The Rise of Microsoft Windows 2x: Extended Thoughts
A few random thoughts and tangents related to Windows 2x
The Rise of Microsoft Windows Part 2
Did Microsoft copy the Macintosh with Windows 2x?
This is probably the biggest question that tends to arise when discussing Windows in the 1980s…were they, either explicitly or implicitly, attempting to copy the MacOS? The topic is probably going to eventually wind up as the subject of a dedicated video at some point, however I am just going to give my opinion in brief here. And to be clear, I’m not going to get too far into the weeds on this, this is just my opinion, with a few explanations as to why I think this. And I will also say that my current opinion has changed over the years as I’ve read and studied, and there is always the chance that I’ll run across some new information in the future that will change my thinking again.
The issue of whether Microsoft copied the MacOS with Windows can be looked at in one of two ways: legally or morally.
Legally speaking the answer is pretty clear-cut. Apple’s initial charges that claimed that Windows copied the Mac interface in 180 (or depending on the source, 189) different ways were quickly whittled down to just ten, and then those ten were dismissed over the course of the various lawsuits and appeals. At the end of the day, on February 21, 1995 when the Supreme Court refused to hear Apple’s appeal, the issue of whether Microsoft copied any of Apple’s copyrighted interface elements was firmly settled in the negative. Legally speaking, Microsoft did not copy any of the MacOS’s distinctive look-and-feel in a way that infringed on any of Apple’s copyrights.
But wait, was Microsoft found legally innocent because it actually didn’t infringe on any of Apple’s copyrights or was it because they were legally covered in perpetuity by the poorly worded contract that John Sculley had signed with them?
According to Judge Schwarzer, who wrote the original opinion on July 25th, 1989, all but ten of the original claimed infringements were covered by the license Apple had signed with Microsoft. And those ten claimed infringements were ruled as being broad ideas that could not be protected by copyright, as opposed to specific features that could be. But did this ruling mean that Microsoft had actually copied 179 Macintosh operating system design elements or did it just mean that even if they did, it wouldn’t matter because they were covered?
I am not a lawyer, and I have not read the opinion, merely commentary on it. Should I do a video on this subject, that is one of the key pieces of research material that I will need to dig through. With that caveat in mind, it appears to me that Microsoft would probably have been found to be in violation of Apple’s copyright…initially, had they not had the agreement as cover. However once appeals and counterclaims wound their way through the system, that would presumably have resulted in everything winding up at the same place, where everything was nullified due to the concepts being too broad for Apple to copyright. I’m definitely open to hearing other opinions however, so please feel free to leave a comment with your thoughts.
My current opinion is that Microsoft did not actually copy the MacOS in any meaningful way, although it is fair to point out that they did acknowledge that Windows 1.0 was “derivative” of the MacOS in the 1985 agreement. However it could be argued that the MacOS was itself derived from Xerox PARC’s Smalltalk operating system, and even that was somewhat derived from Douglas Engelbart’s work at SRI with his NLS system. And let’s not forget Ivan Sutherland and Sketchpad. So morally speaking, where do you draw the line between being “inspired by” versus “ripping off”, or more applicably, where is the line between “derived from” and “copied”?
I tend to think of it in terms of could somebody who was familiar with the MacOS but wasn’t familiar with Windows at all, sit down in front of a DeskPro 386 and not realize that they weren’t using MacOS?
If running Windows 2x was basically indistinguishable from the MacOS, that to me, would mean that morally speaking, Microsoft ripped off the MacOS. And does any version of Windows 2x feel like running MacOS?
The implementation of Windows’ look and feel in both 1.0 as well as the 2x line, was vastly different from the MacOS, not only superficially with Windows being in color, and not only with how it felt to the user, but also architecturally. And I am not just talking about the fact that Windows ran on Intel chips and the Mac ran on Motorola chips.
