The Rise of Windows Part 2: Windows 2x

The confusing, messy, poorly documented, and fascinatingly complex story of Microsoft's follow-up to the initial release of Windows

This is the script for The Rise of Windows Part 2: Windows 2x. I have fixed a word here and there and there may be further edits/tweaks down the road if I spot a grammatical issue, a wrong word, or just decide to update a section in light of new information.

I hope you enjoy :)

Windows 1.0 post launch

After a painfully protracted development process that had made Microsoft the butt of jokes across the entire computer industry, Windows 1.0 had finally staggered out the door in late 1985. However…its future was badly in doubt. The butt of numerous jokes, the product of a frenzied development effort that had dragged on for years longer than originally promised, and a resource hog that ran like a dog on the hardware most PC owners had…Windows 1 was not exactly setting the world on fire.

Not only were actual retail sales a mere trickle, but Microsoft’s attempts to have it come packaged with OEM’s PC compatible offerings as it did with DOS…were also basically a total failure. Hence why Microsoft had reluctantly started selling it directly to consumers in a mostly futile effort to get somebody, anybody to buy it. 

Additionally, Microsoft’s attempts to get developers to write applications for it prior to launch had been largely unfruitful thanks to Windows’ lengthy development cycle, and many developers who had initially been arm-twisted into making a commitment had backed out of their previous commitment to do so just as soon as they could.

The OEMs that Gates had counted on to bundle Windows with their systems along with MS-DOS…well as previously mentioned that had pretty much fallen through as well. For example, Tandy had originally committed to bundling Windows with its PC compatibles, but in November of 1985 they announced that they were instead going to start offering GEM¹ from Digital Research instead. 

And this was only the first of many, as following Tandy’s announcement a number of other PC clone makers indicated that they too were planning on using GEM instead of Windows. Granted, Commodore, Apricot, and Epson were hardly enormous players in the PC compatible market (and yes I see you Commodore loyalists out there, and while Commodore’s PC compatibles were generally well received and sold well in Europe, they were hardly giving Compaq any nightmares), but the cumulative effect of so many companies publicly backing away from Windows and instead indicating support for rival products like GEM…well things didn't exactly look rosy for Windows. The fact that GEM was from Bill Gates’ rival Gary Kildall was probably extra painful to Gates.

And post launch there seemed to be little interest in writing Windows specific applications from any developer large or small. Among other things, it was also apparently a bit of a steep learning curve to develop a Windows application as it not only required the developer to get his head around Windows’ new APIs but also required using Microsoft C. 

As a matter of fact, Tandy Trower, who was then the director of marketing as well as the de-facto project manager for Windows, would admit in July of 1986 that there were no more than a dozen Windows applications, and possibly not even that many. As part of InfoWorld’s coverage² of a database program called Windows Filer 3.0, they interviewed Trower who said that the total number of Windows applications was “between half a dozen to a dozen”, although he stressed that he expected that number to grow “quite dramatically” over the next 6-12 months. 

Trower, the final product manager for Windows 1 after the grueling and stressful development effort had burned through four previous product managers, was still gamely working on pushing Windows forward and trying to get more developers…or really any developers interested in targeting Windows with their applications. To date, his efforts had been rewarded with basically no success. 

However there were some exceptions, a couple companies who decided to take a chance on Windows. Aldus was one of the first significant companies to come on board, who considered both Windows and GEM for PageMaker. According to an interview³ with Aldus co-founder Paul Brainerd, when it came to choosing between GEM and Windows 1, “our conclusion was that Windows was the clear winner in that context from both a technical and a marketing point of view.” 

Aldus wound up releasing PageMaker for Windows, right after Microsoft put out a new updated version of Windows, version 1.03, that added a PostScript printer driver to Windows. In a mildly entertaining bit of trivia, this driver enabled PageMaker to print to, among other printers, Apple’s own LaserPrinters. PageMaker also came included with a cut down version of Windows, basically a run-time version.

This enabled users who did not already own Windows to be able to run PageMaker inside of just the parts of Windows 1 that PageMaker needed to function. So while it looked like the user was running PageMaker under Windows, they were in fact just PageMaker with a Windows shell around it. 

There was also a port of the well received Balance of Power game that had originally come out for the Macintosh in 1985. In 1986 Mindscape released a version of it that also utilized the run-time version of Windows, making it possibly the first major game release that used Windows. 

Yet Windows still was only slowly moving copies. In the classic, and somewhat overused metaphor, Windows was suffering the chicken-and-egg paradox where there just wasn’t enough consumer interest in it to give a large enough install base for most developers to be interested in targeting applications at. And without compelling applications, what was the point of installing Windows? Unless one really wanted to play Reversi of course, Microsoft Solitaire still being some point in the future.

Steve Ballmer made an internal Microsoft video promoting Windows 1.0 that tried to bring a bit of fun to the glum state of Windows software and Windows 1.0 in general. Spoofing over-the-top low budget commercials such as Crazy Eddie’s famous ones, Ballmer enthusiastically portrayed Windows as an unbelievably good deal at only 99 dollars. For some reason this video is frequently portrayed as an actual commercial, but so far as I can tell, it was purely for Microsoft's internal enjoyment. And given Windows 1.0’s dismal reception and dismal software market, a little humor from Microsoft's most frenetically over-the-top executive was probably much appreciated.

In an attempt to unstick the Windows application market, in June of 1986 there was a Windows Development Seminar in New York City that Steve Ballmer ran. The goal was, as Gates⁴ puts it, “to rustle up software, any software for Windows”.

Specifically Windows was to be positioned as a key platform for developers wishing to create applications that were future proof, and would be able to easily take advantage of the new features such as the 286’s protected mode.

On day one Ballmer gave the opening speech and tried to sound confident and optimistic about the benefits of developing Windows applications, saying⁵, “It’s fairly well known from the press, et cetera, that Microsoft is working on a future version of DOS that supports the 286 in its so-called protected mode, where you have the full 16 megabytes of address space and you get access to the memory-management features of the 286. We have committed to make sure that the Windows API exists in a completely compatible fashion between real mode and today’s Windows and this new protected mode OS. We’ve also told people, and I’ll tell you again today, that it’ll be a fairly transparent process to take an application that runs on Windows today on top of DOS 3 and move it to a new version of Windows that runs on top of this new 286 protected mode version of DOS.”

Basically, develop your application for Windows today, and when Microsoft released the next major version of DOS, one that took advantage of the 286’s protected mode, Windows would handle the hard work of ensuring that your application automatically worked with protected mode without you needing to do any more development work. Easy-peasy, lemon-squeezy.

Sadly this seminar, like pretty much every other attempt Microsoft was making to court application developers for Windows, didn't really move the needle much, and Windows remained a platform with very few applications.

Ballmer here is talking about not only a new version of MS-DOS, but also a new version of Windows. This may be the first public confirmation of the fact that Microsoft was pushing ahead with a new release of Windows aimed at taking advantage of the 286’s protected mode. I am indebted to Tandy Trower for setting the record straight on this, as he was able to place the start of Windows 2.0’s development far earlier than I had expected, and far earlier than any of the sources in my research library suggested.

In fact, according to Tandy⁶ “I believe development started relatively shortly after the release of 1.0, though there may have [been] been a few weeks in between in dealing with the organizational changes that I mentioned before (i.e. that the majority of the team moved to the IBM-Microsoft joint development of Presentation Manager…) While I don’t recall the precise date, I don’t believe that was any later than Jan or Feb of 1986.”

However the team working on what would become Windows 2.0 was very small from the start, and the resources given to the project were going to be very minimal. 

Interestingly, the initial Windows 2.0 core team seems to have expanded over time until it reached a similar size to the Windows 1.0 team. This was very surprising to me, as literally every single source I consulted that gave any information about the size of the Windows 2.0 team…was pretty adamant that it was always comparatively small. For example Barbarians Led By Bill Gates claims⁷ that after Windows 1.0 launched,  “Gates would leave in place a tiny three-man team to nurse Windows along, but the maimed and much maligned project would slip to last place on his priority list.”  

However Tandy Trower told me that⁸  “the Windows 2.0 team, even though many devs transferred to Presentation Manager, still was a pretty significant team, somewhat comparable to the size of the 1.0 team. And they always had access to those on the Presentation Manager team, who were typically located on the same floor.” 

So while I think it's undeniable that Bill Gates’ attention was primarily on OS/2 and Presentation Manager, and that was where the bulk of Microsoft's development efforts and resources were going to be aimed at, a rather surprising number of developers worked on Windows 2.0, if not immediately, then eventually as development ramped up.

Matters were not helped by the fact that Microsoft’s focus was clearly and publicly shifting away from Windows in favor of what everyone agreed would be the new hotness in operating systems, courtesy of IBM and Microsoft’s combined development efforts for the newly announced upcoming operating system eventually dubbed OS/2. 

As a result of this, for some reason developers weren’t terribly eager to spend a lot of time creating applications for an operating system that didn’t seem to have a roadmap or future, no matter how much hopium was spread around by Gates and Ballmer. And yes, I still say that Windows 1 was functionally an operating system, even though I also admit that there is a very good argument to be made for it being merely a graphical shell for DOS. 

Although…OS/2 did not exactly get off to a roaring start, with the atmosphere between IBM and Microsoft pretty much strained from the start. I’ve covered the saga of OS/2 in other videos, including a ridiculously long one that I will link to above, so I will mostly be hitting just the highlights necessary for placing Windows’ further development in context. So let's talk briefly about the inception of what became OS/2, starting with the new head of IBM’s PC division as of 1985, Bill Lowe.

Bill Gates and Bill Lowe

Gates had always had a good relationship with the head of the original IBM PC project, Bill Estridge. There was mutual⁹ respect and trust between them, or at least as much as there could be between Gates and anyone at IBM. While the two had engaged in a number of arguments, Estridge was someone who had no difficulty standing up to Gates’ famously confrontational behavior as he was quite technically proficient himself and as a result, the relationship worked reasonably well for both parties. 

