IBM OS2

[louisw3]

I don't think IBM is paying much interest to old OS/2 releases anymore...

They didn't pay much attention when it was current either. They just had some insane mission to prevent Microsoft from cutting into their mid-range turf, and how did that work out...

After seeing Pigskin & Football OS/2 beta's from 1987 it's pretty clear that Microsoft wasn't the one who screwed up OS/2, but rather it was IBM. Xenix 2.x for the 386, and Windows/386 were a thing in 1987, just as OS/2 with the ability to support v86 mode easily could have been a thing, but IBM had to make sure it was 286 only.

[WwOS]

… Microsoft wasn't the one who screwed up OS/2, but rather it was IBM.
… just as OS/2 with the ability to support v86 mode easily could have been a thing, but IBM had to make sure it was 286 only.

OS/2 didn't turn out so bad, but Windows NT was already available. But remember that it was Windows 3.11 and Windows 95 with the huge success, and as far as I remember, Windows 3.1 was 286-capable as well. Only since Windows 95 -- from 1995 -- a 386 was minimum. One might say that Windows 3.x was such a huge success because it still supported the 286. Windows 3.0 even the 8086 (and the 80186). OS/2 1.x was from the late 1980ies, so it definitely needed support for at least 286 PCs. It makes sense. The 386 was very expensive and a lot of people were happy with a 286 in the mid-80ies. OS/2 2.x, developed from IBM alone, started being 386 and better. That was released 1991. So it was the other way around: Microsoft kept the 286 ability in Windows 3.1x and made 386 a requirement starting with Windows 95 four years after IBM.

[louisw3]

… Microsoft wasn't the one who screwed up OS/2, but rather it was IBM.
… just as OS/2 with the ability to support v86 mode easily could have been a thing, but IBM had to make sure it was 286 only.

OS/2 didn't turn out so bad, but Windows NT was already available. But remember that it was Windows 3.11 and Windows 95 with the huge success, and as far as I remember, Windows 3.1 was 286-capable as well. Only since Windows 95 -- from 1995 -- a 386 was minimum. One might say that Windows 3.x was such a huge success because it still supported the 286. Windows 3.0 even the 8086 (and the 80186). OS/2 1.x was from the late 1980ies, so it definitely needed support for at least 286 PCs. It makes sense. The 386 was very expensive and a lot of people were happy with a 286 in the mid-80ies. OS/2 2.x, developed from IBM alone, started being 386 and better. That was released 1991. So it was the other way around: Microsoft kept the 286 ability in Windows 3.1x and made 386 a requirement starting with Windows 95 four years after IBM.

OS/2 turned out really bad in that Microsoft had a Windowing environment already with Windows 1.0. IBM didn't want Windows anywhere near their OS/2 so Presentation Manager works 180 degrees opposite of Windows. No doubt that this antagonized Microsoft to a great deal. Just as in 1987 we saw Xenix for the 386, and Windows/386 v2.01 released. Microsoft was clearly capable of not only using the 386 in 32bit protected mode, but also being able to write VDD's and emulate a PC in v86 mode.

Instead IBM dragged their feet in insisting that OS/2 not just support the 80286, but exclusively support the 80286. The 32bit version of OS/2 was built in parallel with the 16bit version but kept away from end users, a really major mistake.

The next big mistake was the insane cost of SDK's DDK's and redistributes for the OS. Microsoft clearly knew that you needed to court ALL developers, not just big ones to make a platform successful. This also meant making languages cheap and complete enough for small users. QuickC/Visual Basic on Windows 3.0 were the killer apps. Don't forget that OS/2 did have Excel and MS Word, but they most certainly did NOT have any cheap development tools.

The final straw of course was not only that Windows was excluded from OS/2, but Windows 2.x started to get real traction with 3rd parties writing major apps for Windows. And Windows 3.0 was the final game changer as running Windows on top of a DOS Extender shattered the 640kb dos limit, unlocking the capabilities of the 80286/80386 processor, and best of all it continued to run on MS-DOS, not requiring users to change their OS, or go through the hell of trying to find compatible device drivers.

OS/2 NT has shown just how versatile and solid the design was, as it survived first the abandonment of the i860 processor after it was shown to be inadequate for performance, then dumping the primary OS/2 2.0 Cruiser API for the expanded Windows API. Looking at the 1991 previews of Windows NT, you can see that the NTOS + Win32 is clearly in place, and that enough of Windows 3.0 is running on NT to make an OS. Although all the expanded parts of NT are nowhere to be found (additional subsystems, security, etc). There was a long road to get to 1993 for Windows NT 3.1, although the product wasn't really truly feature complete until Windows NT 3.5 where Microsoft finally had released their own TCP/IP stack, along with major performance enhancements (Daytona!) on the 386 reducing the memory footprint and letting Windows NT truly shine.