The two families of operating systems feel very different, and there are some ways in which Windows 1 and 2x are actually more advanced, or at least are trying to be more advanced than the MacOS was at the time, multitasking being a major difference, and Windows/386’s kinda sorta preemptive multitasking being a giant difference, even if it wasn’t terribly polished and was a pretty basic form of preemptive multitasking.
And of course the MacOS was heavily streamlined to be as friendly an operating system as possible, inviting and welcoming to non-computer users especially. That definitely does not apply to the clunky 1980s Windows versions at all. And I will also point out that using the MacOS felt very different from using an Alto, which wasn’t nearly as friendly to non-computer people and was somewhat more geared to programmers than users, what with its ability to let the user basically reprogram elements of the operating system on the fly. The Star was quite user friendly, but still was markedly different from either the Macintosh or Windows operating systems. You would never mistake one for the other.
We also have Tandy Trower’s sworn testimony that he was never told to copy the MacOS, either implicitly or explicitly.
So to me, legally and morally, Microsoft did not copy the MacOS in any problematic way. HOWEVER, it would be rather disingenuous to ignore the fact that the MacOS was a huge influence on GUIs of the era. And the MacOS was a further evolution (some may say devolution) of the Lisa operating system, which was itself influenced by Xerox PARC’s work. I do not feel that either the MacOS or Lisa OS were copies of the Alto/Xerox Star, and condemning Windows for being inspired by the MacOS also requires condemning the Mac/Lisa operating systems for being inspired by PARC’s work. I will admit that my opinion on this has shifted back and forth over the years as I’ve dug into the subject more.
Xerox PARC does get, and more importantly SHOULD get enormous amounts of credit and recognition for pioneering graphical user interfaces and developing the concepts that enabled a GUI to be a useful tool, as their influence in some way or another literally touches every single GUI that came after the Alto. As various members of PARC’s team left Xerox and went to work all across the personal computer industry of the 1980s, the concepts and ideas that they were a part of developing were disseminated even further, and built upon both by them, by the people they were in contact with, and by people who saw the graphical user interface concepts and went about implementing them in their own unique way.
However, let's not forget the sheer chutzpah of Microsoft licensing the rights to overlapping windows to HP for use in NewWave. I wonder at the legal reasoning behind it, or if there was any versus Microsoft just looking to get some money. I mean, if overlapping windows aren’t something Apple could trademark…then Microsoft certainly couldn’t claim to be able to license them out.
On the other hand, if they were a trademark of Apple, but were covered under the licensing agreement that Sculley had so unwisely signed…then I’m not sure what rationale there would be for extending that agreement to allowing Microsoft to license out Apple’s copyrighted design elements. I mean, either way it makes no sense, so I assume there was a hefty amount of cash involved somehow, probably going from HP to Microsoft.
But this is all pure conjecture, so feel free to speculate away in the comments with your own thoughts.
IBM and the 386
One of the reasons that Intel did what it did with the 386 processor is because they apparently weren’t making very much profit on the 286. I don’t know what this translates to so far as margin goes, however according to the book Big Blues, Intel was being paid only about nine dollars per 286. The bigger problem was that since IBM hated relying on a single source supplier, it insisted that Intel license out the rights and designs to its processors to competing companies, which is how AMD really got rolling. Intel was being forced to essentially subsidize its competition, keeping margins low and strengthening its rivals.
And this was happening in an era when Intel’s Japanese rivals were rapidly gaining strength, especially in making RAM chips, which were actually Intel’s main source of income in the early 1980s. 1983 was an absolute bloodbath for Intel, as they lost 114 million dollars in Q3 alone. There was apparently enough concern about survival that between 1983 and 1984 IBM actually invested around 400 million dollars into Intel, which left them owning about 20 percent of Intel with options to buy an additional 10 percent if they chose to exercise it.
Obviously Intel eventually rebounded, and by the early 1990s was in fact sometimes out-earning IBM, but their weakened position in the mid-1980s ensured that they were dominated by IBM and didn't have the leverage to push back against IBM’s demand that Intel license out its 286 chip to other manufacturers.
