Twenty-five years ago, on March 31st, 1992, IBM released OS/2 2.0, the first mass-market 32-bit PC operating system. The road to OS/2 2.0 was quite long and winding, and the OS was a proud member of the vaporware club (just like, say, Windows NT or Windows 95).
Talk of 32-bit OS/2 goes almost as far back as OS/2 itself. In 1987, Gordon Letwin wrote about OS/2-386 in Inside OS/2 and it is clear that future 32-bit extension was a consideration when OS/2 1.0 was on the drawing board. And we now know that in late 1986, Microsoft already started experimenting with using the 80386’s V86 mode to run Virtual DOS Machines (VDMs) on top of OS/2.
1991 OS/2 2.0 6.167 (left) and 1992 GA (right)
Microsoft launched actual development of OS/2 2.0 probably in 1988, the same year development of NT OS/2 (later known as Windows NT) started. The goal was to provide a PC operating system which is upwardly compatible with OS/2 1.x, supports native 32-bit applications, and supports VDMs; the latter was especially important in light of the fact that compatibility with existing DOS applications was a weak point of OS/2 1.x.
The entire history of OS/2 2.0 development is one of missed deadlines and broken promises. In July 1988, OS/2 2.0 was supposed to be released in early 1989. In February 1989, an OS/2 2.0 SDK was supposed to be released by the end of the year and the finished product in 1990.
Microsoft did ship the OS/2 2.0 SDK by the end of 1989—just. On the last business day of 1989, the $2,600 MS OS/2 2.0 SDK became available.
There is no known surviving copy of a MS OS/2 2.0 SDK. The only known fragment is the development tool set (C compiler, assembler, OS/2 headers and libraries, electronic documentation) from Summer 1990:
The only known survivors of MS OS/2 2.0 SDK
The disks aren’t even original—they are copies, made back in 1990 at a certain U.S. Airline which signed up for the SDK program. The disks were donated to the OS/2 Museum about 15 years ago and since then, only parts of the MS OS/2 2.0 SDK printed documentation have been sighted, nothing else.
In the second half of 1990, things fell apart. The tensions between Microsoft and IBM became unsustainable, especially after the runaway success of Microsoft’s Windows 3.0. Irreconcilable differences were followed by a divorce, and the companies parted ways. On September 19, 1990, IBM announced that it is taking over OS/2 development, including OS/2 2.0.
By January 1991, IBM was promising OS/2 2.0 release in March 1992, with free upgrades for purchasers of IBM’s OS/2 1.3. IBM decided that that deadline was going to stick, come hell or high water. And it did.
January 1991 ad for IBM OS/2 2.0
Not surprisingly, IBM’s OS/2 2.0 was a bit of a rush job. Up to 1990, IBM was concentrating on the development of OS/2 1.3, with Microsoft working on OS/2 2.0. When Microsoft quit, it took IBM many months to get up to speed with OS/2 2.0 development.
IBM also had a serious problem in that with Microsoft gone, a 32-bit C compiler (cl386) was also gone. IBM Toronto rapidly put together a 32-bit compiler for OS/2 2.0, released in 1992 as CSet/2 1.0. IBM also aggressively courted third-party compiler vendors and announced agreements to support OS/2 with Borland (May 16, 1991) and Watcom (August 7, 1991).
The most controversial and most heralded feature of IBM’s OS/2 2.0 was the Workplace Shell (WPS), a modern object-oriented user interface. The WPS was not present in MS OS/2 2.0 SDK and not even the early IBM pre-releases of OS/2 2.0 (e.g. Level 6.149 from July 1991); those used the same simplistic shell as OS/2 1.3.
The WPS was very nifty and impressive, but it was also not entirely stable (especially the use of INI files was error prone) and it consumed a fair amount of memory. That didn’t help when memory was already expensive and OS/2 2.0 required 4 MB to run at all, and 6-8 MB to run acceptably well. In hindsight, jamming WPS into OS/2 2.0 (rather than some later, more stable release) seems like an unhappy decision.
