Indeed and a very limited number of drivers, which were also 16-bit (as described in the linked article).

I used OS/2 2.1, 3.0, 4.5 and eComStation (latest stint was probably at the beginning of the century). It was a profoundly singly-user system that crashed quite often, putting it pretty much on par with Windows 95/98/ME.

Windows NT 4, FreeBSD, and even BeOS were far more stable and modern operating systems than OS/2. NT 4 also had far better application support than OS/2 and BeOS had the nicer GUI, filesystem, better multithreading, etc.

Although OS/2 is a bit of youth nostalgia for me, anyone with an objective view would acknowledge that OS/2 was a 16/32-bit patchwork single-user OS that was not ready for the future. Microsoft made the right choice ejecting itself from the project and running with what was originally planned to be OS/2 3.0.

Microsoft convinced IBM to use a single message queue despite protest from IBM. Perhaps they were purposely sabatoging the effort, as they clearly wanted their 386 fork (which became Windows NT) to have the upper hand.

Anecdotally, that was kind of the feeling at IBM about several of the misfeatures of OS/2 in 1990-92, when I worked there with OS/2.

Typical conversation:

"Why the hell does..."

"Microsoft."

"Right."

And the feeling of mis-features being the other company's fault was mutual. Microsoft people viewed the introduction of more UNIX-like features into MS-DOS/PC-DOS 2 as happening despite the strenuous objections of IBM, for example.

People do this. I've seen this happen in places that have nothing to do with IBM or Microsoft. Problems and misfeatures are all the fault of that group of incompetent outsiders in that company over there that we have a business deal with. It is rarely actually true, in my experience.

In any case, it probably wasn't that hard to fix and OS/2 2.0 would have been a good time to fix it.

> OS/2 also had a single threaded event queue

This is another myth, long since debunked (I saw this one debunked in the early 1990s.) to resurface here all of these years later.

OS/2 had multithreaded input handling. Every application had its own event queue. Indeed, Presentation Manager allowed every thread in every application, that actually used Presentation Manager, to have its own event queue. A PM thread first created an "anchor block", then using (its handle to) that anchor block created a message queue.

What OS/2 had was synchronous input handling. Windows NT has desynchronized input handling.

The idea that "OS/2 had a single input queue" whereas "Windows NT has multiple input queues" is an erroneous one based upon an erroneous description of the twain. In fact, both Windows NT and OS/2 have a single system input queue. In Windows NT, this is called the raw input queue, and by the same erroneous logic that OS/2 "had a single input queue" so too does Windows NT, because the queue that people point to as the supposed reason that OS/2 had one queue is the OS/2 equivalent of that raw input queue.

It is not the number of queues, and never has been the number of queues. Presentation Manager had multiple event queues. It is the synchronous nature of processing input events versus the desynchronized way of processing them. This deals not in the number of queues at all, but in how the GUI model copes with things that need synchronization amongst multiple applications, such as (for example) all of the messages that fly about under the covers (amongst those multiple queues) as part of a keyboard focus change.

Something else OS/2 got wrong: Thinking anyone cared about the difference between one linked-list or ten linked-lists with a single producer.

We can't have it competing with 'proper' operating systems on IBM proprietary hardware though, can we?

I first used OS/2 2(.11 I think), which was expensive (but my dad bought it). OS/2 3.0 Warp on the other hand was given away for free with a ton of computer magazines...