> “Originally I'd planned on posting this on Aug 24”
For context, Aug 24, 1995 was the release date for Windows 95. It was a huge deal back then. Microsoft was spending hundreds of millions on marketing, which was unprecedented in the software market.
OS/2 fans had seen the writing on the wall. Everybody already knew what the new 32-bit consumer Windows was like because it had been in beta testing for years.
(Things moved more slowly when a large-scale software release meant manufacturing millions of floppies and CDs, putting them in shrinkwrap boxes, and shipping them to retailers. No online patches either because Windows 95 didn’t even include a web browser. So it just had to work on millions of PCs out there.)
OS/2 2.1 had included a copy of Windows 3.x for users to run 16-bit Windows programs at the cost of IBM paying Microsoft for a Windows license. Later versions of OS/2 could use user-installed versions of Windows, which helped IBM avoid paying for a Windows license.
Since Win95 used Win32 APIs, unavailable to Windows 3.1, 32-bit Win95-apps couldn't run on OS/2 at all. So that spanking new version of Office 95? Not on OS/2.
It's a good way to get the old tiny fast Freecell on Windows 11. The FREECELL.EXE binary in Win32s runs fine on Windows 11 x86-64. (And on Arm64 come to that.)
Nitpick: Win32s was limited, and missed a lot of functionality. From the linked Wikipedia article:
The "s" in Win32s signifies subset, as Win32s lacked a number of Windows NT functions, including multi-threading, asynchronous I/O, newer serial port functions and many GDI extensions. This generally limited it to "Win32s applications" which were specifically designed for the Win32s platform."
There was a MS knowledge base article with the long list of limitations, archived here:
I recall Westwood Monopoly (great game, btw) requiring Win32s to be installed to run the game on Windows 3.x. It also required WinG (which eventually got deprecated for DirectX). I have fond memories of playing around with Freecell and Microsoft Doggie (test app for WinG).
Not a commercially viable target for most companies. Win32s was a stopgap for Win32 support for lower end machines when only WinNT was around. Once Win95 arrived, Win32s's days were numbered (along with 16-bit Win 3.x).
If you're designing a Win32 app for Windows 95, are you going to stick with the Win3.1 look and feel? No. You're redesigning your UI to look like a native Win95 app.
Where are your testing resources targeted? Win32s? Nope. Win95 only.
Well maybe if Win95 sales are slow, you'll do more testing on Win32s...
People lined up to buy Win95 at midnight on release day? [Before Harry Potter book and iPhone releases] Commercial consumer Win32s support died that day.
So running Win95 apps on OS/2 2.1 running Win3.1 running Win32s DLLs wasn't going to be a supported platform for software companies.
IBM didn't want Win32 apps. They wanted OS/2 PM apps.
OS/2 was so far superior to Windows in 1992 that it wasn't even funny. Nevertheless, my company gave up on OS/2 and we had to port our Smalltalk application to Windows 3 instead. Someone had to explain segmented 16-bit addressing to me, and I was totally appalled.
Fortunately, Windows 95 came out a few years later with the Win32 API and I pretty much forgot about OS/2 after that. I still feel a bit guilty, though.
a) Microsoft, then a tiny fraction of what it is now, took $400 million from IBM in co-developing OS/2 and put the profit from that money into Windows 3.0 and later NT (funny how developing NT cost only a stated $150 million just a few years later and the main cost is in software developers)
b) When Windows 3.0 did well enough in the market, MS did what they have always done - they figured out how to break their agreement with IBM, and take their own stuff that they owned 100% and run with it. This is a pattern seen everywhere at all times with MS, ever since the Ed Roberts / MS BASIC days in Albuquerque.
It's funny constantly see a kind of revisionist history where IBM is the misunderstood underdog with the superior technology that was destroyed by the big bad Microsoft. When in reality IBM is this poorly managed monstrosity that was terrible at software development and destroyed their PC market share trying to take back the PC standard and make it proprietary to them.
It's more like an opportunistic, willing to be sleazy, smaller company grew into a very large one by continually taking underhanded advantage of anyone in their path.