Unfortunately Estridge had lost an internal power struggle¹⁰ at IBM, and had been forced out of his position overseeing IBM’s Entry Systems Division, shunted off to the powerless role of vice president of manufacturing. 

The gist of what happened was that Estridge was seen as having overseen two major failures, first the PC Jr and second a major issue with defective hard drives in the AT. These two failures cost him his job, even though I have serious doubts as to just how much direct responsibility Estridge really bore. Still, IBM was full of warring fiefdoms and removing a successful and visionary leader from his position right as his skill and experience was badly needed to manage the next phase of their PC business…well it was far from the first time IBM made amazingly self destructive choices. And of course…it would be far from the last time.

Tragically Estridge was killed in August of 1985 when Delta Airlines Flight 191 crashed at the Dallas-Fort Worth Airport, killing 130 passengers. 

When Estridge was forced out, his former patron Bill Lowe took over the Entry Systems Division. Now Lowe was definitely a smart guy, and IBM’s development of the original IBM PC had come about as a result of his efforts, see my series on the Rise and Fall of the IBM PC for more details. Unfortunately, Lowe did not possess Estridge’s technical knowledge, which meant that he and Gates were basically guaranteed to have some, shall we say, friction, in their working relationship. 

In fact, according to columnist Stewart Alsop: Lowe was¹¹ “a guy…who knows nothing about computers.” Again, not someone who had any hope of gaining Gates’ respect. I’m also not sure it's fair to accuse Lowe of ignorance when it comes to computers, given that he is quite rightly generally considered the father of the IBM PC. Albeit not from an engineering perspective of course, but Lowe still was the man who saw the opportunity for an IBM personal computer, made it happen, and directly birthed the massive PC market that eventually took over the computing world.

So while he may not have been a computer engineer or software developer, he was still a smart, canny man who had demonstrated a solid understanding of what the market was looking for in a personal computer, at least to that date. So with that bit of perspective in mind, let’s talk about the first meeting between Lowe and Gates, which occurred in late 1985 . It umm…did not go well. 

The purpose of the meeting was to go over IBM’s plans for MS-DOS’s replacement, as well as inform Gates what it would be targeted at from a hardware perspective. IBM was very much in the driver’s seat with the new operating system, and had little to no intention of giving Microsoft much input. 

Gates was pushing hard to avoid having to target the “brain dead” 286 chip, he wanted instead to target the new operating system, OS/2, at the powerful new 386 machines, as opposed to the older 286 or 8088 based machines. At this point in 1985 the 386 had only just been announced, and so far as I can tell, there were no consumer PCs based around it as of yet. 

Yet Gates was positive that 386 based personal computers would be coming soon, and he wanted to just completely skip the crippled 286 processor entirely and get ready to hit the ground running with OS/2, targeting it at 386 based systems from the start. At this point Gates probably assumed that IBM would be the first to release a 386 based PC and would set the initial standard, just as they had with the original PC, and then the XT and the wildly popular AT.

The PC cloners would then take that hardware standard and create their own compatibles. The key thing would be that these systems would have the power to run a far more powerful operating system than the poky 8088 and crippled 286 could ever manage. 

Bill Lowe strongly disagreed with Gates’ position, and informed Gates that OS/2 was to be targeted at the 286, no ifs, ands, or buts. IBM had invested heavily into the AT line of computers, and had committed to supporting them. Customers, especially the corporate world that IBM tended to focus most on, had bought, and were continuing to buy large numbers of ATs, and AT clones, not that IBM was happy about the latter case, and IBM intended to follow through on its promise of extended support for that hardware standard.  

Whether Bill Gates liked it or not, OS/2 was going to be targeted at the 286 chip first and foremost, and would not take advantage of any of the 386’s special hardware features or capabilities. When 386 systems came to market, whenever IBM decided to actually develop one, they would run OS/2, but they would run OS/2 pretty much just as a faster 286. And IBM didn’t really see the 386 processor as suitable for a personal computer anyhow, it was far too powerful. And overly powerful personal computers made IBM rather skittish as they would rather not have their juicy mainframe market menaced by powerful personal computers that sold for far less than a mainframe.

So far as IBM was concerned, for the time being the 286 was the perfect chip for their needs, the market was snapping up their AT systems with it, and nobody was clamoring for a new operating system for a processor that wasn’t even in a consumer computer at this point.

Gates was not happy about this, but he had no choice but to follow IBM’s wishes when it came to OS/2 development. And in the interest of being fair, and in light of what was known at the time, IBM was not necessarily being unreasonable in this decision. The 286 chip may have had serious design flaws, but it was also a tested and reliable product by this point, with high chip yields that made for a far more affordable price point than the 386 would presumably get.

And given how many bugs the initial 286 design had possessed, and how hard Intel had been forced to work to squash them, it was highly likely that the enormously more complex 386 would suffer even more teething troubles than its parent had. 

Additionally, Intel had decided to keep the 386 chip in-house, and had not licensed it out to second source chip makers such as AMD. This would enable Intel to charge higher prices for the 386 but also made IBM nervous for a completely different reason. IBM did not like to rely on a single outside source for processors, and had previously required its chip vendors to license their designs to at least one additional source, something that had, for example, helped out AMD enormously.

Furthermore, according to the book Gates, IBM, which had its own fab facilities and could thus make its own chips in-house, had a number of revised 286 designs that included¹² “a superfast low-power version of the 286”. IBM apparently felt that these new 286 designs would give the product more legs, fix some of its issues, and allow them to keep using the 286 for quite a while longer while also keeping costs down by making some or all of these revised 286 chips themselves.

Trusting a proven, if flawed, chip that could be manufactured in-house over an unproven, expensive, and probably bug ridden unproven new one from a single outside vendor, wasn’t necessarily a bad idea given what was known at the time. In fact it actually made reasonable sense from a business perspective.

However in hindsight…this was actually a really, really, bad idea, and one that would wind up helping Microsoft and Windows enormously. But that was in the future, and for the time being, Gates was going to grudgingly go along with IBM’s wishes and follow their lead. He felt that he had little choice in the matter and the future he could see did not seem to have much of a place for Windows. 

With the signing of the JDA, and the assumption that whenever it came out, OS/2 would dominate the PC operating system market, and supplant MS-DOS, Windows’ continued development was very much hanging by a thread. What was the point in continuing the development of a product that had not only made Microsoft the butt of many industry jokes, but also was practically guaranteed to be easily swept aside by the industry dominating titan that OS/2 would surely turn out to be? 

Additionally, as OS/2 development ramped up it seemed to many that keeping Windows alive was tying up valuable company resources that would surely be better served by being reassigned to the OS/2 project. This was a position that Steve Ballmer, who had been tasked with overseeing the IBM relationship, was very much in favor of. In fact Ballmer apparently repeatedly pushed for Windows to be entirely canceled and its few remaining programmers moved to OS/2 development.

Yet Bill Gates, a man firmly convinced of the value of carefully hedging his bets, refused to cancel Windows development entirely, and kept the project slowly moving forward, even as the bulk of Microsoft’s attention was clearly on OS/2 development. It may have been slow, but development on Windows did continue and Microsoft’s applications developers were also making some limited use of it.

This is another sign of how desperate Microsoft was to get more applications using Windows, as while they would do just about anything to close a deal, that did not typically extend to giving away a product for free. Once Windows hit a critical mass of users and applications, the free run times could be revisited, but for the time being, runtime versions of Windows 1.03 were free to anybody that wanted to develop a Windows application. However there were very, very few takers.

And Windows was still badly in need of all the help it could get, even two years post launch. Although by March of 1987 Microsoft was proudly proclaiming sales of over half a million copies¹³ of Windows 1, most of these had either been given away for free or been part of a bundle. As one source¹⁴ puts it, “Windows was the poor relation, the cereal premium of the industry, tossed in the box with computers, mice, video cards, turbo cards, who knew, maybe baseball cards.”

Only about a hundred thousand copies of Windows 1 had actually been installed onto someone’s computer. This was not exactly setting any industry standards, far from it. And the battle over continuing Windows development, or focusing solely on Presentation Manager and OS/2 instead…well it was splitting the company. Windows was seen, even by many inside of Microsoft, as an outgrowth of MS-DOS. And as one source¹⁵ says: “DOS was creaky, DOS was stunted, and DOS, alas, was the foundation for Windows.”

OS/2 and Windows are far more intertwined during this time period than you might expect, and OS/2 and its influence weaves in and out of Windows’ story all throughout the late 1980s. Windows as a product was shaped by its influence in any number of ways, both direct and indirect.

In 1986 Microsoft had been able to tie up a loose end related to MS-DOS, a loose end of the worst sort so far as Bill Gates was concerned…one that allowed a computer company to legally sell PC’s loaded with MS-DOS without paying Microsoft a penny in licensing fees. Just my saying these words probably made a cold chill run down Bill Gates’ back this very minute.

A cornerstone of Microsoft’s licensing policy was that while they were always willing to negotiate the terms of a licensing agreement for MS-DOS in order to lock down a PC manufacturer, they always wound up with an agreement that made Microsoft money. 

Some agreements made them less, and some made them more, but the goal was for every agreement to contribute something to Microsoft’s bottom line. But there was still a company out there with the rights to install MS-DOS on the computers they sold, without sending Microsoft a single penny. And not just one version of MS-DOS, but whatever the current version was, no limits, forever. Fortunately for Microsoft, this company, Seattle Computer Products, was in serious financial difficulties.

Seattle Computer Products’s MS-DOS License

Seattle Computer Products had struggled for years, even with their unlimited and royalty free MS-DOS license. Tim Patterson, the original developer of what was originally called 86-DOS before Microsoft rebranded it, was no longer with them, and was actually in the middle of his second stint as a Microsoft employee, the second of three total employment periods. Seattle Computer Products and its owner, Rod Brock, had no successful hardware or software products anymore, but they did have one thing of value, a hold-over from the original agreement with Microsoft where they had sold the rights to DOS for a mere fifty thousand dollars. 