OS/2 2.0 really suffered with the loss of Microsoft, and IBM had to back-peddle quickly on their Windows stance by creating the Win/OS2 port of Windows to run under OS/2. You can find the first release of this that was bundled with Citrix Multiuser 2.00. The kernel was never fully 32bit, and it really showed with the strain of bigger disks, more memory and adding things like SMP. OS/2 just couldn't scale, and it's MS-DOS inspired config.sys became a nightmare to maintain and configure, things like adding a CD-ROM became a multi day chore for first time users. Binary .rc files for Presentation Manager didn't help things at all either. IBM's reluctance to add in full networking, multimedia and server functionality into the base OS/2 made it a dead end choice by the time NT had gone retail. Even on a workgroup using Windows NT Workstation let you setup a mail postoffice, and let users map drives onto your workstation. This kind of functionality was absent in the few years of opportunity of OS/2 2.0-2.11

By the time IBM had started to think that this internet thing was important, they bundled a GOPHER client of all things, and crippled the consumer version to DIAL UP ONLY. Not to mention again the cheapest tools for writing apps for OS/2 was WatcomC/C++ which was not a cheap endeavour, although with the Watcom tools you could also target 32bit MS-DOS, Windows NT/Win32s along with OS/2, and simply put there was a LOT more MS-DOS / Win32s users than there ever was OS/2 users. There was a bigger 'bang' to get DOS4G/W apps written as they'd run on MS-DOS & OS/2, but something that only required OS/2 was basically condemned to obscurity.

Edit:

To put it in perspective, this is why Windows won:

[WwOS]

OS/2 turned out really bad in that Microsoft had a Windowing environment already with Windows 1.0.

Yes, and Microsoft copied it from Apple. And Apple copied it from Xerox. And Microsoft wasn't the only one, there was also Digital Research with GEM, and then there were numerous others, like Amiga, Commodore, Atari, Acorn, ...

Only, Microsoft was safe since they had a written agreement with Apple, so the copyright violation court case was eventually dismissed.

IBM didn't want Windows anywhere near their OS/2 so Presentation Manager works 180 degrees opposite of Windows.

I don't even know what that means. OS/2 had quite a few similarities with Windows.

Just as in 1987 we saw Xenix for the 386, and Windows/386 v2.01 released.

Xenix was a whole different story. Windows/386 was just a step in the door, nothing more. Windows/386 can hardly be compared to Windows NT or OS/2 32-Bit or even Xenix on the 386.

Instead IBM dragged their feet in insisting that OS/2 not just support the 80286, but exclusively support the 80286. The 32bit version of OS/2 was built in parallel with the 16bit version but kept away from end users, a really major mistake.

Maybe. I still think that back in the late 80ies an exclusive 386 operating system wouldn't have been entirely successful. 32-Bit systems were the server systems of that time and they cost a lot of money. Not even to mention that the situation for programs was not ready for 32 bits. Compatibility issues would have killed the operating system before it started. So OS/2 being 16-Bit and DOS compatible makes kind of sense to me.

The next big mistake was the insane cost of SDK's DDK's and redistributes for the OS. Microsoft clearly knew that you needed to court ALL developers, not just big ones to make a platform successful. This also meant making languages cheap and complete enough for small users. QuickC/Visual Basic on Windows 3.0 were the killer apps. Don't forget that OS/2 did have Excel and MS Word, but they most certainly did NOT have any cheap development tools.

Yes, IBM did lots and lots of things wrong. E.g. they wanted to get their IBM PC back exclusively to them, which failed: the PS/2. It had proprietory hardware that went along with closed and expensive software, just like OS/2. Clearly IBM didn't understand the market as Microsoft did, but then Microsoft made only software and sold it to anyone who made hardware. IBM on the other hand wanted to combine the two, just like Apple did with their Macintoshs. And on top of that IBM had a core business that wasn't the PS/2s or the PCs... Mainframes, business hardware (BIG servers) and business software. Apple didn't have that, neither did Microsoft.

The final straw of course was not only that Windows was excluded from OS/2, but Windows 2.x started to get real traction with 3rd parties writing major apps for Windows. And Windows 3.0 was the final game changer as running Windows on top of a DOS Extender shattered the 640kb dos limit, unlocking the capabilities of the 80286/80386 processor, and best of all it continued to run on MS-DOS, not requiring users to change their OS, or go through the hell of trying to find compatible device drivers.

Interesting. Didn't OS/2 stay 16-Bit at first for precisely that reason? Wasn't OS/2 also DOS compatible? I think it was even later when it was 32-Bit, but more like Windows NT by runnig DOS apps inside a VM, which prevented legacy programs trying direct hardware access from running. That is when they also included WinOS/2...

That said, the primary success was Windows 3.x, not NT 3.x. As I remember NT had exactly this problem: most DOS programs wouldn't run. And there were lots and lots of DOS applications around. What Microsoft did with Windows 95 was genious in this regard: it kept 100% DOS compatibility while providing the new 32-Bit API at the same time! NT and OS/2 couldn't do that.