Ed Roberts of MITS/Altair, STAC, Wang (OLE, lawsuit resulted in MS investing in Wang), DEC (VMS, lawsuit resulted in DEC receiving money), to list just a few.
Whether or not you spell Microsoft with dollar sign it doesn't make every other company the hero.
History has turned OS/2 into some kind of folk hero rather than the bloated, mismanaged, expensive, buggy mess that it actually was. And Microsoft's technology was superior if the job of an OS was to actually run software on computers of the time.
I have read a lot about that time and it doesn't seem particularly biased. Microsoft hated working with IBM and their horrifically old-school way doing software development (e.g. counting lines of code, etc).
The bias, in my opinion, is the other way -- which is exactly what I said at the top.
Letwin says what killed OS/2 was that it could run Windows apps.
This is not true.
That only applied to OS/2 2; OS/2 1 could not run Windows apps.
It didn't matter; there weren't any.
Windows apps followed Windows 3.0. Windows 3.0 followed the failure of OS/2 1.x.
This is his bias: he was one of the project leads and was passionate about OS/2, which blinds him to the fact that it was OS/2 1.x that flopped and there was nothing OS/2 2.x could do to usefully come back from that.
His assertion is that 32-bit OS/2 flopped, because it lacked native apps, because it could run Windows apps. This is not true.
It flopped because Windows 3 was out, and Windows 3 was a hit because it did what 16-bit OS/2 failed to do: it ran DOS apps _well_ and it could do that because it took advantage of the 386 chip.
His bias means he fails to identify the true cause although he named it.
Secondarily, I think we can point to a 2nd, non-technical, personal reason: Letwin was not involved in Windows 3, which was essentially a 1-man skunkworks project.
Bill Gates really was a shrewd businessman, and never had a problem with being malignant if it suited his needs. He just did his sort of dirty deeds with a friendly facade.
I still haven’t forgiven him for what he did to Gary Kildall.
The background is a bit complicated, but the high level is that many believe Bill Gates stole MS-DOS from Gary to license to IBM.
The nuance is that IBM contacted Bill first, who he directed to speak with Gary. While Bill was under NDA, he did give Gary a heads up "Someone is coming to see you. Treat them well, they are important people." Gary did not and missed their visit and IBM went back to Bill.
Bill then bought Q-DOS (a rip off of Gary's DOS) for $75k and rebranded it MS-DOS and licensed it to IBM.
Minor corrections: Gates bought Q-DOS for $50k and Tim Paterson, who wrote it, was happy to sell at that price and never felt ripped off. I worked with Tim at Microsoft and asked him about these points myself.
IBM visited Microsoft about the computer languages and apps for their new PC. When IBM asked about an OS as well, Bill Gates, following the gentlemen's agreement that MS had with Digital Research, deferred to Digital Research, telling IBM that they should go to Gary Kildall at Digital Research for an OS. Bill Gates gives a vague heads-up to Gary Kildall about a meeting with some important people soon.
There's a lot of she-said-he-said about the initial IBM-Digital Research meeting. Gary wasn't there at the start, his wife handled the meeting. In the end, Digital Research treated IBM as just another OEM - if IBM wanted CP/M, they'd pay the same royalty rate as any other OEM. The IBM folks weren't keen on paying a royalty per OS license installed. Plus, their Project Chess PC didn't use an 8-bit CPU. They'd need an OS that could run on a more recent CPU (16-bit or more), And they needed it fast - Project Chess was going to ship in one year, so Digital Research had less than one year to get a 16-bit OS shipping. Gary Kildall wouldn't commit, so IBM walked away empty-handed.
IBM met with Microsoft again and mentioned that IBM hadn't signed a deal with Digital Research. If there was no OS, there would no IBM PC and no revenue from licensing MS-BASIC et al. One of the Microsofties mentioned that they knew of an OS for the Project Chess PC, Intel's 8088. Did IBM want it? IBM said yes. Microsoft offered to act as a go-between or IBM could deal directly with the new OS company. IBM told Microsoft to handle everything.
