IBM OS/2 for Windows 3.1
Posted by FR_fink-roselieve@reddit | vintagecomputing | View on Reddit | 88 comments
My grandfather had this.
Posted by FR_fink-roselieve@reddit | vintagecomputing | View on Reddit | 88 comments
My grandfather had this.
JasonMckin@reddit
So this is a perfect example of IBM's remarkable incompetence in the early 90s.
Their belief was that OS/2 was a replacement for DOS, not Windows. This wasn't a totally dumb belief, because classic Windows was always designed to run as an app on top of a DOS subsystem.
So if users used to run Windows on top of DOS, I think it was IBM's belief that there would be users who would run Windows on top of Workplace Shell on top of OS/2 subsystem. For a while, I think they even marketed this architecture as a "better way to run Windows" It was somewhat legit because Windows apps would crash less often thanks to OS/2’s memory protection and multitasking.
Now, the goofy question is how does IBM ship Microsoft's code as part of this package? Microsoft was more than willing to license the code out and let them do a "replace Windows with Win-OS/2" on their own. But that created the awkward situation where users who already owned Windows might be “buying Windows again” as part of OS/2.
So their next genius chess move was to ship a cheaper version of OS/2 that had hooks into Windows code, but didn't actually ship the Windows itself. So that way, it would just detect the existing Windows and give you the whole enchilada in a different way.
Embedding Windows into OS/2 at all was one of the greatest decisions of dumbassry in the technology industry. They gave ISVs zero reason to actually write native OS/2 graphic mode applications. So the entire app ecosystem started building towards the Windows 3.x platform.
Meanwhile, Microsoft releases Windows NT, a true 32-bit competitor to OS/2. The original was fairly basic and mirrored with Win 3.x user interface.
Then Microsoft releases Windows 95 and just nukes the market with the Windows 95+ / MS-DOS 7.X platform for the next couple years. Maybe the handful of businesses that cared about the security and performance of OS/2 would buy it, but the rest of the universe was aligned to Win95/DOS7. This permanently damaged the prospects of OS/2 Warp 3.x and Warp 4.x.
NT4.0 brought in the 95 interface and by the introduction of Windows XP, it was chess mate for the OS industry (barring rise of personal and enterprise Linux & barring mobile).
Candidly, I miss a lot of the OS/2 Presentation Manager interface. It took a long time to fully embed everything in Windows and I'm not even sure they ever embedded everything. There was something more satisfying about firing up an [OS/2 Command Prompt] than an Windows CMD> session.
If think you about how Google built advertising partnerships with high traffic websites in the early days or how Apple worked with carriers, the one universal lesson is that great code isn't enough, you have to partner intelligently to win chess matches.
PuttingFishOnJupiter@reddit
Windows did not run on DOS. It just needed any compatible DOS to boot up. It would then adjust the interrupt vector table to point to it's own routines, and, assuming at least 386 enhanced mode, would swap the redundant DOS code out. Being self booting is not a requirement for an OS.
mallardtheduck@reddit
It's way more complicated than that. DOS functions are absolutely used by Windows after boot.
Windows 3.x had no filesystem or storage device drivers (*) of its own. All file I/O was performed by DOS. That alone is a pretty massive part of an "OS"; it's the majority of MS-DOS's functionality.
Also, the 16-bit Windows API is "missing" quite a lot of functionality (e.g. time and date functions), requiring applications to call the MS-DOS APIs for such things directly.
Even Windows 9x can use DOS to access storage devices, although it will use its own drivers if possible.
* Windows 3.1 had "32-bit disk access" which optionally replaced the BIOS hard disk access routines with 32-bit code. The driver that ships with the OS is very "particular" and does not work on >504MB or non-IDE drives. It's still DOS calling what it "thinks" is the BIOS though. Window 3.11 introduced "32-bit file access" which did finally bring the FAT filesystem implementation into protected mode (the functionality was actually a backport from then-in-development Windows 95). Note that both of these were only for hard drives; floppy drives, CD-ROMs, etc. all still relied on MS-DOS.
Artistic-Branch7393@reddit
That's a bit misleading, all of the I/O in 3.1 goes through the API in base 3.1 and as you noted VFAT in 3.11 completely bypasses DOS at the driver level. What are you even trying to argue here
mallardtheduck@reddit
That DOS is much more than a "bootloader" for Windows. Especially pre-95. It cannot run without the MS-DOS "kernel" remaining fully active. While "self booting" may not be a "requirement" for something to be called an "OS", not relying on another OS for something as fundamental as file I/O surely is.