OS/2 NT has shown just how versatile and solid the design was, as it survived first the abandonment of the i860 processor after it was shown to be inadequate for performance, then dumping the primary OS/2 2.0 Cruiser API for the expanded Windows API. Looking at the 1991 previews of Windows NT, you can see that the NTOS + Win32 is clearly in place, and that enough of Windows 3.0 is running on NT to make an OS. Although all the expanded parts of NT are nowhere to be found (additional subsystems, security, etc). There was a long road to get to 1993 for Windows NT 3.1, although the product wasn't really truly feature complete until Windows NT 3.5 where Microsoft finally had released their own TCP/IP stack, along with major performance enhancements (Daytona!) on the 386 reducing the memory footprint and letting Windows NT truly shine.

Well, first: the TCP/IP stack was "borrowed" from BSD. So it wasn't 100% their own TCP/IP stack to start with...
Second: The great and solid core of Windows NT, which was new and wasn't Windows 3.x from Microsoft, was largely due to Dave Cutler, the head of Windows NT development. He came from DEC and had something like VMS in mind... On top of this microkernel strategy (which NT is NOT since it uses a hybrid kernel; nevertheless portability was a design goal) they put the Win16 and Win32 APIs. All true. BUT...

As I remember it, Windows NT wasn't the big success. The market didn't like it, as it would have meant that all existing programs would have had to be modified for the new Windows, and then they wouldn't have been compatible with the non-NT Windows anymore, but this Windows+DOS was the success story.

How I see it, Windows NT was a dead duck in the water. Just like OS/2. And this was due to lack of third-party programs. But unlike IBM Microsoft was actively encouraging development for NT and making it way easier and cheaper than IBM for OS/2.

And thinking back Windows NT 4.0 was a very good Windows and it should have been clear that this was the future. But see how many more releases of the older Windows 9x were necessary to make the market move: Windows 98, Windows Me. It took until Windows XP in 2003 until it finally worked to move from DOS+Windows to Windows NT.

OS/2 2.0 really suffered with the loss of Microsoft, and IBM had to back-peddle quickly on their Windows stance by creating the Win/OS2 port of Windows to run under OS/2.

Applications... Without third-party applications an OS is doomed. Yes, there are a few niche systems, AmigaOS, RiscOS, ... But in general you need the third-party software for your OS. Apple also knew that, so what they did with Rhapsody, formally OPENSTEP, formally NeXTStep, was to make it compatible to Mac OS. This was actually a step back. But with this API change, which took 4 years to accomplish, Mac OS X was born and it had the NeXTStep API (renamed to Cocoa) and the Macintosh API, or the sustainable subset of it (renamed to Carbon). And see this: until Tiger released mid-00' they had to struggle with applications. Most ran inside the Classic Environment or as carbonized Macintosh applications. It took more than 5 years until finally there were enough applications that finally used the new Cocoa API... But unlike IBM with OS/2, Macintosh users had nowhere else to go to! On the PC the market was open. Windows was the working alternative to the expensive OS/2 that lacked applications. So the users and the market really went into this already working direction.

It's all about the third-party applications really... the eco-system...
IMHO this is why OS/2 failed.

[Hyoenmadan86]

Xenix was a whole different story. Windows/386 was just a step in the door, nothing more. Windows/386 can hardly be compared to Windows NT or OS/2 32-Bit or even Xenix on the 386.

You're clearly underestimating Windows/386 technology. Yes, it can't be compared to NT or Xenix, not because it is inferior to these two, but it is in a very different OS category. Windows/386 has more in common with the modern Hypervisor OSs like ESXi, HyperV or XENServer, than with standalone NT, or *nixes... with the only difference it can use the loaded BIOS/DOS as a driver and compatibility layer inside a VM too, in order to get all the software and hardware support niceties DOS had then.

[mrpijey]

The topic has been split off from

https://www.betaarchive.com/forum/viewt ... 72&t=38395

Stay on topic, or create a new one!

[louisw3]

IBM didn't want Windows anywhere near their OS/2 so Presentation Manager works 180 degrees opposite of Windows.

I don't even know what that means. OS/2 had quite a few similarities with Windows.

Windows 2 & 3 borrow the IBM SAA 'look and feel', OS/2 1.1 was announced and pushed out in early 87, while Windows/368 was end of 87, with the rest of Windows 2 out between 88-89, and Windows 3.0 again takes the appearance of OS/2 1.2. However the API works VERY differently with the biggest difference being the coordinate system, file formats and other methodology.

Just as in 1987 we saw Xenix for the 386, and Windows/386 v2.01 released.

Xenix was a whole different story. Windows/386 was just a step in the door, nothing more. Windows/386 can hardly be compared to Windows NT or OS/2 32-Bit or even Xenix on the 386.

They are totally in the same range, as Xenix was 100% 32bit flat protected mode, Windows/386 also ran in protected mode, where it could provide virtual LIM-EMS (it didn't need anything like EMM386), along with the v86 mode hypervisor. It was far more than a step in the door. The released versions of OS/2 were chained to 286 protected mode, while the OS/2 early betas did feature v86 VM's and multitasking much like Windows/386 (you can even run them online @ pcjs https://www.pcjs.org/disks/pcx86/os2/misc/football/ ). OS/2 was prevented from being offered in both 16bit/32bit much like how Xenix & Windows were sold at that point, until Windows 3.0 which merged the 286 & 386 product into one retail package.