When it came down to paying for the OS, PC-DOS, Microsoft offered several options:
- per PC license royalty
- flat fee for a non-exclusive license with unlimited installs
IBM chose the flat fee, which meant less accounting work over time, but we all know what that "non-exclusive" would enable - PC clones.
IBM would charge a relative pittance for their PC-DOS, $40. Digital Research came through with CP/M-86 for the IBM PC, but they were charging much more, almost $200 more, and shipped six months after the IBM PC launched. IBM paid $45K to Microsoft for PC-DOS, so after the first 1,125 PCs, PC-DOS revenue was pure profit for IBM.
Digital Research finally relented and lowered the price of CP/M-86 after one year, but IBM released the IBM XT and PC-DOS 2.0 with support for hard disks, subdirectories, etc.
But a couple of months before either of those events, an app written in assembly language was released for PC-DOS - Lotus 1-2-3.
Tim Paterson has basically said that he bought a book on CPM and then renamed the API calls for QDOS. For instance (hypothetical example) if CPM had "OpenSerialPort" then Paterson might have created the same API but named it "SerialPortOpen". Note I'm not saying that Paterson did anything illegal.
This is an interesting video but I am not able to determine what series it is from:
https://dai.ly/x2xzxzo it does have interviews with Gates, Ballmer, Paul Allen, Tim Paterson, and Jack Sams of IBM etc.
Tim Paterson appears about 5m50s into the video and directly states that he had a book on CPM and used that as the basis.
It was true -- but he was flying to visit a client, and he was a programmer not a lawyer. His wife, Barbara McEwen, was the company lawyer and she did the negotiating. It was her job, not Gary's.
I mean Gates made up the infamous apocryphal story that he was flying for fun, careless about IBM and whatnot. Not that he wasn't actually on a plane (coming back from a business meeting).
Gates told that story to reporters. That is fact. There's plenty of info on the internet, starting from Kildall' Wikipedia page and its sources. You don't have to dig that hard.
I'm sure Gary Kildall had his faults or negligence about the deal gone wrong with IBM. That's a complicated story. But still, he was haunted his whole life by the silly "went flying" story. And that's one thing that Bill Gates did to him.
> Microsoft, then a tiny fraction of what it is now, took $400 million from IBM in co-developing OS/2 and put the profit from that money into Windows 3.0
The timing didn't happen the way you're describing it. There are numerous accounts of the tale of how David Weise and Murray Sargent got Windows running properly in protected mode as an unauthorized, personal side-project.
I didn't say anything about protect mode and the URL you included mentions Windows 2.1, which was NOT a commercial success. But Windows 3.0 was the first time that traction was truly gained in the market. That they added protected mode was great, but that didn't directly affect the MS Office juggernaut.
Yep, if you have in one hand "Inside Windows NT" by Helen Custer (or one of its successors, "Windows Internals"[0]) and on the other "VMS Internals and Data Structures"[1]...
DEC cancelled Cutler's pet projects Mica and Prism, a 64-bit RISC chip and a portable multi-personality next-gen OS to run on it.
He was upset. Ripe for head-hunting. MS did so, but he insisted on bringing his core team.
MS did not know what to do with him and put the team to work on some corner of LAN Manager.
Then Windows 3.0 was a hit, which led to IBM and Microsoft divorcing. IBM kept OS/2 1.x (16-bit, for the '286) and OS/2 2.x (32-bit, for the '386).
Microsoft got OS/2 3.x, a planned CPU-independent portable OS.
It was working on it on Intel's RISC chip, the i860, codenamed N-Ten. The OS was named after it: OS/2 NT.
Then, with Windows now suddenly a big deal, getting a next-gen Windows working was suddenly a big urgent issue.
Cutler got the job: finish OS/2 NT and make it work.
OS/2 NT was renamed Windows NT, and as it was barely a skeleton of an OS, Cutler took a lot of the design of Mica -- the cancelled next-gen VMS -- and built NT around that.