Timeline
The following is a timeline of OS/2 2.0 betas, whatever they were called at the time—SDK, pre-release, Early Experience Program, and more. No complete official information exists but enough can be pieced together from contemporary press reports and newsgroup discussions. The dates may be inaccurate and where known, refer to the label printed on the disks. Users typically received the disks with a slight delay.
- 6.43, 12/89, first MS OS/2 SDK
- 6.78, 7/90
- IBM took over OS/2 development in 9/90
- 6.123, 2/91, last MS OS/2 SDK, also shipped by IBM
- 6.149, 7/91, first IBM OS/2 2.0
- 6.167 (Comdex preview), 10/91, included WPS
- 6.177 aka 2.0 LA (Limited Availability), 12/91
- 6.304, 02/92
- 6.307 aka 2.0 GA, 3/92, final OS/2 2.0
The information is primarily based on contemporary newsgroup and press reports, there is no official information. Based on the surviving reports, it is unlikely that another OS/2 2.0 pre-release would have existed.
Betas, GA
A big OS/2 2.0 push started at Fall Comdex 1991. For that occasion, IBM put together OS/2 2.0 build level 6.167, also known as Comdex beta. OS/2 2.0 6.167 was reasonably close to the finished product. It sported the Workplace Shell, fully supported VDMs, full-screen Win-OS/2 sessions, and came with Boot Manager.
OS/2 2.0 build 6.167 disks
OS/2 2.0 6.167 was shipped in a cardboard box with a modicum of printed documentation. It was clearly a pre-release product but not just a hastily produced beta with labels from a dot matrix printer.
OS/2 2.0 6.167 documentation and license
OS/2 2.0 6.167 was a good preview of the final (GA, or General Availability) release. It included all major OS/2 2.0 functionality with the exception of seamless Win-OS/2. The boot-up screen did not look quite like OS/2 2.0 GA:
OS/2 2.0 6.167 booting
There was a nice-looking graphical installer, a clear improvement over OS/2 1.x:
OS/2 2.0 6.167 installer
The installed product looked a lot like the finished product, with the familiar WorkPlace Shell:
OS/2 2.0 6.167 desktop
One clear visual difference was the program/folder icon on the left of the title bar, removed in the final version.
Full-screen Win-OS/2 was in place, of course supporting Windows 3.0—Windows 3.1 was released shortly after OS/2 2.0.
A full-screen Win-OS/2 session
All in all, in October 1991 OS/2 2.0 was still five months away, but its shape could be very clearly seen.
But Wait, There’s More
OS/2 2.0 6.167 of course wasn’t the only pre-release of OS/2 2.0 (see timeline above). Some other disks survived as well. For example build 6.304 from February 1992, just before the final release:
OS/2 2.0 build 6.304 disks
Or the older and somewhat mysterious build 6.605:
OS/2 2.0 build 6.605 disks
Why mysterious? Because the 6.605 “level” does not fit into the sequence at all (the final OS/2 2.0 GA was 6.307). Also because the disks are labeled 7/91 but the files on the disks are dated September 1991. It appears that 6.605 was a minor update to 6.149 but the out-of-sequence numbering is not understood.
At any rate, the 6.605 and the newer 6.304 disks sets in OS/2 Museum’s possession are both incomplete, and incomplete enough that even a most basic installation can’t be put together.
However… not long before this article was published, hopefully complete disk sets (at least the base OS) of build 6.605 as well as 6.177 from the end of 1991 turned up. As of this writing, they are still in transit. Watch this space.
I really wish a comprehensive post on the entire OS/2 2.0 fiasco can be written.
A Quarter Century of OS/2! I was hoping the MS OS/2 2.0 had magically surfaced, but it appears not.
So great to read some article about OS/2 again, I thought this blog had changed to a hardware museum.
I joined OS/2 fandom when 6.177 came out. I ordered the full set via a german distributor and received around 60 disks.
Later I joined Team OS/2 and the developer program which was great because it included lots of free software with the subscription, e.g. TCP/IP and lots of internal tools and documentation.
I used OS/2 until a terminal harddisk crash early in 2001 (ironically an IBM Pegasus 1 GB disk), and it served me well as a foundation for a BBS with 2 ISDN and 2 analog lines. Most of the software was self-written with emx+gcc, and I wrote lots of papers during my time at university with emTeX.
