Shining some light on the history of OS/2, Win9x and NT | Hacker News Reader

Shining some light on the history of OS/2, Win9x and NT

liam-on-linux.livejournal.com

54 pointsstargrave24 comments

Mostly wrong IMO. NT is more like VMS rewritten + pieces of MS code integrated (or rewritten), certainly not a "portable OS/2". The OS/2 parts were confined in the subsystem and are now extinct. As for the IBM did not know how to do OSes opinion, that's both kind of a joke and meaningless: in gigantic organizations, there is not a single mind and single centralized knowledge about X or Y. It bowl down to smaller teams, or even sub-teams. And if there were, "IBM" has done plenty of OSes, and some of them have been in service and backward compat for way longer than anything MS did, ever.

NT is a new design by the people who were involved in VMS. Even when they didn't replicate everything, they had already acquired "taste" and "experience" so some solutions of some problems are very similar, to the point that knowing some detail in one OS gives you a good knowledge about the basics of the detail in another. But this is only in some aspects. There are a lot of important solutions that were made for the first time only in the NT.

The team developing NT had the requirements to fulfill that were never put in front of them, and their result was at the end something in many aspects fundamentally better than any competition. There are of course some aspects that the competition did better too, but that was not enough to win the market.

Whoever really wants to learn more about how it was made, including a lot of details about the humans involved, should read the very good written book

"Show Stopper!: The Breakneck Race to Create Windows NT and the Next Generation at Microsoft" (1994) by G. Pascal Zachary

> Whoever really wants to learn more about how it was made, including a lot of details about the humans involved, should read the very good written book "Show Stopper!...

I can't recommend this book enough. It's a very interesting story and gave me an appreciation for what it takes to do big things in the software industry.

Yes - IBM was doing virtual machines back in the mid 1980s (I was using them on IBM's VM/CMS), long before such things appeared on unix-like or windows OSs.

The amount of cool, ground-breaking stuff that IBM has done is severely under-estimated. I wouldn't count OS/2 as a part of that, however.

> I wouldn't count OS/2 as a part of that, however.

OS/2 1.3 came out at the beginning of 1990 and OS/2 2.x came out 1992.

By comparison Windows 95 came out in 1995.

So you need to compare OS/2 to MS-DOS, Windows 3.x and the Apple Macintosh systems of that time and by that comparison OS/2 was very good.

In the early 90's I did a lot of development for OS/2, MS-DOS and Windows 3.x and OS/2 had the big advantage that a bug in your code would only crash your program where asd for MS-DOS and Windows 3.x it crashed the OS and required a full reboot.

The fatal mistake IBM made was to hobble OS/2 2.x to the 80286 processor.

By staying with the 80286 processor it meant OS/2 ran really struggled to run MS-DOS and Windows programs and that meant businesses just stayed with MS-DOS and Windows.

If they had just gone with the 80386 processor (like they did 2 years later with OS/2 3.X) it would have meant the end of MS-DOS and Windows 3.x only because that later version of OS/2 ran Windows and MS-DOS programs better than Windows 95 which was still a year way.

You're getting your versions mixed up there. It's OS/2 1.x that ran on the 286. 2.x and later required a 386.

I just checked Wikipedia and you are correct.

It was the OS/2 1.3 release that should have been targeted at the 386 chip.

NTFS is a evolution of OS/2 filesystem.

it's also an evolution of ODS-2 filesystem, so...

Not only did I live through this era of operating systems, I’m also a voracious reader of the history of tech companies and have read dozens of books on Microsoft and IBM history, and yet never once have I heard many of the claims in this article. I’d take it with a grain of salt if it were me.

I found the claim that OS/2’s difficulty to run in a VM lead to Virtualbox pretty fascinating. I can’t find any direct sources for this claim at a cursory glance but it does seem that Virtualbox supported OS/2 quite early on. Interestingly, it seems that Virtualbox began life in 2007, which makes it confusing if true; why were people so interested in virtualizing OS/2 in 2007?

>why were people so interested in virtualizing OS/2 in 2007?

I'd suspect that driver support negating for new hardware when it came to replacing aging hardware in area's in which it was still used as an OS, played a big part. So banks, that bridge to the mainframe and other niche area's in which they had a system that ran on that OS and never replaced.

I thought it was Parallels that was created in order to be able to run OS/2 in a VM - not VirtualBox.

I recall seeing ATMs running OS/2 until quite recently.

I could imagine that the hardware running that software was getting quite long in the tooth, and that a virtualised platform would be a potential stop gap measure in extending the service life of that investment.

Now I see a good number of embedded windows ATMs (that tell tale click sound is always a dead give away). I assume that the majority of the OS/2 ATM Fleet is long since gone.