DEC found out and sued, and DEC got very sweet deals on NT and Exchange and things as a result. (Compaq totally blew this and died as a result. It deserved worse.)
DEC also salvaged the Prism project and it was launched as DEC Alpha, the first 64-bit RISC chip.
Alpha was the first 64-bit platform NT ran on, and it was also the first non-x86 platform Linux ran on.
Interestingly enough, NT used the Alpha as a 32 bit processor[0]. Although NT on Alpha was used to port Windows to 64 bits in lieu of using the (very slow) Itanium simulator[1].
Actually, this was discussed on HN earlier[2], and I believe you wrote an article for The Register[3] about it :)
R&D funding provided by IBM wasn't classified as "profit". The Windows NT code base didn't use much from OS/2. In the end OS/2 was mostly worthless garbage to be thrown away; it had some good functionality in a few limited areas but was essentially a "local maximum" and had no viable path forward. If anyone wanted to turn OS/2 into a viable modern OS they would have had to start over from scratch.
I'd disagree that OS/2 was "worthless garbage", as MS would keep using it for several generations for their server/LAN OS.
I'd argue that OS/2 2.x was pretty much a ground-up re-write, as OS/2 1.x had been written for 16-bit 286 protected mode and not the 32-bit 386 protected mode which the next version targeted.
IBM did insist that the new version run all of the 1.x 16 bit stuff, but like DOS and Win 3.x mode this was done through some virtualization if memory serves.
The big hurdle OS/2 faced was being memory hungry at a time when memory prices were still high and when the installed base had less then half of the required amount by default.
It wasn't just memory requirements that hurt OS/2. I'd argue one of the biggest hurdles was IBM couldn't stop being IBM for even one second to do any reflection on the realities in the PC market. They still thought they were going to have a renaissance with 386 PS/2 systems for goodness sakes.
And man were they expensive. $195 for OS/2 2.0. That's about $434 in 2024 dollars, and the PS/2 systems that ran it best started north of $2k if memory serves. No one outside of corporate and the biggest enthusiasts were shelling out that kind of money for "a better DOS than DOS" (cough Desqview cough) or "a better Windows (3.1) than Windows (3.1)" (irrelevant and not that impressiive imho).
Love them or hate them, Microsoft has always been good at hitting the 75th percentile, aka "good enough" and "cheap enough". They proved (conclusively) that customers will put up with a mountain of shit if you cost half of what your competitors do. Plus they knew how to make ISVs a lot of money too. IBM just couldn't get it's head out of it's ass, I mean the 1960s.
And for the record, I salivated over OS/2. I scored a beta of 2.0 when I was 14 off of a BBS associate. I had an IBM laser printer, IBM typewriters, an original IBM PC. I thought back then that IBM meant quality (and to be fair, it did).
But they didn't get it and Microsoft did. Windows 95 had pizzaz and hype, the CD had Weezer's Buddy Holly video (video!). They were working to get games running right under 95. They were courting all of the biggest application vendors. They were doing shit. IBM? Too busy being IBM.
> They still thought they were going to have a renaissance with 386 PS/2 systems for goodness sakes.
Yeah but...
If OS/2 1.x had been a 386 OS and delivered on the promise: great DOS compatibility, multitasking including of DOS apps, and built-in networking, with a passable GUI based on Windows 2...
I think it could have been a hit.
I was there, supporting this stuff in production back then.
DOS was a PITA and getting networking working on DOS and still having enough of that all-important first 640kB of RAM left to run anything was hard. I was a master of it. My skills at it landed me several jobs.
NT made all that disappear. RAM was just RAM, each DOS session got all of a virtual instance's RAM dedicated just to it -- and a network drive looked like a local drive. It was like black magic. It was amazing.
Big drives, with a solid filesystem. Long file names. TCP/IP in the box, as standard.
NT 3.1 was amazing stuff, but that was in 1993 and you needed a £5000 PC with 32MB to run it.
OS/2 2.0 delivered this, smaller and faster, in a quarter of the RAM, the year before...