Instead IBM dragged their feet in insisting that OS/2 not just support the 80286, but exclusively support the 80286. The 32bit version of OS/2 was built in parallel with the 16bit version but kept away from end users, a really major mistake.

Maybe. I still think that back in the late 80ies an exclusive 386 operating system wouldn't have been entirely successful. 32-Bit systems were the server systems of that time and they cost a lot of money. Not even to mention that the situation for programs was not ready for 32 bits. Compatibility issues would have killed the operating system before it started. So OS/2 being 16-Bit and DOS compatible makes kind of sense to me.

There is no doubt that a 386 in 1988 + 8MB of RAM + 100MB disk space were INSANELY expensive. OS/2 1.x needed a bunch of RAM as well, and it's MS-DOS compatibility was restricted to a SINGLE MS-DOS 3.3 session. This is where Windows/286 shined by multitasking MS-DOS although being restricted to running it all in real mode, and Windows 3.0 would just page out Windows entirely but you could have several paged out MS-DOS sessions to swap between. The 286 really did stick around for too long, the real killer app that pushed the 386 everywhere really was split between early internet & DooM.

The final straw of course was not only that Windows was excluded from OS/2, but Windows 2.x started to get real traction with 3rd parties writing major apps for Windows. And Windows 3.0 was the final game changer as running Windows on top of a DOS Extender shattered the 640kb dos limit, unlocking the capabilities of the 80286/80386 processor, and best of all it continued to run on MS-DOS, not requiring users to change their OS, or go through the hell of trying to find compatible device drivers.

Interesting. Didn't OS/2 stay 16-Bit at first for precisely that reason? Wasn't OS/2 also DOS compatible? I think it was even later when it was 32-Bit, but more like Windows NT by runnig DOS apps inside a VM, which prevented legacy programs trying direct hardware access from running. That is when they also included WinOS/2...

That said, the primary success was Windows 3.x, not NT 3.x. As I remember NT had exactly this problem: most DOS programs wouldn't run. And there were lots and lots of DOS applications around. What Microsoft did with Windows 95 was genious in this regard: it kept 100% DOS compatibility while providing the new 32-Bit API at the same time! NT and OS/2 couldn't do that.

Windows NT found great success in business. There was plenty of 'power users' clamoring for Windows NT Workstation. It's not secret that Windows NT 3.1 didn't really sell, Microsoft gave away so many copies that I really don't know if anyone actually bought it. On ebay you'll find Windows NT 3.1 workstation boxed copies always showing up with the promotional item stickers on them.

The real big wins for Windows NT were in the areas of file/print sharing, and the new and exciting opportunities of building 3 tier applications as even VB users could now build 'server applications' and run them on a 'real windows server' where before something to be server grade had to be written for Unix, or worse NetWare! Yuck! So instead of picking up a new API and new dev environment you could take your VB program and put it on NT server and now it's a server! And since Microsoft put the networking into NT you could make client sever apps right out of the gate. Just as MS NetPipes run over TCP/IP IPX/SPX NetBEUI etc so Windows NT gave you far more 'bang' for the buck in corporate dev.

Also don't forget that unlike Windows for Workgroups, or even OS/2 of the era, Windows NT supported SMP, even on that one compaq 386 server, along with the new and exciting (and at the end of the day pointless) RISC processors like the MIPS, Dec Alpha and later PowerPC.

It all depends on where you were, a fair number of people I went to school with and worked with ran NT 3.5 workstation as it had fantastic PPP support, LAN support along with the ability to run Mosaic, unlike all the things to load and install to get Windows 3.1 onto the internet with Trumpet, Win32s, Video for Windows, and finally Mosiac. Not to mention being able to run VisualC or any other program language, debug stuff without crashing everything was really nice.

There is no denying it though, if you were primarily running MS-DOS apps, OS/2 did a far better job of multitasking them, but the TCP/IP addon for OS/2 2.0 was of course an addon that most certainly was *NOT* sold retail or stocked by anyone.

OS/2 NT has shown just how versatile and solid the design was, as it survived first the abandonment of the i860 processor after it was shown to be inadequate for performance, then dumping the primary OS/2 2.0 Cruiser API for the expanded Windows API. Looking at the 1991 previews of Windows NT, you can see that the NTOS + Win32 is clearly in place, and that enough of Windows 3.0 is running on NT to make an OS. Although all the expanded parts of NT are nowhere to be found (additional subsystems, security, etc). There was a long road to get to 1993 for Windows NT 3.1, although the product wasn't really truly feature complete until Windows NT 3.5 where Microsoft finally had released their own TCP/IP stack, along with major performance enhancements (Daytona!) on the 386 reducing the memory footprint and letting Windows NT truly shine.

Well, first: the TCP/IP stack was "borrowed" from BSD. So it wasn't 100% their own TCP/IP stack to start with...
Second: The great and solid core of Windows NT, which was new and wasn't Windows 3.x from Microsoft, was largely due to Dave Cutler, the head of Windows NT development. He came from DEC and had something like VMS in mind... On top of this microkernel strategy (which NT is NOT since it uses a hybrid kernel; nevertheless portability was a design goal) they put the Win16 and Win32 APIs. All true. BUT...