The most glorious single point in time was when I could replace my last piece of DOS software with something native and go PROTECTONLY.
Boot Manager possibly appeared in 6.149, called “Multiple Operating System Tool” back then – we’ll know once the disks have landed.
Maybe in another 25 years 🙂 For example those Multitasking DOS 4.0 disks showed up after about 28 years. So there’s hope.
Without hardware there’s no software 🙂 And Xenix isn’t hardware, although it is not OS/2 either. What is interesting is that at least Gordon Letwin clearly knew Xenix well and used it as inspiration when designing OS/2.
So what was on those 60 disks, and more importantly, where are they? 🙂 I assume there was the OS itself, the Toolkit, some kind of compiler? Any networking bits? Extended Services?
I used emTeX myself in the late 1990s. And I didn’t realize it at first but emx+gcc was one of the first 32-bit OS/2 compilers.
Lots of BBSs ran on top of OS/2 because it was a way to operate a BBS without using a dedicated system, OS/2’s multi-tasking took care of it.
One important anti-trust exhibit for example is http://www.groklaw.net/pdf/iowa/www.iowaconsumercase.org/011607/0000/PX00307.pdf
You will soon be able to verify the following.
The old os2.misc listings showed 6.605 as using the OS/2 1.3 interface.
GA code listed as 6.307 had a number of revisions indicated with suffixed letters; at least C, D, E, and N are mentioned.
One thing I had forgotten was that the 1991 betas were offered through a BBS. 9600 baud which meant a 5 hour download for the entire 18 MB. Any error on the download required redoing the whole thing. Later on, the download was redesigned to be broken up into smaller chunks before IBM canceled the entire idea of downloading OS/2 betas.
Yes, 6.605 used the OS/2 1.x user interface. It looks like OS/2 1.3 with a slightly different color scheme. It’s from September 1991 and must have been the last beta without WPS. It also still uses the LE format for 32-bit executables so it can’t run any OS/2 2.0 32-bit executables.
The OS/2 successor ArcaOS 5.0 just missed releasing almost exactly 25 years after OS/2 2.0 – https://www.arcanoae.com/arcaos-5-0-launch-on-hold-for-a-few-more-days/
Not sure what the market is for an OS/2 successor in 2017 though.
P.S. today is the anniversary of another IBM creation that ultimately wasn’t very successful: The original PS/2 line was released 30 years ago today on April 2, 1987.
Yep, 30 years of PS/2 today. I think Microchannel lasted about 10 years, and the PS/2 brand less than that. A shame, really.
Portions of the PS/2 still live on though…. VGA, PS/2 style mouse, and keyboard ports (still available on new motherboards). While the industry shunned Microchannel, the other IBM backed standards the machines came with were quickly adopted.
VGA, PS/2 keyboard/mouse connectors, EBDA and a few BIOS functions are all that remains of the PS/2 now though the EBDA didn’t seem necessary since there was unused space in the BDA (beginning at 40h:ACh) and segment 50h was almost completely unused.
MCA was definitely advanced for the time. It was killed by IBM’s foolish attempt to forcibly dominate the market.
I thought MCA only lasted about 5 years. Were there any PS/2 systems released after 1992 ? I can’t find an authoritative list of PS/2 models other than the IBM PS/2 and PC BIOS Interface Technical Reference from Sept 1991. I remember by 1994 when I was at IBM they were already back to ISA systems with the ValuePoint series.
There were a few models released after 1992. The vast majority were 486SLC machines that likely retailed for much more then other vendor’s full 32-bit 486 machines of the time (at least the Reply sourced Model 53 planer got local bus video). OS/2 2.1x was a pre-install option too.
http://ps-2.kev009.com/psref/ps2book.pdf
Microchannel didn’t go away after the PS/2 died, there were a few machines (PC730, PC750) with the “select a bus” option of MCA or ISA. The machines had swappable risers that contained a PCI-to-ISA bridge or PCI-to-MCA (!) bridge.
http://ps-2.kev009.com/pcpartnerinfo/ctstips/ee2e.htm
I think the IBM PC 700 Series machines were the last MCA systems. As far as I can tell, the last IBM PC 750 with MCA support was introduced in October 1995 (Pentium 133).