It was amazing. It was a phase shift in the industry. But the networking was extra, TCP/IP was extra, etc. etc.
And it was late. I just wrote about the beta of MICROSOFT OS/2 2.0:
But if IBM hadn't screwed the project in 1985 or so, OS/2 1.0 could have done that in 1987 or so.
Before Linux (1991), before Windows 3 (1990), before 386BSD (1989).
It could have been the amazing thing that the hype promised. The tech was there and it worked. It could have been got ready in the 1980s.
A PC industry controlled by IBM would be no better than one controlled by Microsoft and it would have been more expensive.
But we all suffered years more of the crap of DOS and DOS memory management and Windows 3.x, because IBM fscked up.
I don't know what would have happened.
Maybe it would have forced the Unix folks to adopt Arm 20 or 30 years earlier and make RISC boxes that were cheaper and cooler-running than x86? Maybe those expensive IBM OS/2 x86 machines would have forced BSD onto Arm and what happened 25Y later with Apple kit happened a generation before with Acorn kit.
I am 100% not saying it would have been a better world... but it would have been a much more different one than the closed narrow imaginations today portray.
> IBM did insist that the new version run all of the 1.x 16 bit stuff, but like DOS and Win 3.x mode this was done through some virtualization if memory serves.
Kind of, but IIRC it was a fundamentally different and more low-level kind of "virtualization": The 80386 processor's virtual[1] 8086 processor mode. Not much for them to do in the OS code itself, compared to writing a whole VM to run DOS apps on (he confidently says, never having written anything of the kind himself).
[1]: Probably not exactly the right word, but I can't recall the correct term right now.
OS/2 had true preemptive multitasking, which couldn't be matched on Windows systems until Windows 95. Far from worthless garbage, it was one of the only systems available that could do true multi-tasking on cheap PCs, at least until 95.
I used OS/2 Warp to run a 2 node BBS on one PC in 94-95 time frame. The only other OS (that I'm aware of) which had that capability was DESQview¹ and that lacked any sort of GUI.
Can you be a bit more explicit in describing what the problems were? I don't immediately see why that wouldn't apply way more to Windows at the time.
I did not use OS/2 much, but.. I did write some communication software so that we could work on a remote computer (this was via X.25, which was the option at the time for companies staying connected), and that was easy. Modern, I would say. Thread model easy to work with. I wrote a DMA driver as well. Again, a great API to work with, and overall a good experience.
The user interface (GUI)? Any problems? A roadblock for the future? Can't really remember how it worked, but again - compared with Windows (even NT, which came about later), I can't recall anything "worthless garbage" with no path forward.
It sounds like the grandparent poster is maybe talking about OS/2 1.x, which was definitely hamstrung by 286 compatibility.
But OS/2 2.0 was practically a 32-bit rewrite of the entire system with a new GUI too. It certainly was competitive with the first versions of Windows NT.
The most fundamental problem with OS/2 is that it wasn't designed as a multi-user OS from the start. It had some multi-user functionality but there were core architectural problems that made it impossible to move forward without a complete rewrite.
On the UI side it also had a single input queue. Which was such a stupid design decision that I can't imagine what they were thinking. Fixing that would have also required a major rewrite.
It was garbage while Microsoft was working on it. That NT wasn't garbage is due to it's being written by people from VMS who knew what they were doing. IMO.
That reminds me of when I briefly worked for IBM - an unpleasant corporate culture. I don't think it has anything to do with the end results, however. i.e. good people still ended up working for them.
1.2 was problematic and 2.0 worked. 1.2 was Microsoft and 2.0 was IBM. From what I have read the NT team didn't want any of the MS OS/2 developers - this is just memory - and it was related to perceived code quality or something like that.
Well if OS2 had succeeded we would have PC's controlled by the OG evil tech monopoly which have had their business dealings in far worse things (https://en.wikipedia.org/wiki/IBM_and_the_Holocaust) than microsoft ever did.
Especially since it's now the erstwhile upstart frenemy that sits in the driver's seat. They have institutional memory of where they came from, so they know to be on their guard against the same thing happening to themselves.