However Intel decided to take a gamble with the 386, something that was quite risky at the time. Once again in the words of Big Blue, “A maker of computer processors is in a peculiar position…a maker of processors has to rely on the kindness of strangers in a world where only a few strangers can help–those being the handful of major customers, such as IBM, who sell huge volumes of PCs that might use the processor. When IBM showed little interest in using the 80386 soon, Intel was faced with the choice of either finding someone else to champion the chip and bring it to market in a PC or sitting around and waiting until IBM adopted the 80386–all the while making little money from the prior-generation processor because of the heavy competition in that part of the market.
So basically IBM played a large role in putting Intel in a position where they felt that partnering with Compaq to get the Deskpro 386 made was worth risking their business relationship with IBM for, as successfully getting the 386 into consumer computers would supercharge their profits and market share, since they controlled the 386 and did not plan to license it out. Obviously it could be reverse engineered by companies such as AMD (who could and did), but that would require enormous amounts more time to do, time that Intel could use to get even further ahead.
And obviously Intel’s plans worked, the 386 was a huge success, and they went from needing a bailout from IBM in the 1980s, all the way to being so successful in the 1990s that they were sometimes worth more than IBM was.
However, there was one second source supplier for the 386 that could have given Intel a serious problem…and that supplier was IBM themselves. Or rather, it could have been. Apparently Intel gave IBM the rights to produce the 386 if they wanted to, including the right to modify Intel’s design for it. This was done to help mollify IBM after Intel continued to resist their demands that it license out the 386 design to its competitors so IBM could have a second outside source for chips. You know, for when IBM finally produced a 386 based system of its own.
IBM did have a lot of very talented chip designers working for it, and according to Big Blues, IBM actually tried not once but twice to produce a far more powerful version of the 386 that they would have owned the rights to. Both attempts failed due to IBM’s amazing tendency to stifle innovation with bureaucracy.
As per Big Blues, the most interesting new 386 design attempt was done by one of IBM’s top scientists, Glenn Henry. Henry tried to marry a new version of the 386 with a workstation chip that IBM was working on. The goal was to produce a powerful new chip that was compatible with all PC programs, but also was compatible with the IBM workstations that would use the new workstation chip. This new hybrid chip would have let IBM merge their workstation market into a PC market, while controlling the whole thing and freezing out Intel entirely.
But of course IBM’s culture could not abide any sort of challenge to the workstation hegemony, and apparently Henry wound up spending most of his working week traveling to various meetings around IBM, trying to keep support up and defend the project. With only one or two days a week left to actually, you know, WORK on the project, he finally gave up in frustration and went to work for Dell.
IBM wasn’t able to bring a homegrown version of the 386 to its computers until 1991, an astonishing half decade after Compaq kickstarted the 386 market with the Deskpro 386. Until that point, the 386 based systems that IBM was shipping used other manufacturers’ processors (presumably just Intel’s).
IBM’s consolation prize…of sorts, was that in 1986-1987 they sold off their 20 percent stake in Intel for 625 million dollars, over half of which was profit. Intel had wanted IBM to sell off its share so Intel could more freely deal with IBM’s competitors in the PC market. While it was kind of IBM to sell off its stake…I guess…had they held onto it for another 6-7 years, it would have been worth billions. In 1993 for example, IBM’s 20 percent share would have been worth 5.4 billion.
IBM also been given the right to buy a further 10 percent of Intel, a right it never exercised, and one that would have been worth at least a couple billion more in the 1990s. Money that IBM could probably have really used as it entered the era of downsizing, losses, and collapsing market share. Imagine what could have been done for OS/2 Warp if IBM had a billion dollars to incentivize OEM’s to preinstall it instead of Windows. Of course that would have required IBM to have people in marketing who would actually do that…
Ahhh IBM…few other companies in history have been so relentlessly self-sabotaging.