Sure, Windows provides its own file API, if it didn't it wouldn't be able to ensure files are closed if a program ends without doing so. It's still calling DOS to do the actual work, at least for most users (even under 3.11 VFAT was often not enabled; documentation suggests it was only enabled by Setup if you installed certain network components, otherwise you had to do so manually).
Artistic-Branch7393@reddit
First of all on your definition of file I/O somehow being a requirement of a legitimate OS, you are inventing that post-hoc and citing nothing to back it up, presumably so you can try and reach thec onclusion you want here. Nobody at all sees that as a requirement of an operating system.
Second of all DOS is not "much" (with the italics) more than a bootloader for Windows 3.1. When Windows 3.1 starts up, WIN386 runs at ring 0 and demotes DOS to a virtualized, preemptively multitasked guest.
As for your argument that it calls DOS to do the "actual work", when an app in Windows calls the file API the path goes through 3.1's KRNL386.EXE, which issues an INT 21h afterwards. That interrupt is then intercepted by DOS' VxD layer, specifically IFSMgr inside of WIN386.EXE, before it reaches the (virtualized) instance of DOS. The VxD controlls that interrupt. It can either route it to DOS from there or route it to the virtualized guest, and even if it routes to DOS it is executed as a virtual machine under WIN386.EXE's supervision
mallardtheduck@reddit
I don't see any citations from you either...
"Nobody" thinks managing the filesystem is a core part of an OS's job? Really?
How's this for a citation:
Or this:
Or this (obviously written for a less technical audience):
Sure, there are historic and specialised operating systems that don't feature a filesystem at all, but even they don't require another OS to do it for them (although it is, of course, possible for an application running on such an OS to implement a filesystem that the OS itself knows nothing about).
The exact technical details of the "routing" of filesystem calls vary between versions and configuations of Windows, but basically it goes:
Application -> Windows -> DOS (usually) -> BIOS (usually) -> Hardware
Whether the DOS/BIOS code is running in genuine Real Mode or within a V86 Mode virtual environment is a fairly minor detail, as are things like the names of Windows modules and the interrupt numbers used. Running an OS inside a virtual environment doesn't make it stop being an OS.
Artistic-Branch7393@reddit
As for your citations even granting that because arguing such semantics would turn into a red herring, you're confusing using DOS' services being used in any way by 3.1 with 3.1 being controlled by it. DOS which you claim is so much more than a bootloader for Windows cannot preempt Windows, or reclaim memory from it or terminate one of its processes while Windows can do any of these to DOS. Linux makes BIOS calls during init, so would you say that makes it less of an OS? Of course not. It calls INT 21h while being far more privileged than DOS is at runtime
As for your point that everything I mentioned is a fairly minor detail, no it isn't, not at all, and I was using them to establish DOS' relationship with Windows which is far more subdued in its nature than you read it to be. If you wish for me to simplify it, Windows basically creates and preemptively multitasks virtual DOS machines after taking over. The sequence is as follows;
WIN is typed > WIN . COM executes and takes the CPU into protected mode > KRNL386.EXE and WIN386.EXE take control > DOS is virtualized under VMs that Windows preemptively multitasks.
As for your point about historic and specialized operating systems I don't see how those are relevant here, as with applications running an OS being able to implement different filesystems the OS knows nothing about unless you're trying to poke at 3.1 being that, which I've already established is false
mallardtheduck@reddit
Can you point out where I said this? That's not what I intended to say at all. I'm saying that Windows (mostly pre-95) absolutely depends on DOS for core functionality and cannot run without it. DOS is and OS. DOS + Windows together can be called an OS. Windows alone is only part of an OS.
I don't believe it does... The fact that the exact same kernel image can be used for both BIOS and UEFI boot suggests otherwise. The bootloader (usually GRUB) calls the firmware (i.e. the BIOS when using "legacy" boot) to load the kernel and initramfs; those contain everything the system needs to continue the boot without firmware assistance. About the only case I know of where BIOS services might be used by a running Linux system is when using VBE for graphics (either for X or the kernel framebuffer).