As someone with both feet in the embedded development world, reading this from 1995 is fun:
Note that even the RS232 port itself is a standard. And an inferior one; sending stuff at 9600 baud over a 7 wire connection is a travesty by modern standards.
Note: RS232 is still around, and still relevant on modern systems. Not perhaps "actual" RS232 with the large voltage swing and negative voltage, but the UART signalling.
Also, I wonder what he would have preferred, USB was really really young then.
Around 2010 I met a former coworker for lunch and he told me how they were running serial stuff really fast. I told him I was also working on high speed serial stuff. I asked him how fast and he said 115 kilobit/s. I told him I was working on 8 gigabit/s or around 70,000 times faster. At that point I had not used RS232 in over 10 years. I have a friend who still uses RS232 for lots of industrial factory type stuff in 2024.
Yup. Windows 95 first release did not support it. In only appeared with Windows 95 OSR2 a year later, alongside FAT32.
In 1996 I got an IBM Pentium PC to review at PC Pro and it had these weird little rectangular ports on the back, with no pins, and a trident logo. I called IBM and asked what they were, and after bouncing around a dozen people until I got a techie, he told me they were "this new Intel thing" called a "unversal serial bus". I asked if it was some debugging kit, and he said no, it's meant to be this universal I/O standard.
Then I got another long game of phone tag at Intel where nobody press-facing had ever heard of it.
But nobody had any drivers for it, and there was nothing to plug into it, so it never even became a story.
I can't remember now if I even wrote about it, but probably it was a passing mention. If so, I think I probably gave it the first coverage in the world.
If Apple hadn't adopted USB for the iMac, especially the fruity colored iMacs, which then spawned hundreds if not thousands of brightly colored accessories, who knows if it ever would have taken off on the PC side of things. I still see the occasional motherboard with a PS/2 port :p
Long story short: Gates and Microsoft were in domination mode and everybody else was in experimentation mode. Microsoft's management are par excellence opportunists.
And if I recall correctly from watching one of the Bill's interviews from the 1990s, he even couldn't believe and wrap his head around why IBM didn't do the OS themselves and why the hell they decided to outsource it to Microsoft.
> Standardization is a big plus in the computer field. You're much better off
having thousands of products and vendors compatible with a single standard,
even a mediocre one, than having dozens of products, one or two each for
each of a dozen fragmented standards.
A lesson North America decided to relearn the hard way for EV charging. Europe got it right.
North America's eventual EV charging future is CCS charging with the J3400 plug. But it's still got a ways a go with wasted money on adapters and retrofits in the meantime.
In August 1995, OS/2 was already dead. Windows NT 3.1 was already released in 1993, Windows NT 3.5 in 1994, and NT 3.51 (Daytona) around the same time as Windows 95. Those NT releases eventually replaced Windows 9x entirely. IBM wasn't even focused on the real product because they were rope a doped into making the world's worst character-driven graphics UI.
Yes, OS/2 was dead software walking by then. That is not a big revelation: it is the central theme of this email, and it was only a year before the last ever major version of OS/2, Warp 4, released 25 September 1996.
So, yes, we know that.
What is more interesting for me is that Letwin's post misses the real point -- he mentions it, then moves on, not realising he just stated the core problem -- and that you seize upon a trivial, minor point of the message.
> rope a doped into making the world's worst character-driven graphics UI.
What?! NO!
That was TopView. That was a whole decade earlier. It was before OS/2, before Windows 1.0, before this story even started. Forget TopView.
TopView: March 1985
Windows 1.0: November 1985.
The only thing that matters about TopView is that it inspired DESQview, and that is entirely irrelevant to OS/2.
No, it's not about UI at all.
The real domino falling that killed OS/2 wasn't that OS/2 2.x ran Windows apps. That's a red herring. It's true but it was a symptom of the decline, not a cause of it.
The reason why is trivial to work out without deep knowledge of the history, so I will parenthesize it.