Windows 2x bundle deals with OEM’s
I would love to know more about what OEM’s shipped Windows 2x with their systems, but I haven’t really found much of anything on it. I am positive that Compaq wasn’t the only company shipping Windows/386 preinstalled on its 386 based systems, but I am not sure who, if anybody else.
I found an ad for AST that boasts that their systems “take advantage of the next-generation , 32-bit operating systems and environments, including Microsoft Windows/386, Unix System V/386, OS/2 and more.”
This doesn’t sound to me like AST was shipping Windows/386 preinstalled, it sounds more like it was simply an option you could get. Of course it's also possible that AST simply let the customer pick which operating environment they wanted their new computer to ship with, and technically were shipping Windows/386 with their systems.
When did Compaq start working with Microsoft on Windows/386?
A viewer named nickwallette6201 brought up a really good question about the development of Windows/386 that I’d like to expand on. Given the secrecy surrounding the Compaq Deskpro 386, and with IBM not prototyping any 386 systems at all, how could Microsoft have started developing Windows/386 without access to the actual hardware it was supposed to run under?
The questions of exactly when Microsoft learned about the Deskpro 386, when they started working on Windows/386 for it, when Compaq started working alongside Microsoft as a codeveloper and how heavily they were involved are interesting ones, but we just don’t have much information to go off of.
We know that Microsoft started working on Windows/386 in early 1986, after Gates came back from Armonk, frustrated over their refusal to target the 386 with OS/2. Or at least that’s what the sources seem to say.
But it's also possible that a key impetus for Gates deciding to do a 386 version of Windows was his inside knowledge that Compaq was developing the DeskPro386. It’s also possible that Compaq came to him first, asking him to do a 386 specific version of Windows for their upcoming Deskpro, and offering their own software engineers to help speed the process along.
Right now, all we can say for sure is that at some point Compaq was co-developing Windows/386 with Microsoft and the questions of when exactly they got involved as well as how much of a role they played…is a significant unknown.
It's also certainly possible that it was all three things, Gates’ knowledge that Compaq was working on a 386 based machine, plus his frustration with IBM’s lack of interest in targeting the 386, plus the seductive offer of additional software engineers courtesy of Compaq’s coffers instead of his own.
I purchased Rod Canion’s book while I was editing Rise of Windows part 2 but I haven't read it yet. But I did search it for references to Windows/386 and didn’t see any, it doesn’t seem to rate a mention at all. There is a section on page 99 that talks about the Deskpro 386 announcement and says “Bill Gates…talks about how excited he is about the Deskpro 386, how Intel, Microsoft, and Compaq have worked together to make it happen, and how Microsoft will support its advanced features with its operating systems….<t>he company is working to bring an advanced version of DOS to market soon.” The “advanced version of DOS” referred to here is probably referring to OS/2, not Windows/386.
The answers may only lie in Compaq’s and Microsoft’s email and memo archives for that time period. Of course Hewlett-Packard purchased Compaq in 2002 and I am not overly optimistic about the fate of those records, especially not electronic ones. For all I know, the answers are lying inside a bunch of old, forgotten data tapes tossed into a pile of moldering boxes in some forgotten HP sub-basement. I’d love to be proven wrong on this, but I’m betting the only place that might have answers is Microsoft's own archives, and I haven’t thought of a good approach to take…or who to pitch the aforesaid proposal to, in order to attempt to get access to Microsoft's archives.
I have reached out to HP in the probably vain hope of finding the right person to talk to that could give me access to Compaq’s archives from 1985-1988, and I let them know that I’d be more than willing to travel as needed in order to get access. However, as of writing this, I have not heard anything back. And it’s been multiple weeks with no response, so I’m pretty sure my message has been consigned to the metaphorical circular file.
To be honest, I would be very surprised if these archives even existed, let alone were accessible, but it would be nice to be proven wrong.