Anyway, even then, the BIOS was designed as an interface for the OS. It's very clearly not an OS itself. UEFI on the other hand is almost a 32/64-bit "DOS", it fits pretty much any common definition of an "OS", but ceases to be active once a "full" OS calls "ExitBootServices" early in the boot process.
Again, that's not what I intended to say. I said things like the names of the various modules and the interrupt numbers are minor details. Running DOS/BIOS routines in V86 mode under a "host" of some sort wasn't at all uncommon on DOS; EMM386 (or QEMM or 386MAX or any of the similar memory managers) did that too, albeit without multitasking.
With the exception of the pre-emptive multitasking (and the names of the modules), that's basically what happens when any protected-mode DOS program is run(*). Is DOOM an OS (early releases did display the message "DOOM Operating System" during startup after all)? What about other environments that support pre-emptive multitasking of DOS sessions (e.g. DESQview)?
* Depending VCPI/DPMI host (and Windows is one of those too...) and configuration the system may use V86 mode to call DOS/BIOS routines or may revert to genuine Real mode. I believe the use of V86 mode was most common, but even if not, it's at least common enough for Windows use of it to not to be too novel.
You stated that having a filesystem wasn't a requirement for something to be an "OS". That was my attempt to acknowledge that; it's possible to have an OS that has no filesystem. It's difficult to argue that a system/program/environment that depends on another OS for filesystem access that it can't run without can be called an "OS" on its own.
As I said previously, the filesystem implementation is the majority of MS-DOS. As stated above, I have no issue with suggesting that DOS + Windows can be considered an OS. Windows alone, notsomuch.
Artistic-Branch7393@reddit
As for your comparison to DOOM, no, that's not a good comparison because DOOM does not do any of the things 3.1 does, it does not manage virtual memory or run a VxD layer at ring 0, neither does it preemptively multitask virtual machines or host cooperative multitasking. You've also moved here from "it had no file I/O of its own, all of its I/O was performed by DOS" to "Windows + DOS is an OS" of which would still require DOS to be doing anything meaningful here
DESQview did not do any of these things either, it explicitly ran as a DOS application running in real mode so your comparison there is also a bit faulty.
Your comparison to UEFI actually works as an example of the relationship between 3.1 and DOS quite well. That analogy practically describes what Windows does do the latter when KRNL386 and WIN386.EXE take over.
As for your remaining position about 3.1's filesystem, as I mentioned the INT 21h calls from Windows applications go through WIN386.EXE before they reach any of Windows' virtualized and preemptively multitasked DOS machines. The DOS machines handling all of these calls are machines that Windows created and can manage, schedule, and terminate. It's like saying a hypervisor depends on its guest OS just because it occasionally routes its operations through it. 3.11 also removes this, which you seemingly posit as a major part of your argument here, from the question entirely of which your only rebuttal seemed to be it having low adoption rates, which is completely irrelevant when discussing the capabilities of an OS.
As for your concession that DOS + Windows could be considered an OS, the work DOS and Windows do even within that is too heavily lopsided for DOS to be posited on equal ground with Windows as it provides everything I mentioned before while DOS serves as a bootloader and compatibility environment.
mallardtheduck@reddit
"A VxD layer" is a Windows-specific term. VxDs are a specific type of Windows driver module. Of course no other application/system has those (although OS/2 has some quite similar concepts). Since it's a DOS program, it naturally does run entirely in "ring 0" (unless running under a more "advanced" OS such as OS/2 or Windows 9x). I did say "without multitasking"; although the way DOOM's "thinker" state machines operate could be considered a form of co-operative multitasking/multithreading and the asset loading/caching layer kinda-sorta looks a bit like "virtual memory" if you squint.
While DESQview could run purely in real-mode on pre-386 systems, on a 386-or-later it's a protected-mode "supervisor" running multiple V86-mode DOS sessions (at least in later versions). Since we're ignoring how Windows worked in real/standard mode, we can ignore how DESQview worked on older systems too.
Except that UEFI provides no run-time services to the OS. Once "ExitBootServices" is called, that's it. DOS continues to provide services to Windows that Windows cannot run without.
Yep, that's what my simplified flow of "Application -> Windows -> DOS (usually) -> BIOS (usually) -> Hardware" also stated. Adding specific module names like "WIN386.EXE" and interrupt numbers doesn't contradict that.