> why were people so interested in virtualizing OS/2 in 2007?

A lot of PLCs used OS/2 software to debug / reconfigure them, so that’s one possibility other than the ones mentioned, but I still need evidence that:

A) OS/2 depended on Ring 1

B) VirtualBox supported it better than alternatives

C) OS/2 had anything else to do with VirtualBox’s creation.

Indeed, no mention of https://en.wikipedia.org/wiki/LAN_Manager let alone how NT initially also supported the Alpha (Dec) processor as well as x86.

As for OS/2 rings, not seeing ring 1 used here: https://en.wikipedia.org/wiki/Protection_ring#Implementation...

I also lived thru all this and kinda interesting seeing how word of mouth chains have panned out over time. Makes you wonder about how history will remember and has remembered some things.

NT supported x86, MIPS and Alpha in the mid 90s from the single source code

> DEC cancelled it. Cutler and his team left and went to Microsoft. MS didn't know what to do with this élite group so it gave them OS/2 3 to finish.

Based on the Showstoppers book -- it's more like: Bill Gates knew the days of Windows 3.x were limited, and needed a new, next generation operating system.

Gordon Bell convinced Bill Gates to call David Cutler, of VMS/DEC fame (Bill Gates was always impressed with DEC as an engineering company).

Cutler was like, "what, come work for that dinky company that makes crappy Office docs and a shitty OS? Psscht."

Bill persisted, saying it'll be a brand new OS that he'll lead. Cutler insisted only if he got to bring his DEC team (like, 200+ people) and basically got whatever he wanted. Bill obliged.

NT development started in '89.

Such a pivotal moment in computing history. If I ever got to chat to Bill Gates, I'd ask him what he was more proud of: a) getting Cutler on board for NT, or b) orchestrating the backward-compat mindset that set up the Windows 3.11 -> Windows 95 -> Windows NT -> Windows XP transition.

I would say Windows 3.1 -> Windows NT + Windows 95/98 -> Windows 2000.

The transformation from the sausage factory of the indestructible, but inflexible, NT and the fragile set of hacks which was 95 into 2000 environment was the real story in all this.

This combined with to migrating from Netbios and NT Domains to tcp/ip and active directory was really the basis for everything they've done in the 21st century. They did more in those 6 years than they've done in the 20 years since.

It is hard to pull an accurate history from a couple decades ago. All I know is that when I worked at IBM in those days, a few coders told me that OS/2 was supposed to be an OS for mainframe controllers, but that it got usurped by people with other ideas, and was just another project that wasn't going to fly very far.

I didn't know whether that was true, why they said it, or what really was going on with the people involved. But I remember the comment and always take what I hear about the history of projects with a grain of salt.

There's some misstatements in there. "IBM developed OS/2 2 on its own" is definitely not true. I was working with OS/2 1.x and receiving frequent drops of early development versions of OS/2 2.x all clearly from Microsoft on floppies with Microsoft OS/2 2.0 all over them. I even went to MS's Redmond campus for several weeks of OS/2 API programming. What is true is that MS was tasked with the OS/2 3.x effort while IBM carried on with the OS/2 2.x line.

At the time, the Windows 9x vs OS/2 2.x didn't have a clear expected winner. In the end the doom of OS/2 was due a lack of applications. There were other contributing factors like how IBM simultaneously launched the PS/2 line and the OS/2 1.x version were rarely run on non-IBM hardware (v1.3 ran like a dream on PC clones though.)

The WinOS2 development was an incredible marketing and technical marvel. Codenamed Ferengi and sold as "OS/2 for Windows" it allowed you to 'upgrade' your Windows-license installation to run OS/2 with Windows 3.1 running on top of OS/2. It would be like installing Linux on a Windows machine and getting Linux and Wine. Before it was done, most everyone was saying that it wasn't possible. It ran most of the native windows drivers the main thing that was tricky was sharing the screen to have seamless Windows rectangles overlapping with OS/2 rectangles.

I personally didn't care for running cooperatively scheduled Windows 16-bit apps on a 32-bit OS.

Mark Russinovich, the current CTO of Azure at Microsoft, wrote an article years ago pointing out that Windows NT was more than "influenced" by VMS and at the deepest levels it was closer to a rewrite of VMS from VAX assembly language into C.

Components evolved during the rewrite and, of course, there were many new things added, (Win32, NTFS, etc) but the base is the same.

https://www.itprotoday.com/compute-engines/windows-nt-and-vm...

VMS was mostly written in Bliss, a high-level language, not VAX assembler. When you bought a VAX, it came with a box of microfiche cards showing the VMS source code.

Reply on news.ycombinator.com