Windows/386 Preemptive Multitasking
A viewer named serhiymarchenko8361 left a comment asking about the preemptive multitasking of Windows/386, in light of the fact that a big feature of Windows 95 was its implementation of preemptive multitasking. Digging into it took me down a really interesting rabbit hole where I found something surprising (at least to me).
So far as I can tell, Windows/386 was the sole version of Windows 2x that supported preemptive multitasking, but it only had a limited implementation of it, and it was limited in a surprising way. An InfoWorld article from May 8, 1989 explicitly confirms that Windows/386 did have preemptive multitasking, albeit not to the same level as OS/2, saying that “While Windows/386 used the same kind of multitasking as OS/2 –called “preemptive multitasking” – it does not support multiple threads.”
In a somewhat more detailed feature from the July 1988 edition of Byte, there is a really good lengthy article on multitasking, called Weighing the Options: Comparing the Many Flavors of Multitasking. This article not only delves really deeply into what exactly multitasking is, but also the history of multitasking and briefly covers how different operating systems of the time implement multitasking. With regards to Windows/386, the article says that “Microsoft Windows/386 provides the same user interface as Windows 2.0, but it uses the virtual 8086 mode of the 80386 processor to let any MS-DOS application run in a window concurrently with other programs. Multitasking is preemptive. All programs can also run in the background without user intervention.”
In light of this, I think it’s safe to say that Windows/386 did have preemptive multitasking, the first Windows version to do so. However it seems that its implementation was pretty basic, lacking multithreading and also being limited to the virtual machines that DOS applications ran inside of. From an architectural perspective, this would seem to put Windows/386 ahead of the MacOS at the time, but behind OS/2.
But here is the interesting part…so far as I can tell, only the DOS virtual machines were preemptively multitasked, it seems that Windows applications themselves were cooperatively multitasked. So Windows/386 is an odd hodge-podge of preemptive multitasking and cooperative multitasking, depending on whether it was multiple MS-DOS applications or multiple Windows applications being run concurrently. And the rest of the Windows 2x line strictly used cooperative multitasking.
Again, I could be wrong here, but every reference I found seemed to be talking about preemptive multitasking in the context of running multiple concurrent virtual MS-DOS sessions, something that only Windows/386 could utilize. This would also neatly square with Windows 95 introducing preemptive multitasking, it wasn’t the first version of Windows to have it, but it was the first version of Windows to have it for Windows applications.
Who All Worked on Windows 2x
This list is undoubtedly incomplete but it's the most complete one that I have seen. Figuring out just who worked on Windows 2x is incredibly challenging, not least because 95% of the original Windows 1.0 team transferred over to OS/2 and Presentation Manager. Additionally, some devs apparently floated between PM and Windows 2x. Finally, there were devs that transferred elsewhere in Microsoft during Windows 2x’s development.
Then there are the confusing recollections of, for example, Gabe Newell, who stated in a 2007 interview that he was “the producer” on the first three releases of Windows. However Tandy Trower thinks that Newell transferred elsewhere in Microsoft sometime post-Windows 1.0’s release and during Windows 2’s development and I haven’t found anything else that mentions Newell in a management role with Windows 2 at all. However this doesn’t necessarily mean anything, given how fuzzy many of the details of this time period are. Regarding Newell, it's also interesting that he says he was involved with the first three versions of Windows…does he mean 1.0, 2.0, and 2.1 or does he mean 1.0, 2x, and 3.0? I assume it’s the former, but we all know what assuming makes you…
I did email Gabe directly, asking about this but I haven’t heard back yet and TBH, I don’t really expect to. But I figured it was worth trying at least, because you never know! And since Newell says he worked on the first three versions of Windows, there may be some hope for those of you patiently waiting for a certain third entry in a certain famous video game series…or else he dropped out before the third version of Windows was done, in which there may be a joke somewhere in there about the problems Newell has with completing third installments…but I am far too mature to make such a joke of course.