As I remember it, Windows NT wasn't the big success. The market didn't like it, as it would have meant that all existing programs would have had to be modified for the new Windows, and then they wouldn't have been compatible with the non-NT Windows anymore, but this Windows+DOS was the success story.

How I see it, Windows NT was a dead duck in the water. Just like OS/2. And this was due to lack of third-party programs. But unlike IBM Microsoft was actively encouraging development for NT and making it way easier and cheaper than IBM for OS/2.

And thinking back Windows NT 4.0 was a very good Windows and it should have been clear that this was the future. But see how many more releases of the older Windows 9x were necessary to make the market move: Windows 98, Windows Me. It took until Windows XP in 2003 until it finally worked to move from DOS+Windows to Windows NT.

The applets source to the FTP client uses the BSD source, and that just shows how good Winsock was/is. Although the Microsoft stack most certainly isn't BSD, all the source leaks would have shown so.

As I mentioned above, Windows NT really targeted a different market. I never put my parents onto Windows NT, not until XP Home was a thing did I move them over to that. Had Neptune been a thing I'd have moved them to that, Windows 2000 was very close to being consumer friendly but it really needed more work.

OS/2 suffered as it's best abilities were running other things. NT was really too busy eating old Unix/Netware installs in corporations and Universities, and promoting people to write for Win32. Basically in the 3.1/3.5/3.51 days you were basically going to DIY or have some kind of server application to run.

OS/2 2.0 really suffered with the loss of Microsoft, and IBM had to back-peddle quickly on their Windows stance by creating the Win/OS2 port of Windows to run under OS/2.

Applications... Without third-party applications an OS is doomed. Yes, there are a few niche systems, AmigaOS, RiscOS, ... But in general you need the third-party software for your OS. Apple also knew that, so what they did with Rhapsody, formally OPENSTEP, formally NeXTStep, was to make it compatible to Mac OS. This was actually a step back. But with this API change, which took 4 years to accomplish, Mac OS X was born and it had the NeXTStep API (renamed to Cocoa) and the Macintosh API, or the sustainable subset of it (renamed to Carbon). And see this: until Tiger released mid-00' they had to struggle with applications. Most ran inside the Classic Environment or as carbonized Macintosh applications. It took more than 5 years until finally there were enough applications that finally used the new Cocoa API... But unlike IBM with OS/2, Macintosh users had nowhere else to go to! On the PC the market was open. Windows was the working alternative to the expensive OS/2 that lacked applications. So the users and the market really went into this already working direction.

It's all about the third-party applications really... the eco-system...
IMHO this is why OS/2 failed.

[/quote]

Yes Rhapsody was doomed as basically except for a hand full of people nobody wrote anything big for NeXTSTEP. Or OPENSTEP. Rhapsody wasn't going to magically change this, OS X's "great idea" was Carbon. In the same way that Win32 is an expanded Win16 API as Windows 3.0 was ported to NT, porting the MacOS API onto Rhapsody was the only way out, as now developers didn't have to re-write the majority of the application. And like Win32s, you could write a Carbon application that'd run on both classical MacOS and OS X giving users a bridge to the new world.

[johnlemon647]

IBM made a lot mistakes with OS/2 Not support Intel 386 in still OS/2 2.00 in 1991 was problem. While Windows 386 2.01 as been running Virtual86 mode though Command Prompt in 1987, While in 1988 OS/2 1.10 and OS/2 1.21 was running Intel 286 Protected mode was problem. IBM wanted Microsoft killed Windows and MS-DOS, Windows 2.00 and MS-DOS 4.00 post be last version Windows and MS-DOS together. That backfired on IBM that made Microsoft more focus Windows and MS-DOS. Windows 3.00 start out as secret stankworks project to push Windows 2.00 Protected Mode this was success. Windows 3.00.14 ISV Alpha Build came out this stankworks project. Windows 3.0 was born. Windows 3.00 was success sell over 4 millions copy put deathnail to OS/2. Windows 95 what killed OS/2 in the end. Microsoft wanted IBM to have 386 version OS/2 from start IBM said no at first. It true that Early OS/2 1.00 Alpha did had some 386 support, but still it take OS/2 couple years to have first 386 version OS/2 that was OS/2 2.0 in 1992. When 386 version OS/2 came off it too little too late for OS/2 to recover, the markets when to Microsoft Windows that was spell end for OS/2.

[louisw3]

IBM made a lot mistakes with OS/2 Not support Intel 386 in still OS/2 2.00 in 1991 was problem. While Windows 386 2.01 as been running Virtual86 mode though Command Prompt in 1987, While in 1988 OS/2 1.10 and OS/2 1.21 was running Intel 286 Protected mode was problem. IBM wanted Microsoft killed Windows and MS-DOS, Windows 2.00 and MS-DOS 4.00 post be last version Windows and MS-DOS together. That backfired on IBM that made Microsoft more focus Windows and MS-DOS. Windows 3.00 start out as secret stankworks project to push Windows 2.00 Protected Mode this was success. Windows 3.00.14 ISV Alpha Build came out this stankworks project. Windows 3.0 was born. Windows 3.00 was success sell over 4 millions copy put deathnail to OS/2. Windows 95 what killed OS/2 in the end. Microsoft wanted IBM to have 386 version OS/2 from start IBM said no at first. It true that Early OS/2 1.00 Alpha did had some 386 support, but still it take OS/2 couple years to have first 386 version OS/2 that was OS/2 2.0 in 1992. When 386 version OS/2 came off it too little too late for OS/2 to recover, the markets when to Microsoft Windows that was spell end for OS/2.