That hold-over was a license that gave Seattle Computer Products the right to use MS-DOS as much as they wanted, on as many systems as they wanted, forever. And this license also included the rights to any future versions of MS-DOS that Microsoft would create going forward, and was open ended. For as long as MS-DOS was an industry standard, Seattle Computer Products could use it for free on any computer that they sold.

Unfortunately, possession of this license had not translated into success for Seattle Computer Products, their IBM compatible systems weren’t really selling, they had no other software products, and the company was facing bankruptcy.

Brock started to think about what could be done if he sold the company and this lucrative license to the highest bidder…what would Dell, Compaq, or Tandy think of an opportunity to never have to pay Microsoft a licensing fee for MS-DOS ever again?

Brock made the decision to sell the company and its license to the highest bidder, but before doing so he decided to let Microsoft know of his plans, and offer them the chance to buy the company first, telling Microsoft that he felt that 20 million dollars was a fair price.

Microsoft responded positively to the letter, appreciating how much they owed to Seattle Computer Products for so cheaply selling them the rights to what became MS-DOS, and giving Microsoft control of the PC operating system market. 

Of course I’m joking, what actually happened was that Microsoft president Jon Shirley fired off a letter back to Brock literally the same day that he received Brock’s letter. In this letter, Shirley accused Brock of having an “exaggerated interpretation“¹⁶ of the original licensing agreement, then told him that the license was non-transferrable, couldn’t be sold, was valueless, etc, and basically told Brock to please go out of business quietly and without fuss, thank you very much and have a good day.

Unsurprisingly, this compelling argument failed to convince Brock of his errors and thus he immediately hired one of Seattle’s top law firms, the ironically named Bogle & Gates, and took Microsoft to court, filing a lawsuit in the King County Superior Court in February of 1986.

In the lawsuit, Brock demanded¹⁷ that Microsoft either allow Seattle Computer Products to sell its MS-DOS license to whoever it wanted to, or else admit that the original agreement that had given Microsoft the rights to what became known as MS-DOS was null and void. In the latter case, Microsoft would be required to hand over the vast majority of revenue that it had earned from licensing MS-DOS out, a massive financial windfall that would possibly bankrupt the company and would definitely make Brock a very happy, wealthy man. 

Gary Little

The original judge assigned to the case was Gary Little, a major stroke of luck for Microsoft as Little had been Gates’ constitutional law teacher years prior, and who seems to have known the Gates family socially. This did not result in Little recusing himself from the case, however Seattle Computer Products’ lead lawyer, Kelly Corr, got suspicious and demanded a different judge be assigned. He got a severe chewing out by Little, but did succeed in getting a different judge, one Gerard Shellan, assigned instead of Little.

Gerard Shellan

The case finally reached trial in front of a jury in November of 1986 and lasted only three weeks, with Gates and Paul Allen both being called to testify and Gates being present in court almost every day. While the jury was still deliberating, Brock accepted an out-of-court settlement whereby Microsoft bought back his DOS license for 925,000 dollars, less than five percent of what he had originally sought, but almost double what Brock had apparently been willing to settle for.

According to several¹⁸ sources¹⁹, Gates didn’t want to settle at all, saying that he had already paid Seattle Computer Products for DOS once, and had no intention of paying for it again. However in spite of this, Microsoft, undoubtedly at the advice of its legal team, very quickly started trying to settle, starting with an initial offer of 50,000 dollars, an offer that went up by around 100,000 dollars every couple hours as the jury deliberated, until Brock finally capitulated. 

Kelly Corr later said²⁰ “I used to sit there in court and say to myself, ‘Why’s this guy [Gates] in trial here wasting his time everyday? His time is so much more valuable. If he could buy Brock out for a half-million, it would be money well spent…I wasn’t sure if he was just trying to squash Brock on the principle of the thing or what… These guys play hardball. They grind people. It almost doesn’t make good business sense.”

Once the trial was over, Corr went out and bought stock in Microsoft, figuring that people who played as ruthlessly as they did, would probably keep doing well, even more so now that they had secured the only remaining royalty free MS-DOS license. 

I have no idea how long Corr held onto his shares, however I would imagine he did quite well off of them as the opportunity to buy Microsoft shares in 1986 would have obviously been extremely lucrative if they were held onto for any length of time.

The New and Improved DOS: CP-DOS/286DOS/DOS 5

IBM’s new operating system had been initially called CP-DOS by IBM, and 286DOS or DOS 5 by Microsoft. Because of course it needed multiple names. But its name did eventually morph into OS/2, after a lot of work. To avoid confusion however, I am going to just refer to it as OS/2, although I will also be referring to its graphical user interface by its given name, Presentation Manager, wherever appropriate. Presentation Manager was eventually replaced by Workplace Shell, but that was in the early 1990s and does not factor into the timeframe we are looking at.

OS/2 was in trouble in 1986, as the work that IBM was putting into it began to majorly spiral out of control, with feature creep beginning to bite with a vengeance. IBM hadn’t even managed to get any worthwhile progress on its graphical user interface by the time Gates and Ballmer flew out to Boca Raton in April 1986 for a meeting where they would pitch IBM on a super not-difficult, marginally inconvenient way of helping out CP-DOS. 

Their pitch? Use Windows as OS/2’s graphical user interface instead of developing a new one from scratch. Gates cheerfully assured IBM that since Windows had already been released and thus was a mature product, it was feature complete enough to easily port it over to run on top of OS/2, just as easily as it ran on top of MS-DOS. IBM brought up TopView compatibility, the zombie operating system that IBM so far refused to let die. 

Gates and Ballmer crossed their fingers and piously promised to add in support for the TopView API into Windows, enabling the one or two programs that ran under it to run correctly under Windows as well. 

Someone from IBM, none of my sources say who, told Microsoft about a company called Dynamical Systems Research, who had developed a TopView clone that was multiple times faster than IBM’s TopView as well as much smaller. This unknown IBMer jokingly suggested that Microsoft just buy this clone in order to speed things up. 

Microsoft thought that sounded like a smart idea, and so trotted down the road, or rather flew across the country, to Berkeley where Dynamical Systems Research’s tiny headquarters was located. DSR was unwilling to just sell off their product, possibly benefitting from Seattle Computer Products’ example, and thus Microsoft wound up signing a letter of intent to buy the entire company. All of this occurred only two weeks after the original meeting with IBM in Boca Raton, Florida. 

The purchase of this company would wind up having far-reaching consequences for Windows development and eventual success, none of which could have been foreseen at the time. In the words of one source:²¹ “...DSR would arguably be Microsoft’s single most important acquisition, changing the future of the company forever.” But that was in the future, right now the only thing on Gates’ mind was making sure that he could tell IBM that Microsoft could give them TopView compatibility in OS/2 via Windows. 

The next hurdle for Gates? SAA.

Windows’ and SAA 

I’ve briefly touched on SAA before, essentially it was IBM’s plan to unify its various computer offerings, from mainframes to micros, with a common API. Or in other words, a layer of software would sit in between applications and the actual hardware that they ran on. To me it sounds like a virtual machine, a virtualization layer that would be very portable and would give a single target for application developers.

This would allow programmers to just target this layer of software, and ideally mean that an application would run on any one of IBM’s offerings. Any loss of efficiency from running everything through a translation layer, would be more than offset by the money saved by getting rid of the duplicated effort required to write the same program multiple times for different IBM machines.  

It wasn’t a bad idea…in theory. As a matter of fact, that had been one of the major selling points of the enormously successful S/360 series back in 1960s, a family of computers that could all run the same applications across all lines big and small, specs permitting of course. However, so far as I am aware, this compatibility had all been accomplished through hardware, not software, and this philosophy had long since been abandoned by IBM anyhow. 

The successor to the S/360 line, the creatively named S/370 line, was not compatible with IBM’s System/3X minicomputers, although the same product lines maintained compatibility internally, i;e the S/370 family was all hardware compatible across all models, just as the S/360 line had been. The S/370 line was also mostly backwards compatible with the S/360 line, making for a theoretically smooth transition. However, the idea of offering a single line of computers that covered all sizes and needs from small to large, and was easily scalable…well IBM was no longer doing a very good job of it. 

Contrariwise, IBM’s competitors in the mainframe and minicomputer space had not been idle and Digital Equipment Corporation in particular had been making significant inroads into the market with their VAX line of hardware compatible computers. DEC sales reps were happily spouting DEC’s slogan of “Digital has it now” and closing deals based on the customer’s ability to start with a small VAX system and then easily migrate upward as their needs grew, no retraining or reprogramming needed. It was a winning formula and one that IBM had once pioneered when they solely focused on the S/360 series two decades prior.

IBM’s plan was to combine SAA with another standard, called Common User Access. SAA would focus on making programs able to run across the full range of IBM’s hardware offerings while CUA would ensure that things like menus and function keys would also stay consistent.²² And it went without saying that SAA and CUA would be entirely proprietary to IBM. 

In July of 1986, a mere three months after the April meeting where Gates and Ballmer had sold IBM on using Windows as OS/2’s graphical interface, and promised to update Windows to include TopView compatibility… Bill Gates and Steve Ballmer had another meeting in Armonk with Bill Lowe and several other IBM executives. This meeting started off by smacking them across the face with SAA and IBM’s plans to reshape its strategic direction around it.  

For starters, TopView was now part of the past, it would not be part of SAA and would not be considered part of IBM’s plans for moving forward. This came as a shock to Gates and Ballmer, who were no-doubt looking forward to bringing IBM up-to-date on the acquisition of Dynamical Systems and the way it would allegedly make it comparatively easy to add TopView compatibility into Windows running on top of OS/2.