(OS/2 2.x didn't get native apps because it could run Windows 3.x apps.
This means: All the important stuff was on Windows.
This means: Windows was already dominant.
This means: a good version of Windows was already out when OS/2 2.x appeared.
In other words: OS/2 2.0 followed Windows 3.0.
This means: OS/2 had already lost the battle. The war was already won.
Why? Because OS/2 1.x had flopped.
Why? Because OS/2 1.x was limited to the 16-bit 80286.
So? So because OS/2 1.x flopped, Microsoft launched a product that fixed the key weakness of OS/2 1.x. That was Windows 3, which ran on 286 machines but the same binaries ALSO ran on 32-bit 80386 PCS, and used the hardware to give better multitasking of the market-dominating DOS apps than OS/2.
Win3 beat OS/2 1.x, because OS/2 1.x was crippled, so it flopped, allowing Windows to gain the upper hand.)
The really important bit of Letwin's post isn't what he thinks it is.
(And it's definitely not that footnote about text-mode UIs.)
The really important bit is not Windows apps; it's why Windows was able to gain that dominance.
It is this part:
«
The miscalculation came about with the 386 coming out sooner than we
expected.
[...]
When the 386 did come out earlier than expected and we saw what was happening,
Microsoft wanted to abandon OS/2 1.0 before it was released and work on a 386-only version, one that would be able to emulate more than one DOS box and do a better job, at that.
»
From what I understand, IBM didn't want to make the leap to 32-bit earlier because they were still making a boatload of money selling 286-based computers (PS/2 50, 60)
They were indeed. At that time I was installing and setting the things up.
99% of them ran plain unaugmented PC DOS 3.3 and DOS apps, and the customers didn't give a fig about OS/2.
IBM lost a $multi-billion market through its short-sightedness, and if it had more vision, it could have couriered a 386 planar to every 286 PS/2 owner who wanted OS/2 and sent an engineer to install the board and then the OS... and still made more money overall.
The gist of it: IBM botched the "extend" part of embrace and extend. OS/2 was better but not by enough to migrate and stay. It had no "killer app" that drew the crowd and developers kept targeting Windows which OS/2 supported.
Also IBM f*ed up the OEM deals. OEMs preferred Microsoft which wasn't a direct competitor in the PC business.
By the time OS/2 came out, IBM wasn't a serious competitor for most OEMs -- they were high end, so were mostly competing with the likes of Compaq.
The big issue that no one is mentioning is that OS/2 needed 8MB of RAM to run decently, preferably more, but this was when most machines were 2-4MB and extra RAM was still a big cost.
When I was a student, around 1992-93, I did shift work for a few months in an MS production facility in Dublin. When it was slack I got assigned to degaussing floppies - most of them were "Microsoft OS/2 Version 1.0" install media. I never kept any for posterity :/
OS/2 2.X was superior to Windows 3.0 but lacked the apps and marketing to compete with it. IBM Client Access/400 for DOS ate into the 640K layer and OS/2 used Extended RAM to keep more of the 640K for WIN-OS2 and DOS programs. Once Microsoft released Visual BASIC for 16-bit Windows, IBM had nothing for OS/2 mode like Visual BASIC. Borland had Delphi and Turbo C++ for Windows 3.0 but not for OS/2.
OS/2 started as Advanced DOS at Microsoft and IBM, by the way. An open-source project called OSFree (OS/3) wants to write OS/2 from scratch: https://osfree.org/
At my employer, I had to downgrade OS/2 2.X systems to MS-DOS 6.2 and Windows for Workgroups 3.11 before OS/2 Warp 3.0 and Windows 95 came out. Microsoft also had affordable license fees for DOS, Windows, and Office. We had to have everyone on the same OS for compatibility reasons and OS/2 lost out.
> IBM Client Access/400 for DOS ate into the 640K layer and OS/2 used Extended RAM to keep more of the 640K for WIN-OS2 and DOS programs.
Nah.
OS/2 1.x was a native 80286 OS and it used one big block of 16-bit RAM. It did not have the concept of "Extended memory" (XMS) -- that was a DOS and Windows 3.x thing.