Except that the "root VM" under "WIN386" is actually running the entire Windows 3.x GUI, as well as all the file I/O from Windows applications. It cannot be terminated without terminating Windows itself.
The overall architecture for Windows 3.x (in enhanced mode) was basically the same as Windows/386 2.x. Windows loads the "supervisor" (WIN386) then immediately starts a DOS VM to run what is effectively just "Standard mode" Windows. Additional VMs can be started at any time to run DOS applications and these can be terminated.
The only significant architectural difference between Windows/386 2.x and 3.x is that 3.x added support for running 16-bit protected mode (and Windows itself was ported to that mode) and even 32-bit protected mode(*) applications within its DOS VMs.
* Of course, initially there was no way for 32-bit code to call Windows APIs as 32-bit support was intended only for DOS applications using DPMI. It was quickly realised that there was nothing preventing applications from switching the "root VM" into 32-bit mode (thanks to co-operative multitasking between Windows applications) and third-party (e.g. Watcom Win386) and later Microsoft's own "Win32s" libraries enabled 32-bit Windows applications on 3.x.
Note that this is one of the things that made Windows 3.x driver development (particularly video driver development) quite difficult, since you needed both a ".DRV" driver than ran in that "root VM" (or accessed the hardware directly if Windows was running in Standard mode) as well as a "VxD" (usually named ".386") driver that virtualised the hardware between multiple VMs (one of which being the Windows GUI running the ".DRV" driver). Of course, in the latter days of Windows 3.x it wasn't uncommon for developers to just put more-or-less everything in the VxD and make the ".DRV" just a wrapper, dropping support for Standard mode.
This architecture is actually pretty reminiscent of what we have today; Windows 11 (usually) runs as a VM within Microsoft's hypervisor and additional VMs can be started/stopped (e.g. by WSL or Hyper-V) at any time... And the same "what is an OS?" question comes along when you realise that nearly all I/O (graphics, disk, keyboard, etc.) has to be routed through the "root" Windows 11 VM since that's what runs the drivers that talk to the actual hardware. That's one rabbit hole I'm not going any further down today.
Optionally and only for FAT16 (and maybe FAT12) hard drives. Since Setup often didn't enable it by default, floppy disks were common CD-ROMs were at least somewhat common in the era and nowadays it's completely possible to run Windows 3.x under versions of DOS with FAT32 support (and even at the time "odd" setups, like running Windows from a network drive, existed). Real Windows 3.11 installs still commonly rely on DOS for quite a lot of file I/O.
It might be irrelevant if we were solely talking about the "capabilities of an OS", but we're not. We're (or at least I am) talking about the actual real-life operation of the OS. Windows 3.11 was capable of bypassing DOS for file I/O in certain circumstances, absolutely, but it was also capable of "falling back" to DOS (even Windows 9x could do that) and in common scenarios (e.g. saving a Word document to a floppy disk, launching Encarta 94 from CD-ROM, etc.) had no other option.
The HAL (or HALs; although I don't think there's been more than one on a "normal" version of NT for a good while now) is a core part of NT. Ships with NT. Cannot run independently. Is not an OS itself. NT does not depend on another OS.
Picking apart an OS and saying "it's not an OS without this bit" makes no sense. You can do that for any OS.
Not sure what you're getting at there. That's not new for Windows 2000 (perhaps you're confusing the use of INT 13h with the use of "extended" INT 13h that was possibly new for W2K). Of course the bootloader uses BIOS services, that's pretty much the sole reason they still existed after DOS went out of common use. All(*) bootloaders that support BIOS-style boot (including the previously mentioned GRUB) work that way. Point is; the BIOS is not an OS. The bootloader is not an OS. NT does not depend on another OS.
* While it's theoretically possible to write a BIOS-boot bootloader that accesses the hardware directly and never uses BIOS services, it would be tied to very specific hardware, so nobody (that I know of) does that.
That's not what I said; I said the dependency (at runtime) on another OS makes Windows (3.x) not a full OS in itself.
Artistic-Branch7393@reddit
As for your point on DESQview, that is correct but V86 sessions DESQview runs alone are still not enough to produce 3.1's architecture which further supports my point.