Instead IBM would basically make a pretense of supporting TopView itself, since a handful of companies had actually bought into it based on IBM’s promise to support it, but no more real effort or resources would be expended on it in any form. IBM had grudgingly decided to accept the fact that TopView was clearly a failure and there was no point in spending time and labor to give compatibility with it to OS/2, whether Microsoft was doing the gruntwork on it or not. 

And while Gates and Ballmer were still reeling from the news that the still uncompleted purchase of Dynamical Systems had apparently been rendered completely pointless, Lowe hit them with another punch. In order for IBM's mainframe graphics terminals to support SAA, they would be using a brand new graphical display API called GDDM, which stood for Graphical Data Display Manager. GDDM was being developed under the codename Hawthorne by a team of IBM’s programmers based out of Hursley, England. Critically, GDDM was an entirely different way of doing graphics than the Graphics Device Interface or GDI API that Windows used.

Lowe informed Gates and Ballmer that OS/2 would use Hawthorne, not Windows as its graphical interface, and then finished by cheerfully informing Gates that he felt he should know this before Gates had lunch with the chairman of IBM, John Akers, the following day.

So Gates, ever the master salesman, responded with one of his greatest sales presentations ever. If you ever doubted that Gates had the ability to rise to an almost Steve Jobs’ level of salesmanship and reality distortion, the trick he now pulled to save the Windows deal with IBM should put them to rest.

Gates somehow managed to sell Bill Lowe on OS/2 having two graphical interfaces, Windows and Hawthorne. Lowe initially overrode the objections of his own team members to give Microsoft some time to come up with a viable way to work with IBM’s Hawthorne team in Hursley England.

Microsoft put together a six person team led by Gates himself and Nathan Myhrvold, and then this six person team flew to Boca Raton to face off against a thirty person IBM team. After two grueling weeks of daily meetings, Microsoft triumphed and the decision was made that OS/2’s graphical interface would be a combination of Windows and Hawthorne, formally dubbed Winthorne.

The Microsoft team was ecstatic, with Myhrvold saying: “We thought we had won big time. We had snatched one out of the fire, gone over and gotten the business. And we knew that we’d made some compromises in doing so. We knew that there would be a few changes in Windows that would be likely as a part of this. We knew that we’d also have this additional piece of crap, this GDDM layer.”

However the GDDM layer would prove to be a malignant tumor that would eventually nullify most of the Windows code in OS/2, making a mockery of Steve Ballmer’s June 1986 claim of compatibility between Windows and OS/2. One Microsoft employee quipped that²³ “In general Windows and Presentation Manager are very similar. They only differ in every single applicable detail.” 

And it didn’t take long for this to happen, as by October of 1986, Ballmer was apparently quietly, and privately, admitting that Windows applications probably wouldn't just run under OS/2, and would probably have to at the very least be recompiled. 

But there was another thing that came about as a result of, or rather in spite of, the July 1986 meeting. Even though the TopView compatibility was no longer needed or wanted, Microsoft was still obligated to purchase Dynamical Systems, thanks to their signed letter of intent.

They certainly could have tried to back out of the purchase anyhow and fought it out in court, trusting in their deep pockets and ability to bury their opponent under a wave of litigation until they hopefully gave up. This was entirely in keeping with Microsoft’s way of doing business and would be far from the first or the last time they resorted to similar tactics.

 Nevertheless they decided to go through with the purchase, and while the original rationale for that decision no longer existed, they would gain six experienced new programmers, including one named David Weise. 

Weise is notable because in just a few years he would play a critical role in making Windows the success it eventually became. In the end, Microsoft's purchase of Dynamical Systems would turn out to be one of the turns of good luck that played such a pivotal role in the company’s success. But that’s getting ahead of ourselves, and that’s a story for the next entry in this series.

Windows 2’s Ongoing Development

There really was little impetus to develop further versions of Windows, and the market certainly wasn’t demanding a new and updated version. Plus there was the fact that the bulk of Microsoft’s attention was firmly focused on the all consuming resource suck that was OS/2 development. However, Microsoft’s very first Windows application, Excel, needed a graphical interface to run under, and the unloved original version of Windows wasn’t going to cut it. 

Now, Windows 2.0 development had been quietly ongoing since the end of 1985, with one source calling it²⁴ “that ugly little Microsoft orphan, Windows” so it was already at least somewhat along the road towards release. However the focus of it was going to be primarily to give Excel what it needed to run under, one of which included one of my least favorite Windows conventions, MDI or Multiple Document Windows. This is what lets Excel and other Windows applications spawn multiple windows INSIDE of the main application window, although I think Excel 2010 was the last version to utilize it. This has nothing to do with the main topic of this video, I just felt the need to mention it.

Windows 2.0 may have been ongoing for most of a year by this point, but its development was going to move forward with at least somewhat of a renewed purpose now, since it was going to hopefully be kept compatible with Presentation Manager on top of the existing goal of fixing the shortcomings of Windows 1.0 and giving Excel a good platform to run under. 

Teamwise, it’s really hard to nail down exactly who was on the Windows 2.0 team…Barbarians led By Bill Gates says it was really just David Weise, Rao Remala, and Bob Gunderson, which is definitely incorrect. Barbarians Led By Bill Gates also claims that Windows 2.0 development didn’t start until September of 1986, which is at least nine months later than when it actually started.

Tandy Trower was kind enough to try to come up with a list of people for me and what they primarily worked on, although he states that one of the complications with nailing down just who worked on Windows 2.0 is the fact that some programmers floated between both Presentation Manager and Windows 2.0 development as part of Microsoft's attempt to keep the two projects in sync. Additionally, some people like Gabe Newell, left the Windows project at some point during Windows 2’s development.

So with those caveats in mind, here is what Trower remembers:²⁵ “there were developers like Steve Wood (Kernel) and Neil Konzen (User) and possibly Marlin Eller (GDI) that were straddling both development efforts. You also had the recent acquisition of Nathan Myrvold’s Dynamical Systems, which included David Weise. While most of Nathan’s team focused on the IBM JDA world, as noted Weise was the exception and it is possible that others who came with him may have had dual work. For example I think Konzen may have helped on getting Windows back to overlapping windows. But back to the core dev team for Windows 2.0, I believe it also included Walt Moore and Paul Klingler. I am also trying to recall which team Lynn Shaw transitioned to. I think she ultimately was part of the Presentation Manager team before leaving the company. 

On my project management/marketing side, I think I was still assisted by Joe Rehfeld for PR/marketing and Paul Davis, who now worked for me, still focused on ISV (Independent Software Vendor) “evangelism”.

As previously mentioned, Tandy Trower was still the project manager for Windows 2.0, but Windows 2.0 development itself was led by one of Microsoft’s top application developers, Bob Matthews. Matthews, just like Trower, had to deal with the headache of trying to keep Presentation Manager and Windows 2.0 compatible, a truly thankless task. Once Windows 2.0 was released, Matthews returned to the Applications division of Microsoft, from where he had been plucked in the first place.

As previously mentioned, Tandy Trower definitively told me that Windows 2.0’s development began within at most a few weeks of Windows 1.0’s release. And Trower also states that one of the goals from the outset was for Windows 2.0 to run on top of the 286 in order to better support a port of Excel from the Macintosh, saying²⁶ “I do know [that] part of the 2.0 plan was to support the 286. Excel was not a skinny app, so for us to be hosted under Windows and succeed against Lotus 1-2-3, memory would have been a top priority.”

Most likely there was a major meeting in September of 1986, but it wasn’t a meeting to start development of Windows 2.0. It was either to clarify the existing development goals, and/or nail down some final goals for Windows 2.0. Given that the development of Windows 2.0 had been ongoing for most of a year at this point, most of the direction and goals for Windows 2.0 had already been defined so I doubt there was much to tweak or change, but this is my best guess as to what happened.

Steve Ballmer

Ballmer told them that Windows 2 was going to happen solely because Excel needed it, and then said²⁷ “This is it, after this we’re not going to have any more Windows. It’s all OS/2.” The expectations for Windows 2 were so scaled down from the original aspirations for Windows that it’s rather amazing. 

Back in 2010 Tandy wrote an article for the Technologizer website that covers a lot of Windows development history as well. From the article “Meanwhile I would continue to manage a second release of Windows (2.0) for which the Microsoft Applications Group would use to target a version of its increasingly popular Excel and Word applications. However, Ballmer moved most of the core Windows development team to the new joint development project with IBM. Even I had a partial responsibility for working with IBM to try to keep the interfaces between Windows 2.0 and OS/2 consistent so users could easily transition.” 

Tandy Trower was kind enough to share with me a lot of fascinating memories and information about this time period in Windows development. I have and will be quoting extensively from his emails to me, as they fill in a lot of detail that is frankly difficult-to-impossible to find elsewhere. 

Regarding the situation he faced at the start of Windows 2.0’s development, Tandy told me that:²⁸ “The only problem was (that) to develop the new OS,” referring to OS/2 and Presentation Manager,  “Ballmer needed 95% of the original Windows team. So starting out Windows 2.0, I was given a new dev manager from the Microsoft Apps team. This was also to ensure that Windows 2.0 would have all the necessary requirements to run a port of Excel, something that Gates made a top priority. But in addition to working with a new dev team, I also had been given the responsibility to be Microsoft's advocate for UI design of OS/2 and Gates wanted to ensure that Windows and OS/2 would be very similar. 

That would have been hard enough if I only had to interface with the IBM OS/2. However, with the success of the IBM PC, many organizations inside IBM now wanted to have control over its future, and some saw the need to unify all of IBM's UIs. So I ended up in design committee meetings with some that recognized the value of GUI, while others wanted to ensure that the new OS's UI was more like the data terminals IBM sold. 

For example, in Windows we mapped the Esc key to cancel any menu, but there were numerous people at IBM that thought that we needed to put "Esc = Cancel" at the bottom of each menu. To further complicate things, IBM would run usability studies to make their argument, while I had no such support other than that we had successful applications for the Macintosh.”