Anyhow, the below list is almost certainly incomplete, but it's the best one I have right now. If any new names come in, I will add them to this list. Where I have information (usually courtesy of Tandy Trower) on their role in Windows 2x development, I have added it as well.
Tandy Trower (Product Manager/Liaison between OS/2 and Windows)
David Weise (Memory management/probably many other things)
Bob Gunderson
Rao Remala
Chris Larson
Bob Matthews
Steve Wood (Kernel)
Neil Konzen (User)
Marlin Eller (GDI)
Walt Moore
Paul Klingler
Joe Rehfeld (PR/Marketing)
Paul Davis (Independent Software Vendor evangelism)
Gabe Newell (producer?)
Chip Anderson
Ron Gery
Why was the 286 considered crippled?
The 286 was a considerable step up from the 8088/8086 chips that powered the original IBM PC and clones based off of it. The enormously successful PC AT from 1986 was built around the 286 chip, and the 286 was not only considerably more powerful (Intel said 3-6x faster) than the original 8088 that powered the original PC, but also addressed a number of that chip’s shortcomings. And it was capable of being expanded to some surprisingly fast speeds, one later version from Harris apparently was capable of running at 25 MHz.
However…the 286 had a problem, a problem caused by one of its powerful new abilities. The 286 had two modes of operation, in the first one, called real mode, it was sort of a very fast 8088, still limited to directly accessing the same limited amount (either 1 MB or 640k, depending on how you look at it. Practically speaking, it was 640k) of RAM that was all the 8088/8086 could directly use.
Switching the 286 into protected mode allowed it to directly access up to 16 MB of RAM, a fairly ridiculously large amount of RAM for a personal computer of the era. This mode would however potentially break programs written for the original IBM PC and its 8088 processor, depending on how these programs handled memory access. As I understand it, in real mode memory is directly accessed, whereas in protected mode memory is indirectly accessed. The easy solution would simply be to switch the chip back and forth from real mode to protected mode, depending on whether you were running an older program or one written for the 286.
At this time, apart from TSR applets (tiny useful programs that lived in RAM and could be called up via a keystroke), personal computers mostly ran one thing at one time, and thus it would seem to be simple to just switch between modes based on what application was being launched. After all, if the user is only going to be running one application at a time, what does it matter if it's in real mode or protected mode? Just switch to whichever one the application requires.
Well…here we see the 286’s major flaw. It could be switched into protected mode from real mode without issue, however it could not revert back to real mode without a hardware reset. And the 286 was not capable of multitasking real mode applications either, only protected mode ones. And although users in the early 1980s were content to just run a single application at a time, by the mid 1980s people were starting to look for ways to do some form of multitasking (Andy Hertzfeld’s Switcher for the MacOS being a prime example).
This is probably the 286’s biggest Achilles heel, especially as computer technology and software advanced. The 386 was incredibly powerful for its day, but it's real selling feature lay in its ability to not only shift back and forth seamlessly between real mode and protected mode without any performance hit from literally rebooting the chip, but it also could run multiple DOS sessions in virtualized real mode, running under protected mode.
The 286 could do none of this, although there were ways of adding additional hardware around the chip to enable it to switch between modes transparently to the user. Although again, since you are doing a hardware reset, the user would certainly experience a significant and noticeable slowdown from what I can tell, with one source saying the computer would lock up for about one second during this. And of course, this did nothing to fix the issue of the 286 being unable to multitask real mode applications. Additionally, I suspect freezing the computer for a second would frequently have a calamitous effect on whatever application you were running at the time, especially if it was an MS-DOS one. But that’s purely a guess on my part.
Obviously these problems mostly are a non-issue if you never run any applications requiring real mode, and just run in protected mode all of the time. But the applications most PC compatible users were running at the time, were going to be targeted at the largest possible base of uses, and that usually meant being designed for the original IBM PC.