As for your point on the system vm that's somewhat sophisticated of a point but it not being terminatable by Windows does not mean it governs Windows. Win386 still governs it, the system VM can't modify its page tables or access the PC's hardware without VxD. Also, Microsoft chose to make the VM unterminatable so it's not really a constraint imposed on Windows as much as an architectural choice.
And if we grant your point then by your logic Hyper-V is also hosted by the root partition it depends on
I don't think the point you made about DOOM is serious so I'll move on from it without addressig it, you could've just said I was right there rather than trying to salvage it by saying it's kinda-sorta cooperative multitasking
On the UEFI providing no runtime services, even the services you say DOS provides at runtime are intercepted by WIN386 before DOS sees them. In simple terms Windows decides whether it wants to involve DOS. It is more windows delegating to a guest under its control rather than DOS providing and hosting services.
As for your realworld usage point, we are discussing the architecture here of the OS, I never agreed on talking about "actual real-life operation"
As for your point on NT's HAL, here you've alreay stated Windows manages its own DOS instance. DOS being shipped seperately and marketed as an OS doesn't change the relationship at runtime, WIN386 has DOS under its control regardless of that. Windows 3.1 was compatible with PC-DOS and (almost didn't) DR-DOS. If it were such a dependency it would need to be a controlled component, yet it is completely interchangable between vendors
Artistic-Branch7393@reddit
Yes, the "3.1 was just an operating environment/shell/etc" is very persistent and also completely wrong, it is an OS and more of one than 9x
JasonMckin@reddit
Okay, but can we agree there is no way for win.exe to load into memory without config.sys and command.com - unlike the case for Windows NT?
PuttingFishOnJupiter@reddit
You don't need cpmmand.com either! If you put
shell = c:\windows\win.com
...that'll go straight to Windows. But..DOS prompt work so probably not worth mucking around with!
JasonMckin@reddit
Really?? Are you sure about this? Even Windows 95 shipped with DOS 7.0 and a command.com. Why ship it if it wasn't necessary?
How do you get long file names in Windows 95 OSR2 without FAT32 in DOS 7.1?
PuttingFishOnJupiter@reddit
DOS is still there. Command.com is just a text based user interface. MS-DOS.SYS is still loaded as usual.
Without command.com you can't interact with DOS. Windows itself doesn't need it though. It's there for compatability reasons.
JasonMckin@reddit
Okay...this is a layer of subtlety that isn't totally clear:
"Windows did not run on DOS."
"DOS is still there."
"It just needed any compatible DOS to boot up."
"MS-DOS.SYS is still loaded as usual."
evert@reddit
Command.com is to DOS what Bash is to the Linux kernel. Command.com us certainly part of DOS but it's also a program running on the DOS kernel.
PuttingFishOnJupiter@reddit
Yeah, it's a bit odd!
PuttingFishOnJupiter@reddit
Boot process:
Read boot sector and load bootstrap code. Use that to loaf io.sys Load msdos.sys (other DOSes are okay too) Load device drivers according to config.sys
At this point the normally route is: Run command.com, implicitly passing autoexec.bat as a parameter. Command then executes each line in autoexec.bat. Then drops you at a DOS prompt (which is really command.com)
You now type "win".
win.com loads into nemory, modifies the valies in the interrupt vector table such that any calls to the "dos interrupt" (21h) are nown handled by windows, rather than the aforementioned ms-dos.sys
When you exit windows, the original vectors are restored and you are back at command.com.
Does that make it clearer?
Few-Subject-8142@reddit
I ran it back in the day, so I could keep my BBS running, and still do school work. :)
Few-Subject-8142@reddit
I have an OS/2 box just like I'm this. OS/2 Warp 4.
seeingeyefrog@reddit
I loved many things about the workplace shell.
Font and color palettes made customizing windows and apps a breeze. You could even edit your own icons.
And you could even multitask with dos and windows.
It was so ahead of its time, and then deliberately crushed by Microsoft.
JasonMckin@reddit
I think you could technically edit icons in Windows, but you had to install a separate app to do it, they didn’t package the editor with Windows like IBM did with os/2.
I would argue IBM deliberately crushed itself but bundling Windows as Win-OS/2.
Microsoft hacked 8.3 before NTFS with FAT32. It wasn’t the same, but you could use long file names in classic Windows as of 1996.