So on top of managing the overall development of Windows 2.0, Trower also had to do a significant amount of coordination with IBM to keep Windows 2.0’s interface somewhat in sync with OS/2, and this coordination frequently required him to take long flights from Seattle to not only IBM’s East Coast offices, but also flights all the way to IBM’s main OS/2 developer center in Hursley.

Trower had to make Windows 2.0 happen with far fewer resources than Windows 1 had been given, and a compressed development schedule of only eighteen months. Yes you heard that right, even after becoming an industry laughing stock for missing so many deadlines with the original release of Windows, Microsoft still decided to give Windows 2 a very compressed development timeline. 

But Trower took his task seriously, and was determined to fix some of the most annoying things about Windows 1. In his Technologizer article from 2010 he said, “I had about eighteen months to come up with Windows 2.0, and so I tried to schedule all those things I had been unable to get into the first release. I had the interface changed back to overlapping windows, added a proportional system font, and tried to make as many UI improvements as I could as well as any changes required to maintain some level of user interface compatible with OS/2 Presentation Manager as it evolved, all with a mostly new development team (since the former seasoned developers were now working on OS/2).”

Trower told me that his Windows 2.0 schedule was basically²⁹ “split between driving the changes I felt the product needed (with a new dev team), keeping our own apps group happy, and negotiating UI design with IBM.” 

As a way of helping Windows 2 be able to use more memory, Microsoft had briefly put aside its rivalry with…basically everybody who had a competing application, and joined with Lotus, Intel, and AST to come up with a new version of the Expanded Memory Specification. The original version of EMS had been characterized by Gates using the inspiring words³⁰ “It’s garbage! It’s a kludge!”, but the new version, EMS 4.0 did allow DOS to utilize memory capacities up to 32 megabytes, which was of course a colossal amount of memory for a personal computer of the 1980s. 

Tying in with EMS 4.0, Microsoft also attempted to improve Windows’ performance on original IBM PCs and clones by developing and releasing a product called the Mach 20 board. This board enabled users to upgrade an 8088 PC to the much faster 286 processor, which would allow original PCs to run the original Windows release as well as the soon-to-be-released Windows 2.0 at something other than a snail’s pace and would thus hopefully increase the pool of potential Windows users, something Microsoft fervently desired. This add-on board, formally named the Mach 20 Performance Enhancement System, was a three-quarter length board that came in several processor and memory variants. 

The Mach 20 Performance Enhancement System

According to an InfoWorld article from September 1987, the Mach 20 base model included an 8 MHZ 286 and 16k of RAM. Users also had several options for frankensteining a daughterboard onto the Mach 20, such as the Memory Plus board, which in its base configuration added another 512k of RAM and was EMS 4.0 compatible, or a floppy disk controller called Disk Plus. 

These additional add-on cards for the Mach 20…add-on extended it from a three-quarter length all the way to full length, which is why only one could apparently be used at a time. Mach 20 cards were also advertised as a way to get OS/2 to run better, since in 1987 everybody assumed that OS/2 was going to be the operating system of the PC future and Windows was only hanging around as basically Bill Gates’ hole card in case he needed it. 

However nobody really cared about spending 495 dollars on the Mach 20, especially one with a slower 286 than other upgrade options on the market and thus these cards joined other unlamented Microsoft products such as the Mac Enhancer and the PCjr Booster in the same place that most of Google’s products seem to wind up in.  

The Two Faces of Windows 2

IBM may have rejected the idea of targeting OS/2 at the speedy new 386 systems, but Bill Gates saw an opportunity to give Windows a boost by producing a version of Windows 2 aimed explicitly at 386 based machines. However he once again hedged his bets by splitting Windows 2 into two separate versions. Windows 2.0 itself would run on standard 8088 and 286 based systems, but there would also be a version of Windows 2.0, descriptively called Windows/386 that would be solely aimed at the 386 processor, and would take advantage of some of its special characteristics. 

This was a very, very canny move on Gates' part, and in my opinion, is one of the first concrete actions he took that shows he was starting to think of Windows as less of a back up plan, and more of a strategic product in its own right.

According to Hard Drive, the decision to split Windows 2 into two separate versions came about as a result of a meeting Gates had with Microsoft’s application development team in early 1986. In this meeting, after ranting about IBM’s strategy of focusing on the AT PC with its crippled 286 while ignoring the 386 entirely, Gates finally said³¹ “Screw the people with ATs; let’s just make it [the new operating system] for the 386, so they can upgrade.” 

The early 1986 decision to create a version of Windows 2 specifically for the powerful 386 would prove to be extremely fortuitous. In September of 1986, the largest and most powerful of the PC cloners, Compaq, released a consumer PC built around the 386. This powerful new machine’s processor was explicitly called out in its name, the Deskpro 386. And it took the market by storm.

And fortuitous may be the wrong word here as it suggests a large element of luck, and while that certainly played a role in Windows' eventual success, it also appears that Microsoft was aware that Compaq was going to release a 386 based computer. According to the book Gates, Microsoft worked³² “in close conjunction with Compaq” to develop Windows 386. If this is correct, then it's another example of Gates’ uncanny ability to hedge his bets in the most successful way possible.

The 386 processor had only been announced in October of 1985, and it initially was available in a 16 MHz version that was close to triple the speed of the 6 MHz 286. When it launched in September of 1986, Compaq’s Deskpro 386 initially sold for 6,499 dollars³³ and came with 1 MB of RAM. This was not a huge amount more than a specced out PC AT. And for that price you got a system that was obviously considerably more powerful, even with the limitations imposed on it by running MS-DOS.

All the other cloners had been waiting for IBM to release a system based around the 386, and in so doing presumably set the standard for 386 based systems that the cloners would then focus on replicating. However none of them had wanted to take a chance on introducing such a radical new machine themselves.

Once Compaq had blazed the trail however, competing systems quickly emerged and prices for 386 based compatibles started to head downhill, with Advanced Logical Research bringing the ALR Access 386 to market for a starting price of only 3,990 dollars, and PC’s Limited, soon to be known as Dell, releasing the most powerful 386 compatible to date in early 1987. This last system, descriptively named the PC’s Limited 386, started at only 4,499 dollars for a system with 1 MB of RAM and a 40 MB hard drive.

Thanks to IBM’s belief that the 386 was too powerful for a consumer system and was better suited to minicomputers, they hadn’t even started prototyping a 386 PC by late 1986. 

According to Hard Drive, IBM hadn’t even ordered a single 386 chip³⁴ from Intel when Compaq released the Deskpro 386. And since this new system was a game changer, it quickly became evident that it was Compaq setting the standard for 386 systems, and IBM wasn’t even in the race. 

However, the new owners of these powerful new systems had to deal with a number of unpleasant compromises if they stuck to MS-DOS, bottlenecks that negated a lot of the extra capabilities of these powerful new systems. It didn’t matter how powerful the processor was, they were still hobbled by MS-DOS’s 640k RAM limit, and had to use the same clunky workarounds to access more memory that older systems did.

This was a rather obvious gap in the market and due to Gates' decision at the beginning of the year, Windows 2 was going to be able to take advantage of it once it was released. But that release was still a year away, and when the Deskpro 386 was launched, Microsoft hadn’t even publicly announced that they were doing a new version of Windows.

But partially as a result of the Deskpro’s release, IBM was feeling some serious pressure that was going to push it into making a decision that would set Gates and Windows on the path to success. And it started with a meeting Gates had in Armonk, in late 1986.

In the leadup to this meeting, IBM had belatedly realized that not only had they seriously erred by focusing on the AT and its 286 chip rather than the 386, but they also needed a graphical user interface sooner rather than later. The Macintosh, although not threatening the PC juggernaut, was starting to sell decently, and more importantly the buzz around graphical interfaces seemed to be growing by leaps and bounds. And OS/2 was still far behind schedule, the Winthorne based graphical user interface was nowhere near done, and OS/2 was a long way away from shipping.

Windows and Presentation Manager 

And what did Lowe want from his meeting with Gates? He wanted OS/2 to take advantage of the 386’s features, while still maintaining focus on the 286. But more importantly for Gates, Lowe told him that IBM was willing to license Windows to be OS/2’s graphical user interface, however he wanted the name changed to Presentation Manager. Lowe also pressed his advantage with Gates by also adding some additional stipulations. A number of elements for Windows needed to be rewritten to incorporate elements of IBM’s mainframes. 

Now here is where the record is a bit fuzzy. Hard Drive talks about this meeting as if occurred after the release of the Deskpro 386 in September of 1986, but requiring Windows to adopt IBM’s mainframe graphics system is definitely referring to the Common User Access or CUA that was discussed in the meeting from July of 1986 when Microsoft was told that Windows no longer needed TopView compatibility but did need to support CUA and SAA.

Since I cannot nail down exactly what happened at the meeting, here is my best guess as to what happened. And I stress that this is a guess. We know that the decision to adopt Windows as OS/2’s interface was already done before the release of the DeskPro 386, this was the Winthorne graphical user interface. I think that references to this are confusing the timeline of the decision to make a version of OS/2 that could take advantage of the 386’s power with the decision to use the Winthorne graphical user interface for Presentation Manager.

So my quasi-educated guess on the timeline is that in July of 1986, Gates and Myhrvold successfully convince IBM to utilize the Windows graphical user interface for OS/2, merging the Hawthorne and Windows interfaces into Winthorne and agreeing to incorporate SAA and CUA into it. At this point though, OS/2 and Winthorne are solely targeted at 286 based PCs. They would obviously run on 386 based computers, but would be unable to take advantage of any of the 386’s advanced features.

Then Compaq takes the computing world by storm in September of 1986, just two months later. This makes IBM nervous, and they realize that they need to make a version of OS/2 that targets the 386, and they need it and its graphical user interface as fast as possible. So they decide to double down on using as much of the Windows graphical user interface as possible, hoping to save considerable development time.