It's 'skunkworks' not stankworks.

https://en.wikipedia.org/wiki/Skunk_Works

Windows 3.0 sold well over 4 million units, the number 4 million is from the first year alone.
http://techland.time.com/2013/05/07/a-b ... 5-present/

And Windows was on the market for 2 years.

OS/2 was already dead by 95. Windows 95 only gaurenteed that IBM was basically done as they weren't going to do a Win32 API.. Linux was eating up a lot of the OS/2 momentium, along with Windows NT. Team OS/2 had basically imploded between OS/2 3.0 & 4.0. By the time Merlin actually shipped it was all done. I remember that out of all the OS/2 fans I knew I was the ONLY one that bought OS/2 4.0. And I used it for less than a week, as the world had moved beyond Win32s, as mentioned and you really needed either NT 3.51 (NewShell made it at least look nice) or yes, Windows 95. And in the world of Linux well it suffered the same fate as OS/2 as any 'big time app' like Real Audio, Netscape, WordPerfect were only available for SCO OpenUnix or UnixWare making things like iBCS manditory. It took quite a while from 1991 to about 1999 for Linux to pick up enough Unix kills to start to get mainstream adoptation for applications. And ultimately Linux stopped the massive tide that was Windows NT on the internet, and in corporations as people fundamentally wanted free (gratis) software, but free software as in freedom is something companies still graple with.

Keep in mind that in 1987 the biggest and most useful 386 feature was v86 mode, as that majority of applications were MS-DOS based. Windows having been written for the 8086 mean that it was a natural fit for the 80286 as it too dealt in 64kb segments. But the tools were there for 32bit, Xenix didn't use the pcc (Portable C Compiler from Bell Labs) but rather Microsoft C & MASM. The only thing really lacking was Microsoft standing up on their own and making the future, and that didn't happen until Windows 2 showed that they could finally get other companies writing for Windows, and I'm sure the sales of Word & Excel sure helped. Being able to get Windows 3 run in protected mode basically had made OS/2 redundant in their minds. Windows 3.11 started what would be the heart of Chicago by moving more and more of MS-DOS functionality into protected mode.

It's really no surprise after moving a million units of Windows 3.0 in the first few months that OS/2 NT was going to be changed into Windows NT. Microsoft had pulled off all the OS/2 devs to work on porting Windows to NT. It just goes to show how impressive the design of Windows NT is where the primary API could be dumped, and all the other infastructure work wasn't lost. That would be catastrauphic for any other OS.

[louisw3]

Here is an interesting interview with Balmer on Windows NT

https://books.google.com.hk/books?id=2j ... &q&f=false

October, 1991 so this also is in the same time as the 1991 CES reveal

[johnlemon647]

Thank Louisw3 I just misspell skunk works. I know what skunk works is. You right about IBM push 286 version OS/2 while Microsoft wanted 386 version OS/2. I Think 386 version OS/2 1.10 with Presentation Manager could work in 1988. OS/2 probably could more famous today with they put GUI on OS/2 1.00 and a loud OS/2 1.00 to have Intel 386 V86 Mode like in OS/2 1.00 Alphas.

[os2fan2]

IBM was committed to supporting existing hardware, rather than just new machines. The requirement to support 286 machines is because there were a lot of 286 machines out there. Machines were still quite expensive.

I don't recall seeing any NT boxes outside a computer show for many years. OS/2 was the mainstay of servers and ATMs. Novell Netware was also used. OS/2 4.0 ran quite well on 16 MB ram, if you disabled the file-system (HPFS) and not too many graphics. You could even use something like BOOTOS2 to make a PM shell with Filebar as the shell in 9 MB hard disk. This was what I used to burn cdroms in at the start.

Windows NT 4.0 struggled to boot on 20 MB ram on a 486, although it did load. 2000 required a pentium, and to install Windows 98 on a 16-MB machine, you need to use 98lite in the install, and then disable IE. You could install IE as a separate download, but the performance was terrible.

The thing with windows NT, well up to W2K, was that if you "don't know what it is, you don't need it". That was the official line. For most of the machines at 1995, NT simply would not load. Even on PEKO, my fancy top-notch box (16 MB ram, 486dx66, 2*120 MB hd), OS/2 2.x and 3.x installed quite well. 4.x was a bit of a struggle. Windows 95 was painful that you ran it from the front boot menu when you needed some win32 proggie to do things. Windows NT 4.0 barely installed (at the time on 1 G disks), and Windows 98SE was a no-go. Most of the time, I used PC-DOS and Win 3.11 (not w/g).