I don’t think NTFS was based on HPFS. I thought the Dave Cutler team rebuilt it from scratch? And IBM never built a work around like FAT32.
itsasnowconemachine@reddit
NTFS came out in 1993 with Windows NT 3.1
JasonMckin@reddit
Fair enough. So classic windows users were “stuck” relative to os/2 for a couple years.
Orbitalsp3@reddit
Yep https://www.youtube.com/watch?v=NAUr6zyVjKg
Quints_beercan@reddit
Os/2 ran the majority of atm machines for many years
Navydevildoc@reddit
Plus almost every Voicemail system in the 90s and 2000s.
khooke@reddit
I worked in a office in the early 2000s that had a pc sat in the corner of a storage room running the office voicemail on os/2
PuttingFishOnJupiter@reddit
Yep, that's about right. It just sits there doing its job. Until a hardware failure occurs and suddenly you are being asked "why didn't this have a failsafe or backup" and they get really grumpy when you tell them:
And after all that it's somehow the fault of the IT department.
Me? Bitter?
PuttingFishOnJupiter@reddit
Also private PBXs. It did it well.
Impossible-Driver69@reddit
Nothing beat OS/2 security. IBM based it off mainframe security which is bulletproof.
McLayan@reddit
IBM mainframes were not very secure, it was rather that very few people knew the platform. The only people who know about mainframe architecture tend to be the people who make money keeping ancient businesses software alive and live in this walled garden whete IBM is the only authority on what is good design.
There was a huge breach of IBM mainframes operated by Logica for the Swedish government about 14 years ago. A security consultant gave a speech about the bugs that allowed it and how poorly security was implemented on z/OS back then (e.g. using improper password hashing with algorithms long known to be too weak)
RevolutionaryRush717@reddit
Didn't use RACF or couldn't afford ACF2?
gnntech@reddit
It was the preemptive multitasking that was based on mainframe concepts. Security meaning processes are not able to step on other processes.
OS/2 didn't really have any kind of user security until later on and even that was not anything I would call secure.
Right_Stage_8167@reddit
Sorry, but there was no actual security features. It was more like MS-Dos for protected mode.
LiqvidNyquist@reddit
Also some important live train routing & safety software for one of the big railways in Canada, accrding to one of my programmer buddies who worked for them back in the 90s.
Just_to_rebut@reddit
An OS for an OS? I don’t understand…
LousyMeatStew@reddit
OS/2 had a feature called Win-OS/2 which allowed you to run Windows programs within OS/2. The problem was that they did this by licensing a full-fat copy of Windows from Microsoft and that was included in the cost of the software. So they made a version of OS/2 that shipped without a Windows license that required the user to have their own copy of Windows pre-installed.
Just_to_rebut@reddit
>that required the user to have their own copy of Windows
That’s interesting, so Windows 3.1 had to have had API (?) to allow OS/2 to run programs with it’s kernel but appear in the OS/2 GUI?
Thanks for explaining. I had it completely backwards. I thought it was a version of OS/2 that would run inside the Windows interface.
LousyMeatStew@reddit
No, the opposite. OS/2 natively supported DOS, so it ran Windows as a DOS app. Win-OS/2 was the translation layer that allowed for stuff like graphical acceleration, clipboard integration, etc. Closest analogy might be the way OS X would run OS9 in Classic Mode.
Windows was running in standard mode so if one app crashed, it still crashed Windows but OS/2 itself was unaffected.
One cool thing you could do was run multiple copies of Windows simultaneously. If you did it this way, you could run e.g. Word and Excel and OS/2 would take care of stuff like clipboard integration and if Word crashed, it wouldn’t take Excel down with it. But since you were running two copies of Windows simultaneously, this required way more memory.
Just_to_rebut@reddit
Oh… thanks for technical (well, at a level I can understand) explanation. You really made me understand why this was a useful thing back then.
CrankyD@reddit
OS/2 had a subsystem for Windows 3.1 allowing it to run the Windows 3.1 Program Manager and programs. Similar to how Windows 11 can run Linux programs using the Windows Subsystem for Linux. Basically just a built in virtual machine.
CrankyD@reddit
OS/2 could run DOS programs natively, and Windows 3.1 was just a DOS shell so as far as Windows was concerned it was just running on DOS.
JasonMckin@reddit
They didn’t drop it, the market dropped it for them.