The late 1986 meeting between Gates and IBM mostly is there to tell Gates that IBM now wants OS/2 to work well on 386 systems, and anything that can be used from Windows to speed up the development of the newly dubbed Presentation Manager graphical user interface is now a priority. I stress that this is my educated guess, but it's the only timeline that makes sense to me. If anyone watching this has any more light to shed on this timeline here, please leave a comment below with your thoughts.  

Given the fact that Compaq didn’t really broadcast their development of the Deskpro 386, it also seems reasonable that IBM initially didn’t take any rumors about its existence seriously. I found an InfoWorld article from April of 1986 called “386-Based Machines Due in ‘86”, that vaguely cites a couple unspecified manufacturers possibly including Compaq as possibly working on 386 based PCs, but gives no specific details. 

As development work on Presentation Manager steadily ramped up, Steve Ballmer actually pushed to cancel Windows entirely,³⁵ arguing that Microsoft was wasting resources developing two separate versions of Windows 2.0. 

According to one source,³⁶ “It came down to Ballmer and Gates having it out. Ballmer wanted to kill Windows. Gates prevented him from doing it. Gates viewed it as a defensive strategy. Just look at Gates. Every time he does something, he tries to cover his bet. He tries to have more than one thing going at once. He didn’t want to commit everything to OS/2, just on the off chance it didn’t work.”

Still, development resources for Windows 2 continued to be pretty minimal. The overriding focus for Microsoft was clearly getting OS/2 and Presentation Manager finished and out the door where they would no doubt set the new standard that everyone would have to deal with. One of the changes IBM required for Presentation Manager was changing some of the words used for commands to be more user-friendly, changes that would become somewhat of an accepted standard for windowing based GUIs going forward to the present day. For example, Windows 1’s “Zoom” and “Icon” buttons were re-labeled to “Maximize” and “Minimize” for Presentation Manager’s windows.³⁷

Interestingly, on December 26, 1986, Microsoft and IBM signed a new agreement that covered the money IBM would be paying Microsoft for its work on Presentation Manager AND also…you ready for this? IBM also agreed to start selling Windows, at this point still just Windows 1.0 but presumably once Windows 2.0 was released it would be also sold. Now this didn’t represent an official endorsement of Windows, as it was simply one more operating system that IBM’s direct sales force was selling, along with the somehow-still-kicking TopView and Digital Research’s GEM. Still, this was something that Gates could use to generate useful PR spin for Windows. 

In March of 1987, IBM went a bit further, or at least a part of it did. The IBM Publishing Systems business unit publicly announced that it would be standardizing on DOS, Windows and Adobe PostScript. Gates carefully straddled the proverbial fence, publicly stating that this was a nice boost for Windows while also stating that this only applied to desktop publishing and people shouldn't read more into it.

Windows 2’s Release and Reception

In April of 1987 Windows 2.0 was announced and even though Windows and Presentation Manager were already no longer compatible by this point, Microsoft engaged in some rather clever phrasing. 

As the book Gates puts it:³⁸ “In an opaque six-page press release, Microsoft announced Windows version 2.0, which boasted “visual fidelity” to something called the Microsoft Operating System/2 Windows presentation manager. This was a handy way of pretending that Windows and Presentation Manager were still almost the same thing, which they no longer were and never would be. Ballmer would now publically mutter something about a “mechanical recompilation step” to translate Windows programs to the new graphical interface–a wild oversimplification. By this time, the links between Windows and Presentation Manager had been irrevocably broken.”  

In September of 1987 both Windows 2.0 and Windows/386 were officially shown to the public, with a promised release date of October for Windows/386. I am not sure exactly what the promised date for the standard version of Windows 2.0 was, but it was undoubtedly either October or November.  

Interestingly, Compaq is credited by InfoWorld as well as the book Gates as a codeveloper of Windows/386, which does make sense in light of the fact that the Deskpro 386 would greatly benefit from the 386 specific version of Windows. Additionally, Compaq was the first OEM to bundle Windows/386 with its 386 based systems, giving it a solid advantage versus competing systems from other cloners.

 Assuming that the 386 version of Windows was started in early 1986 after Gates got frustrated with IBM’s refusal to target the 386 with OS/2, the development of it fits reasonably neatly into the 18 month development timeframe that Tandy Trower has mentioned Windows 2.0 was given.

Given the fact that the normal version of Windows 2.0 went into development only a few weeks after Windows 1.0’s release in late 1985, it seems to have had a slightly longer development, but still probably comes in at somewhat less than two years total. And it may even be basically eighteen months, it all kind of depends on when it was considered finished and they started the process of releasing it to manufacturing and creating the retail boxes and all of that.

InfoWorld, January 4, 1988

The timeline of Windows 2 and Windows/386 releases is also a bit murky, depending on how you choose to define “release date.” The first version of Excel for Windows was released in October of 1987, and it ran under the Windows 2.0 runtime, albeit one that was slightly older than the primary release version of Windows 2.0, which wouldn’t actually hit store shelves until January of 1988.

According to a quote from one of the Excel developers³⁹ “We wanted to get it out the door. We didn’t want to have to wait” for the final release of Windows 2. Hence the runtime probably had some bugs that weren’t present in the shipping product of Windows 2.0, however this is just a guess of mine. I’m also refraining from cracking any number of cheap jokes about bugs at Windows expense, because this is a serious video doggone it.

Compaq was apparently shipping Deskpro’s with Windows/386 preinstalled in October as well, however this was apparently a somewhat rushed and buggy version of Windows that required a lot of bug fixing. Presumably Compaq worked with Microsoft to iron these bugs out, its status as somewhat of a co-developer of Windows/386 probably helping this occur more quickly.

The retail version of Windows/386 also seems to have hit store shelves in January of 1988, alongside the normal version of Windows 2.0. However some sources will also cite a December release for both versions of Windows, and I saw one source claim November.

This is a good point to briefly summarize the various release dates that sources give for Windows 2.0 and why figuring out just when it actually was released has proven essentially impossible so far. Keep in mind that there are two different releases of Windows 2 to deal with, the 386 specific one that Compaq assisted somewhat with development and the standard one.

Technically there were three if we count the runtime version of Windows 2 that was released slightly before any other version, but let's not overcomplicate things. One other thing to keep in mind is that there seems to be a difference between when Windows 2 was available and when it was being sold as a retail product.

 The earliest claimed release date I can find for Windows/386 is September of 1987. Well that’s not quite true, the book Computer Wars: The Fall of IBM and the Future of Global Technology claims⁴⁰ that Windows 2.0 came out in 1986, which is definitely wrong and is so obviously incorrect that I tend to think it's a misprint and they meant to put 1987.

Both the website betawiki and os2museum state that Windows/386 came out in September 1987, with betawiki stating that the September 1987 release of Windows/386 was actually shipping with Compaq computers in September, before seeing its retail release alongside the initial release of the standard version of Windows 2, on December 7th, 1987.

Windows/386 box

However the os2museum website has a picture of both the front and back of a shrink-wrapped Windows/386 retail box that it states is from September 1987. I originally thought this might be something that only shipped out with Compaq DeskPro 386’s, but there is nothing I can see on the packaging with Compaq specific branding, it all looks like normal retail branding. 

My initial guess was that the build on the disks was possibly from September 1987, with the retail packaging itself from several months later, however since the box is still shrink-wrapped, the disks have clearly not been accessed to check for a date. The picture is from 2011 and is low enough resolution that I cannot quite read all of it so it's possible that the September 1987 date appears on it somewhere. Either that or the os2museum is saying its from September 1987 because that’s the date their source says Windows/386 was released to retail.

A commentator on the os2museum post about the Windows/386 box suggests that while this box looks like a retail box, it actually was a limited retail release that only shipped with Compaq Deskpro 386. I tend to think this is less likely since the box has no Compaq branding and in fact explicitly lists other competing systems as compatible with it. However it's also possible that Compaq didn’t want their name on any retail box, since I doubt they were getting any royalties from Windows/386 sales even though they played an unknown role in developing it.

Wikipedia says that both versions of Windows 2.0 were released to manufacturing on December 9, 1987, which would actually work with the January 1988 retail availability that InfoWorld reported in their January 4th, 1988 issue and is the same month that the book Gates cites as when the retail versions of Windows 2.0 hit store shelves. The book Hard Drive: Bill Gates and the Making of the Microsoft Empire actually ignores 1987 altogether as well as the 386 and standard versions of Windows 2 and just says that⁴¹ “In 1988, Microsoft had introduced a second version of Windows designed for computers with the powerful 386 chip.” Which I guess works with a January release date but might also be concatenating the release of Windows 2.1 in mid-1988 together with the original release of Windows 2.0.

Unless I have a very good reason to believe it's wrong, I will always trust a contemporary source over something written many years or decades after the fact, and the January 1988 InfoWorld article is about as contemporary as it gets, with Gates being published a scant five years later, in 1993. Of course Computer Wars was also published in 1993 and it completely messes up the date so…here is why it's critical to use as many sources as possible when researching. 

So my educated guess is that the retail versions of both versions of Windows 2.0 were available in retail stores in January of 1988 and not before. However there may have been some form of retail packaging for both versions of Windows 2.0 available from specific computer companies prior to this, or alternative distribution channels, and Compaq Windows/386 disks were definitely available in October 1987, probably using a September 1987 build of Windows/386. 

This timeline also works with what the book Barbarians Led By Bill Gates states,⁴² which is that Windows 2.03 was released on November 17, 1987, but was not in consumer hands until January 1988. If companies were shipping Windows 2 with their systems, Microsoft probably gave them priority over getting the retail version of Windows 2 into stores, especially since Microsoft’s original plan for Windows way back when Windows 1 was being developed, was to sell it to OEMs and completely avoid dealing with retail. 