The PC Server 520 was introduced in June 1996 with the 8641-MD0 having 6 MCA slots and 2 PCI slots. Curiously, the MCA version was offered with Pentium 133 and 32 MB while the EISA equipped counterparts all had Pentium 166s and 64 MB. I think there was one last MCA model released in 1997 but a quick search doesn’t turn up a matching announcement.
The PS/2 lineup had one major marketing blunder. IBM frequently pointed out how the design made automated manufacturing easy and reduced costs. IBM’s sales prices did not reflect that. Never, never, brag about extreme margins.
I said in another thread that even Ed Roberts mentioned in http://www.virtualaltair.com/virtualaltair.com/mits0011.asp that “We should have copyrighted the name or patented the bus”.
How long was OS/2 used for in embedded systems (e.g. ATMs, POS) ? It seems like it was used through the early 2000s and then largely replaced by Windows XP Embedded.
Until at least the mid-2000s. Possibly later. IBM stopped selling OS/2 in 2006 but there was still some maintenance going on after that, with at least two unreleased FixPacks (FP5/FP6) that weren’t available to all customers. From my own experience I know that in 2005, new OS/2 systems were still being rolled out.
Thanks for looking that up, so in servers MCA survived even a bit longer. It is then more or less accurate to say that the PS/2 brand lasted 5 years and MCA 10 years. Probably lasted longer than EISA 🙂
I really liked OS/2 2.0. I came to it from Windows 3.0/3.1, and it was a massive improvement on them. I don’t actually remember when I got it–my Dad bought it for me sometime around the 1992-1995 timeframe–it was definitely before Windows 95 came out – I was around the age of 10–13. Sometime later my Dad also bought me the Hobbes CDROM, around 1994-1995 I think, and I used that to install EMX, and started using bash and Emacs and GCC and so on. Next thing you know, I was installing Slackware, and left OS/2 behind. But EMX for OS/2 was an important waystation on my adolescent journey from DOS/Windows to Linux.
I was recently interested to discover that Workplace Shell originally was not going to be shipped in OS/2 2.0, but in OfficeVision/2 LAN. https://books.google.com/books?id=fBIEAAAAMBAJ&pg=PA51 Does anyone know anything about this? (And how much does OfficeVision/2 have in common with OfficeVision/MVS, OfficeVision/VM and OfficeVision/400 – something? Nothing but the name?)
I’d also be interested if anyone knows anything about Workplace Shell for DOS, which I believe was cancelled before release. Was it true WPS, with SOM etc, or just some improved DOSSHELL with the “Workplace Shell” label stuck on it?
I have had a quick look at Workplace Shell for Windows, which was IBM employee-developed software, never a product. I get the impression it is more of something that looks like WPS on the surface but isn’t really inside. I haven’t tried running it yet, but I had a look at the source code, and I can’t see any signs of SOM in it.
The Micro channel based PC Servers were withdrawn in February 1997 (https://www-01.ibm.com/common/ssi/rep_ca/5/897/ENUS997-025/). The last PC Server 520 system was introduced in November 1996: the IBM PC Server System/390. In 1997 there were some updated RS/6000 models: 7012-397, 7015-R50 (perhaps the last one until August 2000).
Workplace Shell for DOS was developed by some contractors for IBM. It would have looked like the OS/2 version but no API so it would have been a glorified DOSSHELL (which was highly reviled). Only problem was it didn’t work at all, crashed on the simplest of operations (like Alt-H for help) and was written by idiots who didn’t understand DOS or x86 assembly language and barely understood C. This steaming pile of crap literately had several hundreds of bugs opened against it (approaching a thousand) and half of the DOS department working on it until it was finally canceled because it was completely unsalvageable. Fortunately for me I was working on the core DOS kernel and utilities and largely avoided it though I had to do an analysis of the code (on a weekend). It was easily the worst code I have ever seen in my career and my recommendation was to get Norton Utilities Wipeinfo and delete it all. The whole idea was stupid considering that most users hated DOSSHELL but clueless upper management wanted this thing, probably because some corporate customers wanted it since they really didn’t respond to the general consumer market at all.