Microsoft used the code from PMEXEC to make progman, it had for many years, the same bugs. If you run pmexec in a Presentation Manager for Windows (3.5 or 4.0) session, then pmexec will populate its windows with all of your explorer groups. Also, progman and explorer listing icons in a vertical list, when programs are dropped on its window, directly corresponds to pmexec's behaviour.

In the years from 1992 to 1995, OS/2 was a viable option. Even well past this, OS/2 could still outperform Windows. When Windows 95 arrived, it more or less killed off the DOS world, at least in the trade press, but companies still plugged away with their DOS 5 / Windows 3.10 dekstops, and OS/2 2.1 in the server room.

The first time i saw windows 95 in the real world, was when it was a client to a netware server. These replaced DOS 5 / Windows 3.1 workstations, and used a lot of power, that they blew the site UPS. These were upgraded to NT 4.0 clients.

When you read the comments here, you get the impression that Windows was great at its release dates. This was not true. Windows 3.1 was released when 4 MB ram and 80 MB hd were common. An install of 20 MB for windows meant it soaked up 1/4 of the hard drive. Ami-pro 3.0 at seven disks is a bigger install than the DOS Multi-mate 4.0 (23 disks of 360k = 6 disks), noting that the latter included its drivers.

The actual load and running was slow. It is not for no reason that we say things like 'slow, dead slow, windows', or 'Windoze' (you could nap off while it booted). Even something like 'ami-slo', you could read the splahscreen as it loaded. Progman was much more restrictive than the DOS menus we used. I mean, my computer came with a dos menu where you could put groups or items at any level. Progman made you use single-level groups for everything.

You read about the story of the explorer task-bar. They showed windows 3.1 to rocket scientists, and got them to make a text document. They just could not do it. It's like that. I was supporting university-educated folk on how to use programs and so forth.

[louisw3]

IBM was committed to supporting existing hardware, rather than just new machines. The requirement to support 286 machines is because there were a lot of 286 machines out there. Machines were still quite expensive.

And that was a massive mistake. Instead we were stuck with 'AT' nonsense until the arrival of PCI! The 286 to the late generations Pentium was a graveyard of preserving a bad decision with the 5170.

I don't recall seeing any NT boxes outside a computer show for many years. OS/2 was the mainstay of servers and ATMs. Novell Netware was also used.

At the bank I worked at we could install as much OS/2 as we wanted with a blanket fee. Guess what OS we loaded for every little thing?

The thing with windows NT, well up to W2K, was that if you "don't know what it is, you don't need it". That was the official line. For most of the machines at 1995, NT simply would not load. Even on PEKO, my fancy top-notch box (16 MB ram, 486dx66, 2*120 MB hd), OS/2 2.x and 3.x installed quite well. 4.x was a bit of a struggle. Windows 95 was painful that you ran it from the front boot menu when you needed some win32 proggie to do things. Windows NT 4.0 barely installed (at the time on 1 G disks), and Windows 98SE was a no-go. Most of the time, I used PC-DOS and Win 3.11 (not w/g).

I guess it must have surprised you that I used NT 3.5 on a 486SX-20 with 8MB of ram then. And it ran with netscape better than using OS/2 + Windows 3.1 + Win32s + Video for Windows + OLE2 for Windows.

People knew they needed NT as it was the 'bigger one' the more 'expensive one' that wasn't as friendly as 95 on the desk, and they KNEW they wanted it in the datacenter as it was the one with networking built in, and the one that could easily connect to the big mainframe host, not just doing 3270 sessions but building APPN and CPI-C apps. Not to mention the databases, the DCOM, and other remoting technologies to bring apps into client server.

In the years from 1992 to 1995, OS/2 was a viable option.

It wasn't. From 1993 onward OS/2 was dead.

When Windows 95 arrived, it more or less killed off the DOS world, at least in the trade press, but companies still plugged away with their DOS 5 / Windows 3.10 dekstops, and OS/2 2.1 in the server room.

I guess that is the server room you were used to. The one I worked in, I was removing RS/6000's Dec MIPS, Dynix, AT&T 3B2's and all kinds of NetWare & OS/2 machines, which were just backed up wiped and re-installed with NT.

When you read the comments here, you get the impression that Windows was great at its release dates.

That is because it was true.

You read about the story of the explorer task-bar. They showed windows 3.1 to rocket scientists, and got them to make a text document. They just could not do it. It's like that. I was supporting university-educated folk on how to use programs and so forth.

Do I have to dig out the studies on the OS/2 2.0 interface, as the most bewildering interface of all time? It was a brilliant move by IBM to push MS out, so that they could demonstrate once and for all that the only thing they were capable of was a crazed UI that was far too easy to destroy, and all that time making WinOS/2 instead of letting MS put Windows directly onto OS/2 as they had wanted back in the 80's.

The 16bit driver model, and reliance on Microsoft C 5.10/386 to compile OS/2 says it all.