Warp 4 still ran Win16 apps, but by 1996, nobody gave a shit, since Windows 95 supported a basic version of Win32 and ISVs were rewriting apps to that.
So the genius idea of Win-OS/2 literally only had any value from about 1992-1995.
LousyMeatStew@reddit
It’s easy to look at any single vendor like IBM and say “well, they clearly shot themselves in the foot” in retrospect but Microsoft literally owned 90s computing. The 90s began with Windows 3.0 (which allowed Microsoft to take over the consumer market) and ended with NT 4 getting DoD C2 certification (which was the beginning of the end for commercial UNIX).
OS/2 still had value for businesses because line-of-business apps and internally developed apps could linger on as Win16 versions for years. But that niche was killed off not by Win95 but NT 4 which came out in 1996.
And a lot of this was due to the way Microsoft conducted business. Win16 became significant because Win 3 was when Microsoft started bundling licenses and Microsoft just refused to license Win32 to IBM.
JasonMckin@reddit
Microsoft owned the 90s because IBM clearly shot themselves in the foot.
They couldn't build hardware to save themselves.
They couldn't market software to save themselves.
They did bupkis to go after the home consumer market.
LousyMeatStew@reddit
IBM didn’t need saving in the 90s. They had their own niche and were doing fine. ThinkPads, RS/6000 and System/390 were what IBM was doing in the 90s.
The idea of IBM’s consumer products being unsuccessful is largely a myth told at IBM’s expense. IBM was #2 worldwide behind Compaq by the mid-90s in worldwide shipments in the consumer market.
JasonMckin@reddit
Exactly, more PCs for consumers to install Windows on!
bramblefish@reddit
I had a dual boot OS/2 & Windows. Neither worked very well.
psteve_m@reddit
I could never figure out why someone would want half an OS.
itsasnowconemachine@reddit
Half an operating system: The triumph and tragedy of OS/2
Timbit42@reddit
Uses half as much CPU, RAM and disk space. This means it runs twice as fast, can run twice as much, and takes half the space away from your own files.
vabello@reddit
I used to run this in late 1994 into 1995 until Windows 95 came out. I didn’t have the one that included Windows 3.1, but you could supply your own, which I did. After getting better serial driver, it was amazing being able to multitask a DOS terminal program downloading a program from a BBS in the background while playing Doom.
Financial-Brief-1038@reddit
I've actually installed Win 3.11 from a stack of floppies before.
evert@reddit
How else would you do it?
JasonMckin@reddit
Exactly, that was the only way 🤔
geewronglee@reddit
Maybe not? I had the original Creative Labs multimedia upgrade kit and it included Windows 3.1 on a CD-ROM. I recall also buying OS/2 on a promotion where they gave me a rebate on Windows 95. A friend and I tried OS/2 that summer but come fall I realized OS/2 was not the future and went back to Windows.
JasonMckin@reddit
It wasn’t Windows 3.1. It was probably a service pack of multimedia utilities that didn’t come with the base Windows 3.1.
Thr promotion was probably from your retailer, not from IBM. OS/2 cannot run Windows 95 Win32 apps. You would have had to dual boot two separate environments.
geewronglee@reddit
I had these things and they were real. It was a full Windows 3.1 and people sell them on eBay. I agree the rebate might have been a Micro Center promotion but am I pretty sure it was an IBM promotion. You are correct about Win32 and thus why I moved on from OS/2.
JasonMckin@reddit
Hmm, I guess it’s possible that CL was bundling Win 3.1 and maybe some multimedia packs too?
It was a crazy time for bundling for sure.
CrasVox@reddit
Could not stand installing or working with OS/2. Such a mess of a system compared to WinNT 3
mnlx@reddit
Idk why they're downvoting you. It has some pretty cool aspects but it's also a complete mess for consumers. Installation isn't guaranteed, updates are cryptic, it takes too long to boot. Presentation Manager is not a professionally designed GUI, more like a tech demo, the file manager is simply abject. There's barely any comforts, system messages give you codes to look things up elsewhere, it has interfaces for IBM system administrators, not users. And where's the software for it? There's very little compared with other contemporary options.
