Over the last few weeks, two “new” pre-releases of OS/2 2.0 have been found in ancient warez archives.
The first is OS/2 2.0 build 6.64, released in early April 1990:
In general, this build is not substantially different from the MS OS/2 2.0 Pre-Release 2 (build 6.78) from June 1990.
The READ.ME file on the installation floppy opens with the following:
README for OS/2 SE Version 2.0 IBM CONFIDENTIAL
Driver 6.64 - shipped 04/06/90
Driver 6.64 is the third scheduled monthly update of OS/2 Standard Edition
Version 2.0. This is an update of the BASE Operating System, the ToolKit
and the Languages.
Sadly only the base OS has been preserved, not any of the development tools.
As the READ.ME file suggests, this pre-release was part of some kind of closed beta program run by IBM which shipped monthly updates.
It is unclear if IBM even distributed these builds on floppies; it is quite likely that some or most customers received floppy images electronically. If IBM did ship these betas on floppies, they were likely fairly nondescript disks, not intended for any kind of wide distribution.
Build 6.64 is as of this writing (June 2024) the oldest surviving pre-release of OS/2 2.0.
MS OS/2 2.0 PR3
From the other end of the spectrum is Microsoft OS/2 2.0 Pre-Release 3, build 6.123. This pre-release is from February 1991 and it was the last (third) Microsoft beta of OS/2 2.0.
Build 6.123 has been seen before, in a curious packaged IBM release from April 1991. But that box only contained the base OS, not any of the development tools.
The warezed archive of Pre-Release 3 includes the entire disk set, very similar in content to the recently archived MS OS/2 2.0 Pre-Release 2: Base operating system, Toolkit, Languages (Microsoft C and MASM), networking, and debug support.
Several disks (notably the installation floppy) were archived in the form of TeleDisk images, easy to convert into raw floppy images. However, most disks were stored simply as ZIP archives; this makes it possible to restore fully functional but not quite original floppies.
One obvious change from Pre-Release 2 was a graphical Toolkit installer which set up the entire development environment including the 32-bit C compiler and assembler:
Once the Toolkit is installed, everything is in place: Development tools, sample programs, online documentation (in both QuickHelp and INF format).
The installer sets up all the necessary environment variables in CONFIG.SYS, creates desktop groups, and after a reboot, the development system is ready to go.
As is apparent, the look and feel of this build doesn’t quite look like anything else. It somewhat resembles Windows 3.0, but doesn’t really look like either OS/2 1.2/1.3 or like the released OS/2 2.0.
Here’s the development system in action:
A 32-bit debugger (CodeView) was included as well:
I believe the Microsoft C386 compiler version 6.00.054 is the version that IBM kept shipping as part of the OS/2 DDK until the end of OS/2.
The development system in Pre-Release 3 appears significantly more polished than Pre-Release 2, while the base OS doesn’t really look all that different.
The networking components (LAN Manager client) do not install cleanly, just like in the PR2 beta; however, this time a README file told users what to do—rename several networking related DLLs in the \OS2\DLL directory, so that the networking setup is not confused into thinking that networking is installed already.
Unfortunately I was not able to get the networking going, unlike the PR2 beta. All attempts ended up in the OS crashing. Whether the networking actually worked in the PR3 beta is an open question.
All in all, this is quite a find—a year ago we had no MS OS/2 2.0 pre-releases, and now we have two out of three (Pre-Release 2 from June 1990 and Pre-Release 3 from February 1991). I’m sure there is more out there.
I was just putting the parts on pr3 to write up. The cl386 does appear to be the one that was in the IBM DDKs from 1994. It’s got a newer libc as well although I haven’t seen if it thunks.
I was too tired last night I did a strings and pr3 had 6.123 in there although I need to redo the teledisk.
I’m curious if pr3 cl386 is close enough to 1991 NT cl386, and if it can compile GCC.
Like the last beta that was shown, one of the striking things about this is just how complete it is. It also runs acceptably in 4MB of RAM, unlike OS/2 2.0 LA or GA; I think we can blame the Workplace Shell for that.
If they would have pushed this OS over the finish line (which would not have been particularly difficult), this would have been a completely usable higher-end OS, and this came out a solid two years before OS/2 2.0 GA did. It’s a shame Microsoft didn’t go with this instead of putting all their eggs in the NT basket; Microsoft didn’t have a relevant 32-bit operating system until 1995.
In contrast, the earliest NT betas in mid-1991 are barely functional (and relied on an OS/2 development system) and chewed up a lot more RAM.
The new very early Windows 95 (Chicago) beta that just come out has a somewhat similar visual appearance to this (and by “new” I mean “ Windows 95 20 July 1992 build”), but oddly enough, not identical.
The difference was clearly that in ’91, NT was a new OS being busily written from scratch. You’re right that the early pre-releases are very, very rough around the edges and clearly nowhere near to being finished.
In ’91 OS/2 1.x already had ~5 years out in the field behind it and despite everything, it was a solid OS. While adding 32-bit addressing and paging significantly changed the OS/2 kernel, it at the same time kept enough in place (notably I/O) that it wasn’t an entirely new thing.
It really is a shame that MS + IBM didn’t release OS/2 2.0 in spring or summer ’91, basically using what they had at the time. But I suspect Microsoft was basically dead set against it, refusing to help in any way, which meant among other things no compiler (and IBM had to scramble to get a new one).
I am fairly certain it was the other way around. MS wanted to release OS/2 2.0 ASAP while IBM wanted to wait until after finishing 1.3 and moving 2.0 to the unifying strategy of the WPS. MS wasn’t going to release software partnered with IBM that deviated much from IBM’s release; that had already been tried with a lack of success.
IBM’s decision to delay OS/2 2.0 two to three years probably prevented it from having any chance of getting a solid market position. By the time OS/2 2.0 got through its initial problems, Win 3 had been well established and the excitement for the potential of Win 95 was ramping up.