[os2fan2]

It is only a 'massive mistake' to support older hardware, until you start looking at the LOB stuff. What makes or breaks a sale will ultimately does it run our aging proggie. Where I worked, i had to debug a lump of software written in GW Basic, largely to make the thing y2k compliant. They were running this stuff in a DOS box in Windows 3.1, 95, and then NT4.

It doesn't supprise me that you ran nt 3.5 on a 20 MB ram 486. I could install 4.0 on a similar box, and i imagine that 3.5 would have been a less weighty thing, given the same box ran OS/2 3.x and 4.0. The key difference here is that I could install half of OS/2 above the 504 barrier, whereas none of windows nt OS could be so installed. The best it offered was to be installed on hdb.

The third-largest class of Windows 3.1 software was replacements for progman. You read the software manuals from that era, and they talk about 'if you have NDW do ...'. Microsoft did not have the answer in Progman. IBM abandoned Progman (PMExec) for their WPShell in OS/2 2.0. It had a number of quite nice features, but it was pretty frail, because it was all in ea's.

The actual model of the WPS shell: desktop icons and folders, with extensions, were copied directly into explorer. The separate taskbar was copied from Apple's shell, and already had existed in the third-party world of the OS/2-Windows environment. [Filebar for OS/2 implements most of the explorer taskbar in a text-menu, including a running-tasks menu]. It's not like explorer is a 'de nova' work.

Windows and the press may well have been forward-looking, but they pushed it out as an upgrade for older machines too. This was essential for their 'windows everywhere', but it caused a good amount of greif. The software was typically much larger, that it was called 'bloatware' (because it took too much hardware to drive), and booting Windows 3.1 on a machine with 4 MB ram, though possible, was simply painful. (It wasn't called 'Windoze' for no reason).

It was mainly successful because Microsoft engaged in a large number of practices for which they were brought to heel by the various courts.

Windows itself was not all that hard to destroy. My sister's computer was largely destroyed by a 'drag and drop' of the Windows 9x directory into some second directory.

Explorer had a lot of nice things in it. But it's hardly a patch on WPS. Each learnt from the ommissions of the other. Microsoft caused a lot of end-user pain by not honouring old software interfaces. The solution is the thousand-subdirectory thing SxS, where they might run lots of different versions of the same DLL. The user end calls it 'DLL-Hell'.

There was a version of Netscape for OS/2. You did not need to install Windows to run it. The whole point of the internet is that different systems could use it, and that Microsoft sought to hijack it with MSJAVA and a whole lot of MS-only stuff, gives the appearence that Windows was better than other systems. It was more to do with them hijacking the standards, rather than innate ability.

Microsoft used their dominance in the 16-bit market (Windows/DOS), to push their particular 32-bit agenda. It's not improbable that better 32-bit drivers etc can be written for OS/2, it was more the case that even by the time of OS/2 4.x, the 32-bit model had not really taken off.

OS/2 did pretty well in the trade press, despite Microsoft's opposition. You should remember that even today's CMD.EXE has a lot of OS/2 features in it. The one sorely missed is EXTPROC. But KEYS is there, as is DPATH. Neither are used in Windows, but OS/2 has them.

Edlin is a bit-identical file from MS-DOS 5.0. kb16.com is keyb.com from 5.0, there are a few others from that DOS too. Even something like QBASIC is a DOS 5.0 version. But if you ask Windows to make you a dos boot disk, you end up with a DOS 8.0 disk. The fact is that until BartPE started making the WinPE environment available, you had to use a DOS disk to install Windows, or several windows diskettes. Windows can not boot from floppies. You get the VMS substrate from the boot disks, which can't run Windows.

But we were doing things in OS/2 ten or so years before Windows. Booting from cdroms, putting a mini-partition in, were things i did on my 486.

[xelloss]

Microsoft used their dominance in the 16-bit market (Windows/DOS), to push their particular 32-bit agenda.

And it worked pretty well. Perhaps IBM did make some mistakes, otherwise they wouldn't have ended out of the PC business, both software and hardware, that they created.

[louisw3]

Microsoft used their dominance in the 16-bit market (Windows/DOS), to push their particular 32-bit agenda.

And it worked pretty well. Perhaps IBM did make some mistakes, otherwise they wouldn't have ended out of the PC business, both software and hardware, that they created.

It was inevitable. IBM wasn't going to market anything that would threaten AS/400 or RS/6000 sales, nor would they put out products to go head to head with small companies after that whole monopoly thing.

[jbyrneiu]

OS/2 was the best DOS multitasker at the time. We used it to run multiple DOS applications due to its smaller footprint than Windows NT. I was the author of DOS2OS2 which was/is a device driver for OS/2 that allowed separate DOS and OS/2 windows/sessions to "talk" to one another through OS/2 message queues and shared memory. I'll post it to this site for you OS/2 fans!

[jbyrneiu]

A major problem with developing OS/2 apps was that the GUI only supported one input queue for the entire system whereas Windows provided one queue per application. This meant that OS/2 would freeze up the entire system if one misbehaved application corrupted or locked the input queue. Windows would only freeze the offending application but the rest of the GUI would still be responsive. Later versions of OS/2 alleviated the problem but by then Windows NT had the large part of x86 operating system market share.