or 6.3 either, even though they were cheaper than MS-DOS 6.0 and 6.22 and came with much better utilities. Since it had the letters “IBM” stamped on the package, people assumed there must be a catch.

And OS/2 was an even tougher sell. At a time when Windows 3.1 was barely usable and Windows NT 3.1 had such anemic driver support that it wouldn’t run on much of anything other than the Compaq Deskpro computers Microsoft developed it on, OS/2 2.1 and 3.0 should have made perfect sense. The 32-bit multitasking worked extremely well; DOS programs ran faster on OS/2 than they did under native DOS and you could run multiple Windows programs for weeks without them crashing the system or each other. But people didn’t want to look at it.