I was working with this stuff at the time, too. :-)
Extended Memory is just the name for memory above 1MB that could be accessed by the 80286 in protected mode. The 286 was still a 16bit processor, still had segments (although descriptors in protected mode) and "one big block of 16-bit RAM" is 64KB.
The single DOS session available on OS/2 1.0 still ran in real mode.
AFAIK the split into conventional/base and extended memory is largely a DOS concept.
It's of course based on the 8086 real mode being fundamentally limited to 1M and the 286 protected mode not being limited to that.
But if you divorce yourself from the DOS memory model, I don't think there's a fundamental need in protected mode to treat 0 to 640k (or <1M) differently from >1M, or for protected mode to have a concept of extended memory.
"One big block of 16-bit RAM" probably means the entire RAM (accessible in protected mode) addressed in 16-bit segments. If you wrote an OS that uses 286 protected mode and don't care about the DOS memory model, that's what you'd have, and there would be no need for distinguishing between "base" and "extended" memory once you've entered protected mode.
I'm not really that familiar with OS/2 so I don't know how exactly it handled the switch to real mode for the DOS session or what it did about keeping the rest of the memory separate from the <640k of the DOS session.
But in general the base vs. extended memory split is a DOS memory model thing.
Many PC BIOS implementations also reported base and extended memory separately in POST output for a long time but I think even that's just following conventions based on the DOS memory model.
The original comment was about running DOS programs under OS/2 1.0. The comment I replied to seemed to be under the impression that OS/2 didn't care about the DOS memory model, but of course it did, because it ran DOS programs in real mode.
But really calling memory above 1MB Extended Memory was just to differentiate it from Expanded Memory which was memory that was banked into the address space above 640k.
Without large model or something, one segment was a block of 64kB and me, I'd call that an 8-bit block of RAM.
OS/2 1.x allocated all the computer's RAM as its RAM, with no conventional/extended split -- unless or until you ran a DOS box. The handful of times I used it, it was a fileserver with 3+Open and those never ran DOS boxes.
Slight correction - there were some Borland tools for OS/2, but after the split MS stopped putting out new tools; but IBM did have some interesting compiler stuff too.
> In 1992, Apple and IBM cofounded Taligent, based upon Pink, an operating system with a mass of sophisticated object-oriented compiler and application framework technology from Apple. Pink became CommonPoint, the partnership was dissolved, and CommonPoint was absorbed into VisualAge starting with the Compound Document Framework to handle OLE objects in VisualAge C++ 3.5 for Windows. In February 1997, the first mass release of Taligent technology came in the form of the Open Class within VisualAge C++ 4.0 ... The original prototype which led to VisualAge was an attempt "to make something like the NeXT interface builder" within the Smalltalk/V development environment.
Two spaces always put me off typographically in monospace. The sentence-ending dot plus a single space already create a much larger gap visually in monospace than in proportional fonts, and with an additional space it just becomes ridiculous. It would be fine in proportional font I guess.
>As an oldster, I have to force myself to not use two spaces, so I don't reveal my oldness to the youngsters. Once space still looks weird to me.
People make weird claims that two spaces was from typewriter days to keep them from jamming or something, but honestly two spaces is so much more readable and it's annoying that web things and work processors try to remove them.
I think it is both. IBM may missed the point in a big way. They were an slow corporation to market an OS home product. The PC became a home product and IBM was not ready to be on that kind of circus.
For context, Aug 24, 1995 was the release date for Windows 95. It was a huge deal back then. Microsoft was spending hundreds of millions on marketing, which was unprecedented in the software market.
OS/2 fans had seen the writing on the wall. Everybody already knew what the new 32-bit consumer Windows was like because it had been in beta testing for years.
(Things moved more slowly when a large-scale software release meant manufacturing millions of floppies and CDs, putting them in shrinkwrap boxes, and shipping them to retailers. No online patches either because Windows 95 didn’t even include a web browser. So it just had to work on millions of PCs out there.)
reply