And Microsoft might have tested out retail packaging for its shiny new operating system, even though this packaging wouldn’t have been seen in stores as of yet and would only be shipping with OEM systems. I will admit this feels like a stretch, but the existence of nicely packaged Windows 2 boxes that preexist the actual retail release would help explain some of the confusion as to when it was available to buy.

So to conclude, I am pretty sure Windows 2, in both of its versions, wasn’t on retail store shelves until January 1988, but was available through various other channels, and in several different version numbers from October 1987 on.

Arriving somewhere between one and three months late, this was still pretty good for Microsoft, both then and now.

The original tiled interface of Windows 1 that had been carried over to Windows 2 was gone, replaced by overlapping windows reminiscent of the Macintosh. Put a pin in this, we will be revisiting it shortly. In something of a minor miracle, Tandy Trower and his team had successfully gotten Windows 2 out the door fairly close to the original timeframe. 

Once this was done, most of the metaphorical lights for Windows 2.0 were turned out, with Barbarians Led by Bill Gates saying,⁴³ “A couple people were left to maintain the Windows 2.0 code and fix any bugs reported, but the “glory days” of Windows, such as they were, appeared to be dead and gone. Last one out, turn off the lights.”

Interested parties could get a sort of preview of Windows 2 by buying the first version of Excel for Windows, which launched in October 1987 running inside the first run-time version of Windows 2, exactly the same way that PageMaker had run inside a run-time version of Windows 1.

Excel 2.0 wasn’t the second version of Excel for Windows, although I guess you could count it as the second edition of Excel, with the first one being its original Macintosh version. However Microsoft decided to number it to correspond with the new Windows release, which as we stated earlier came about as a large reason because of the need for Excel to have a graphical user interface to run under. 

The question of whether Excel should come out for Presentation Manager first and then Windows, or even just for Presentation Manager had provoked fierce battles within Microsoft. However Jeff Harbers, who was overseeing the development of Excel for Windows, stubbornly held out for doing it on Windows first, in spite of Ballmer’s blustering about Presentation Manager being the critical platform for Excel to be on.

In the spirit of growing market share by whatever means possible, Microsoft also continued to make the run-time version of Windows 2 available to developers who wanted it, for the same low price that the Windows 1 runtime had cost…or in other words it was free. 

Microsoft also hedged its Windows 2 verbiage with some mildly self deprecatory statements that basically said that, while OS/2 and its Presentation Manager graphical user interface were obviously going to be super duper popular, since OS/2 was unfortunately running behind schedule, users could get a sort of preview of the Presentation Manager interface by buying Windows 2. 

Specifically, as per Infoworld’s April 6, 1987 article called “Windows 2.0 To Resemble OS/2 Windows Manager”, Windows 2 “would have a new visual appearance identical to the OS/2 Windows Presentation Manager”, with the article going on to say that “The new appearance is the result of a development agreement between Microsoft and IBM…Windows 2.0 offers the same external appearance as the Windows Presentation Manager, but its compatible with the existing applications and device drivers written for the current Windows.”

Gates was still clearly unsure about the continued viability of Windows, once the OS/2 juggernaut really got rolling. The marketing comes across as trying to have it both ways, both lauding Windows 2 as the new hotness while also meekly trying to say that it was a way to get a preview of what OS/2 would look like, once it came out and inevitably crushed all the competition. And was Windows actually OS/2’s competition? Well…Gates was carefully making sure to avoid positioning it directly at OS/2. He was still riding the IBM tiger and hoping he didn’t get bitten. 

As previously mentioned, Windows 2.0 came in two different versions, one targeted at the good old 8088 line of standard PCs, and another special version called Windows/386 that was solely aimed at PCs equipped with the fast 386 processor and required DOS 3.1 instead of DOS 3 as well as at minimum of 1 MB of RAM instead of 512k. 2 MB of RAM was recommended as a single MB of RAM made multitasking applications of any substance difficult. 

Windows/386 could make use of preemptive multitasking, something that the Mac operating system still was unable to do. Additional updates for Windows 2.0 over the original version of Windows included better memory management⁴⁴ and a move to a more Mac like style of user interface that allowed for overlapping windows instead of forcing all windows to tile. This particular feature would become extremely relevant about one second after Apple took offense to it.

The biggest advantage that Windows/386 brought to end users was probably the fact that it could take advantage of the 386’s Virtual 86 mode, enabling multitasking of both DOS and Windows applications in a more-or-less stable way. Or at least reasonably stable by 1987 computing standards, which are considerably different from modern ones. By which I mean they were a lot lower. 

Windows/386 allowed for multiple virtualized 8086/8088 MS-DOS sessions to run concurrently, each with its own 640k of RAM and the ability to utilize extended memory or EMS 4.0. From a user perspective, multiple DOS applications could be run simultaneously in overlapping windows, and if multiple concurrently running applications attempted to simultaneously access a device or port, Windows/386 was able to just popup a dialog⁴⁵ that asked the user to select which application got to make use of the device or port. 

Byte reviewed Windows 2.0 and Windows/386 in the May 1988 edition. As a side note, I really miss the days when computer magazines were beefy and 300+ pages thick. There is something special about flipping through page after page of glossy ads, detailed reviews, and in-depth features that just cannot be replicated on a website or scrolling through a PDF. Granted, I was only a few years old when Byte and other computer magazines had their heyday, but I love collecting and reading back issues.

Byte’s review of Windows 2 and Windows/386 was overall very positive, with Windows/386 coming in for the lion’s share of praise. Windows 2.0 was praised for being a significant upgrade over the first version of Windows, with Byte starting off its review by delivering a backhanded slap to Windows 1 stating⁴⁶ “[t]he first version of Windows was slow; it did not support concurrency, nor did it directly use more than 640k bytes of memory. Version 2.03 (99$) is faster than the original Windows product, and makes the best of microcomputers without the 80386 chips. The 386 version ($195) offers true multitasking, thanks to the power of the 80386.”

A word on pricing, whereas Byte’s contemporary review of both versions of Windows 2 stated that the price was 99 dollars for Windows 2 and 195 dollars for Windows 2/386, the book “When Computing Got Personal” states⁴⁷ that both versions of Windows 2 retailed for the same price of 99 dollars. I’m not sure about the reason for the discrepancy, but it makes sense that Microsoft would seek to make Windows 2 as attractive as possible, no matter the version.

Definitely a solid review, but Windows/386 was clearly where the main excitement was, with the author stating in the article wrap-up that Windows/386 was not only a solid product that really helped users make better use of the power of the speedy new 386 chips, but also provided OS/2 with some serious competition: “Windows/386 has a greater appeal for the users of 80386 machines, due to its ability to support multitasking to break the 640k-byte memory barrier (making use of expanded or extended memory). Indeed, Windows/386 is a serious rival for OS/2. While OS/2 is seen as being weighed down by its compatibility with the 80286 chip, the Windows/386 system taps into the power of the 80386 tasking software compatibility, taking it to a new level.“⁴⁸

And since it is absolutely required to take a look at what The Computer Chronicles has to offer on the subject of Windows 2 and Windows/386, we find an episode that aired on August 16th, 1988, where a three minute segment was devoted to Windows/386 in particular. Gary Kildall kicked things off by asking the Microsoft representative what the difference was between OS/2 and Windows 386, rather interestingly crediting Microsoft as the one behind both of them, and not mentioning IBM.

Let’s listen to the Microsoft rep’s response here, because it is an excellent example of how murky Microsoft's strategy was at this point and how hard they were trying to spin things, downplay things, and in general muddy the waters as much as possible. 

So let’s unpack exactly what was said here, because it really boils down to three specific statements. First of all, we are told that the same user interface is used across OS/2, Windows/386, and the standard version of Windows 2. It is implied that this is all due to IBM’s SAA and CUA standards being implemented across both OS/2 and Windows 2.

Secondly we are told that because Windows/386 is built solely for the 386 processor, it is a “short-term” product because it won’t run anywhere else beyond the 386 and thus is somehow a dead-end.

Thirdly we are told that the platform of the future will be OS/2, because OS/2 is “built for the 286 processor and beyond and it really is the operating system and a platform for the future”, whatever that means. 

Watching this clip you really do get a good idea of how unsure Microsoft was about Windows. It was still viewed within the company as a product that most likely had a limited lifespan ahead of it, but one that could still be useful for the time being for people who had 386 machines. But critically, Microsoft still wasn’t seeing Windows as a platform, they saw it as a product. OS/2 was seen as the platform, one that would be built on, extended, and improved, and would be the dominant operating system of the future. 

And I have to give the Microsoft rep, Regional General Manager Dave Jaworski, credit here for doing a solid job of dancing along the company tightrope, keeping the conversation smoothly moving along, and giving a good Windows/386 demonstration.

Jaworski worked for Microsoft for ten years and has written a book about his experience with, among other things, launching Microsoft Office. I recently bought the book, but have not had the time to thoroughly read it yet as of this recording, but it looks interesting. Future me editing this may have some other notes on it.

Also Gary Kildall calls Windows an operating system here, not merely a graphical shell for DOS. Which will of course settle that particular argument once for all and I expect zero dissension to occur in my comment section.

By the end of 1988, Windows was shipping somewhere in the vicinity of fifty thousand copies a month.⁴⁹ Hardly setting the world on fire, and still far removed from the original boastful comments made prior to the original Windows’ release, but still at least a modest success. 

According to Steve Ballmer, the main thing keeping Windows alive at this point was actually Microsoft’s own Excel, with Ballmer specifically stating⁵⁰ that “Excel single-handedly was giving mouth-to-mouth to Windows”, although PageMaker was probably also helping out a bit.

It's also fair to point out that since Excel for Windows utilized the Windows runtime, I’m not sure how exactly how much help it was being when you break it down versus people buying Excel and using the Windows runtime to use it, versus people buying Windows and then buying Excel to run on it. And I have not really found any good information as to how that split went. Or how those are counted in the fifty thousand copies a month, or if they are counted. 