@Simon K: If you install OfficeVision/2, you will notice that it adds a few object oriented extensions to OS/2 1.x Presentation Manager like the printer icon will print documents dragged to it. It was part of a strategy to make it easy to add access to mainframe resources to every application. Thus, an OfficeVision supporting word processor could be used to read email instead of having mainframe email only be available through a terminal emulator. WPS spread that feature set throughout OS/2 so every OS/2 32-bit program would be OfficeVision ready once the OfficeVision client was installed. HP had a similar strategy of providing an object oriented user interface to unify all product lines under the name “New Wave.”
WPS for DOS was shown off at a few trade shows. It looked a lot like WPS OS/2 and was intended to be object oriented GUI that would run some WPS OS/2 applications (after a recompile). I think it foundered on the same issues that kill most light products: how to include enough of the API to support applications written for the high end platform without becoming just as big as the high end platform. I don’t know if it was related to the PM for DOS that showed up underpinning the Japan exclusive Webboy browser. IBM had too many incompatible but overlapping projects in the 90s.
WPS for Windows header files shows that a lot of hoped for functionality was not implemented by the time development ceased. I think the goal was to get 90% of the WPS API implemented in a simpler fashion.
Pingback: OS/2 2.0, Summer ’91 Edition | OS/2 Museum
Notice that PX00307 is about “PM vs. Windows” as it was about API calls only, which was the wrong way to make the decision.
Pingback: Weekend tech reading: Hybrid jet announced, Uber banned in Italy, hacker archetypes, gaming on wired vs wireless – Tech Reviews Blog
Michal Necasek Didn’t I used to work with you at CaseWorks in Dunwoody?
Back in 1990? or so???
A couple of things occurred to me while comparing 6.167 to GA: Win-OS/2 was devoid of any IBM branding in 6.167 and didn’t include Adobe Type Manager, and the “Lighthouse” wallpaper introduced in 6.167 was deprived of the American flag in GA.
Definitely not — I wasn’t even in the US at that time 🙂
Pingback: Kurzmeldungen 2017-03 März - WarpServer.de
Thanks for this comprehensive review, it’s enlightening!
I remember to have evaluated two OS/2 2.0 beta releases in 1991,
but I had no idea how many revisions had been released in total.
The graphics engine was (already) significantly different from
OS/2 1.3’s one. After installing a graphics driver for a Nanao
TIGA display adapter, instead of obtaining a vast PM desktop at
1664×1200 pixels, I just got an “exception in SingleQ” trap.
After OS/2 2.0 GA’s release, a cute employee at IBM direct at
Laatzen (close to the Hanover fair with it’s CeBIT hall) sold
me his personal OS/2 1.3 EE package. The TIGA driver worked, and
PM’s Desktop Manager isn’t that bad, but the 1.3 architecture
wasn’t an option (Nanao’s monitors were superior to their
graphics adapters, and I changed the latter’s model soon).
One of the OS/2 2.0 beta diskette sets became corrupted, i.e.
I ran into “sector not found” errors and tried to apply PC Tools’
“revitalize disk” function on it, but realized only later that
the diskette drive was defective (the culprit) and the media
had had been okay … alas, no longer.
My interest in OS/2 2.0 betas grew again when I needed a slim
WPS replacement on a maintenance system. OS/2 1.3 EE’s Desktop
Manager is compact and can run programs (a screen saver) on
startup. While a Warp4-based system with latest fixes applied
boots flawlessly with 1.3’s Desktop Manager as shell replacement
(RUNWORKPLACE=PMEXEC.EXE), it crashes with SYS3175 as soon as a
program icon is selected in the shell (POPUPLOG.OS2 available).
Does someone have an idea whether OS/2 6.149’s PMEXEC.EXE might
do a better job, and where to obtain the binaries (PMEXEC, PMCPL
and PMFILE)? As mentioned before, there had been an issue with
my own diskettes. Yes, Desktop Manager was a nice interface!
All the best Torsten