I've been looking into it for its unique DOS virtualisation capabilities, but the more I interact with it the more I understand it couldn't win. It appears to be solid, if uninspiring, at what it does beyond virtualizing DOS and Win16, I get the industry use, but it's a dead end. Windows 3.1x and 9X weren't stable at all, but they were made for consumers first, they were cheap, very usable, they ran on the lowest specced hardware, they had drivers for everything and loads of software available. NT was a way more advanced and capable OS from the get-go.
JasonMckin@reddit
What consumers? Average consumers weren’t buying PCs in 1992.
The genius of the start menu UI, getting Jay Leno to launch Windows 95, etc was that Microsoft took the PC to average consumers.
IBMs moves were short-sighted, but not irrational.
mnlx@reddit
I was using a Mac with System 6 and Multifinder at home in 1990 and writing in glorious black on white WYSIWYG using Word 3, why would I buy a personal way (and way more expensive if interesting enough) PC? That experience was the future and it already had been around for a while, the writing was on the wall.
So of course my own first PC came with Windows 95 OEM, and then I really had to learn DOS, which I had avoided mostly in college, such a letdown.
khooke@reddit
PM was based on years of design, concepts and UI standards developed at IBM called (if I remember right) CUI (Common User Interface). There was an IBM Red Book that was an inch thick manual that documented the UI conventions.
My understanding was it was rooted in standards developed for 3270 and 5250 terminal based systems, so although there was consistency across systems, this probably held back (depending on your point of view) PM as it felt clunky because of this history.
mnlx@reddit
I was about to edit it, I meant Workplace Shell. I mix them up all the time.
caliman123@reddit
Anyone else remember lotus notes, ran on meaty os/2.. did Novell back then, thought it was completely badass and notes was revolutionary!!
khooke@reddit
I worked for 2 companies in early 2000s that were still using Notes as their email. I rather liked the document storage feature that supported team collaboration… ahead of the game for the time.
JasonMckin@reddit
Every non-Microsoft office product (WordPerfect/Wordstar, Lotus 1-2-3/Quattro Pro, dbase/Paradox, etc) was ahead of its game.
Microsoft arguably held the industry back in the 1990s in pursuit of monopoly power.
rayhaque@reddit
I still remember the TV ads for OS/2 Warp. It was pretty good stuff at the time! It was far more stable than the shit Microsoft was dumping out.
JasonMckin@reddit
Better DOS than DOS, better Windows than Windows!
aardvark_xray@reddit
Best screen saver
QuirkyImage@reddit
Used as my daily driver through college before uni
eusty@reddit
We still use OS/2 on an old machine for PLC programming.
The only thing which annoys me is the menu system... seems like Close and Exit and where it is depends on the application
restlessmonkey@reddit
I have an unopened box of OS/2.
LiqvidNyquist@reddit
I used this for many many years until I finally had to give it up because it just wouldn't properly run all the windows CAD and EDA applications I needed for my job. It crashed less and I got really into programming for the "OS/2 workplace shell" and GUI which was kind of like a parallel but different version of the win32 api. Memories of spending hours feeding floppies into my PC to install or reinstall though!
turtleisinnocent@reddit
It came with a free version of Dragon Dictate, which retailed for a lot of money. You had to train it to your own voce, and in Spanish you'd read the first parts of El Quijote de la Mancha, which is a lot of fun. Win 3.1 binaries ran well, but by then Windows was already 32 bit so it was dated.
laffer1@reddit
I’ve got warp 3 and 4. Bought warp 3 at babbages on clearance back then
Orbitalsp3@reddit
I've never used it, but apparently it had way better multitasking https://www.youtube.com/watch?v=NAUr6zyVjKg
laffer1@reddit
It’s still supported today. You can buy arcaos. It has aptitude package manager with Firefox and several gnu apps plus os/2 warp apps and some newer drivers
Timbit42@reddit
It has pre-emptive multi-tasking, like Windows NT, Mac OS (v10 and up), and Linux do. The Amiga had it before OS/2 came out, but without memory protection.
ZestycloseAd2895@reddit
And voicemail systems.
satsugene@reddit
Yeah, I supported one. OS/2 was fairly uncommon and new to us when we did a new deployment (~2000 or so).
It was our only Oracle instance too.
The only “normal” thing about that server (for us) was Dell hardware.
el_esteban@reddit
I saw OS/2 once at a community college radio station, circa 2002. Actually, that's the one and only time I encountered it "in the wild".
cchaven1965@reddit
I put this on my first self-built PC. Loved it.