Mind you, I can’t imagine PageMaker was helping out more than a small amount as it seems hard to believe that there were many people passing up a Macintosh with PageMaker in favor of buying PageMaker for a DOS/Windows machine. But maybe I’m wrong and there were tons of desktop publishing outfits in the late 1980s that turned their noses up at the Macintosh and PageMaker and were all about Windows and PageMaker. If you were someone who used the Windows version of PageMaker in a production environment back in the day, I’d love to hear from you in the comments about your experience with it.

Interestingly, this time period also saw at least one attempt to build a digital editing and compositing application utilizing Windows 2. I’m not going to go too far in the weeds on this, as the history of digital video editing is going to be the focus of a future series on this channel, but it is worth mentioning in passing.

A company called Digital F/X showed off a new product at the 1988 SIGGRAPH. This system, formally called the DF/X 200 Digital Production System, was⁵¹ “a video post-production system aimed at editors who wanted to enhance their material using special effects, paint and video typography tools included with the system’s software…” 

This was a turnkey solution of both hardware and software, with the hardware based around a 386 processor, and some version of Windows. However, Windows was so limited at this point, especially with regards to video, that Digital F/X’s engineers actually had to write their own low level DOS.

According to one of the engineers, Chuck Clarke, “Of course we were using the latest 386 chips because they could handle multitasking, that’s why we had chosen them but the O/S wasn’t as capable so we had to write a custom version of DOS. The API layer, which we called the Virtual Video Interface (VVI) separated the hardware from DOS and Windows. When you switched from say the real time effects mode to the paint system you were switching state from the Effects DOS to the Paint DOS to the title DOS all with Windows on top.”

One year later, at the 1989 NAB, Digital F/X launched a new turnkey product called Composium. This was a compositing suite that cost 197,000 dollars and still made use of Windows 2.0 for the interface. And it was apparently still very messy code, thanks to Windows. According to Robert Berger, who joined Digital F/X when they were trying to upgrade Composium from Windows 2.0, “The first engineering team had built the Composium system on Windows 2.0 and going forward that was untenable. They had used an incredible amount of hacks just to make the thing work and they had layered on so many changes without having the time to go back and clean up the code.”⁵²

None of this really plays into the history of Windows as a whole, and in fact Digital F/X was looking to move away from Windows for future versions of Composium. Systems that cost hundreds of thousands of dollars were never going to sell in large enough numbers to drive Windows adoption at all, it's not like Microsoft made any money off of the systems that did sell.  

Still, the fact that such an early version of Windows in any form was the target for one of the pioneering systems for video editing and digital effects is an interesting footnote, at least to me. 

Windows 2.1

There was an often forgotten update to the Windows 2 line that was released in May of 1988, just six months after Windows 2.0 came out. Windows 2.1 was a decent step forward for the platform, far more than its slight version number increase would indicate, and came in two versions, one optimized for the 286 processor and one solely targeted at the 386, just like the preceding Windows/386. It also required a hard drive, the first version of Windows to do so. 

However, Tandy Trower was no longer overseeing the project, having left the Windows team after the release of Windows 2.0. His reasons for doing so were simple…Windows 2.0 was supposed to be the final version of Windows, so why stay in a dead end job? 

In Trower’s words,⁵³ “From Steve Ballmer’s perspective, these were intended to be the last versions of Windows, with OS/2 (and its Presentation Manager) replacing them. So it was clear that I either would need to find a way to transfer into a role on the already staffed OS/2 team or find a new job. Since I had become increasingly aware of the need for improving our overall design of Windows and Windows applications, I opted for defining a new group that would be devoted to four things: 1) employing real graphics designers (not developers) to create those interfaces, 2) establishing usability testing facilities and services, 3) defining guidelines for good UI design and consistency for Windows applications, and 4) developing UI beyond the current product development cycles to further evolution of Microsoft’s user interfaces.

I went to Gates with that proposal. As Gates also felt that we needed to do a better job on our products’ user interfaces, he agreed and I transferred over to report to Microsoft’s new VP of Applications…” 

Information on just why the decision was made to press forward with a new version of Windows is scarce to nonexistent. Trower did tell me that he was told that⁵⁴ “whatever plans were in process for post 2.0 would be less about UI and more about improving performance and memory access.” Windows was clearly on the way out, but some updates might be made to it that would give users a somewhat better experience. This is where Windows 2.1 comes in.

Most sources, if they mention version 2.1 at all, give no details on the impetus for its development or who headed up the development. Tandy himself isn’t even sure who headed it up, or if there even was a formal replacement for him, saying⁵⁵ that  “...post 2.0(3), again I had pretty much transitioned out of the day-to-day work on Windows. I had enough to occupy my time between trying to get all of Microsoft to get to a consistent UI design (the hardest turned out to be the PowerPoint team) and my time spent trying to negotiate commonality of UI between Windows and OS/2.”  Other major sources I consulted, such as Gates, skip right over it and go straight to Windows 3.0. Barbarians Led By Bill Gates also omits any mention of Windows 2.1, as does Hard Drive: Bill Gates and the Making of the Microsoft Empire. 

So my guess is that this was viewed as a small upgrade to the already released version 2.0, worthy of very little coverage and almost always lost in the shadow of Windows 3.0. Most likely this is another case of Gates hedging his bets and keeping Windows ticking along, even though he still figured OS/2 was eventually going to drive it out of the marketplace. With a skeletal crew of programmers, the project most likely ran along quietly so far as Microsoft was concerned, with little attention paid to it. 

It's even possible that Gates simply didn’t have anything else for the Windows programmers to do, and rather than lose them, merely let them keep quietly beavering away on what was felt to be a minor upgrade. As always, I welcome any comments that can shed light on this often overlooked iteration of Windows. 

InfoWorld previewed both versions of Windows 2.1 in their July 4, 1988 edition, saying “Microsoft last week released Windows upgrades that demand 286 or 386 power but return greater memory, enticing developers and users who predict renewed interest in the DOS-based graphical interface. The new Windows/286, Version 2.1, which replaces Windows 2.03, provides an extra 50k above the 640K DOS limit when running on an AT-class system with more than 1 megabyte of extended memory available. The 99$ Windows/286 still can run on 8088/8086-based systems but without the new memory features.”

Although the 286 processor was definitely severely compromised compared to the 386, when running on it Windows 2.1’s ability to store 64 kilobytes of itself was a nice trick for squeezing more performance out. 64 kilobytes may not sound like much, but that was an additional 64 kilobytes available for applications to use, something that could be very useful to the memory constrained applications of the era. As the quote from InfoWorld mentions, in spite of the name Windows/286 was still capable of running on an original 8088 based computer, but would have no ability to access extended memory. And as always, Windows on an original PC or clone would be a very poor experience.

Regarding Windows’ performance on 286 based machines, keep in mind that Commodore was still merrily selling the Commodore 64 at this point, so to put things into better perspective let's say that Windows/286 was able to give its users an extra Commodore 64’s worth of RAM. The modern take on this would be Windows 11 somehow granting you an entire Chromebook’s worth of extra RAM, which if we assume is about 4 GB, would enable you, the happy user, to open up to a single additional tab in Chrome.

I’m guessing that this is probably David Weise’s work, along with all of the memory improvements in Windows 2x. Weise would really come into his own with his work on Windows 3.0, but it's important to remember that prior to Windows 3.0 he also made significant contributions to improving Windows 2x’s memory handling. 

Let me address one thing here though, the discrepancy in the amount of RAM that Windows/286 was able to store in an AT’s High Memory Area, was it 50 kilobytes or 64 kilobytes? I am honestly unsure, as the InfoWorld article says 50 kilobytes, but quotes Microsoft’s general manager for the DOS and Windows unit, Russ Werner, as saying that the next generation of EMS plus an upgraded version of Windows/286 would increase that 50 kilobytes freed up to a full 64 kilobytes.

My guess is that this may be referring to EMS 4.0, but one of my other sources, When Computing Got Personal directly states that it was 64 kilobytes freed up and makes no mention of 50 kilobytes at all. Feel free to comment below if you can clear this up. 

Regardless of how much RAM could be freed up on 286 based machines, Windows/286 was a nice step forward for the platform and with its reasonable retail price of 99 dollars, paired with OS/2’s continued struggles to get off the ground, it helped Windows’ sales keep steadily increasing. Nothing spectacular or eye-catching, but that quiet growth was one of the reasons why so few people saw how successful Windows was going to wind up being.

Windows 2.1 also came in an updated 386 specific version, still just known as Windows/386 just like the Windows 2.0 386 version. According to the contemporary InfoWorld coverage, it cost twice as much as Windows/286, or 199 dollars, although according to When Computing Got Personal, it cost the same 99 dollars⁵⁶ as Windows/286.  

The new versions of Windows sold steadily, and both consumer and developer interest in Windows as a platform was picking up. A virtuous circle was forming, where more people were running Windows, meaning it was a more attractive target for developers to target, and as more Windows applications hit the marketplace, it attracted more people to buy Windows.

The fact that Microsoft gave away the Windows runtime version to any developer that wanted it, also gave a significant boost to developers and should not be underestimated as a clever marketing tactic. Albeit I am fairly sure it was always a marketing tactic born out of desperation, as Microsoft tried anything they could think of to entice more developers and users.

The most popular runtime version of Windows 2.1 appears to have been based on Windows 2.11 by the way.⁵⁷ It also seems like most of the applications for Windows, whether using the runtime or not, were still primarily coming from smaller developers or solo developers. Major software development companies such as WordPerfect, Lotus, or Borland were still uninterested in creating Windows versions of their flagship DOS applications. This attitude was going to badly bite them in the butt in just a couple more years.

By 1989, the two versions of Windows 2 had sold about two million cumulative copies,⁵⁸ which may have been dwarfed by the 5 million copies that MS-DOS sold every year, but was still unquestionably a successful product that had moved from being the butt of industry jokes to ever increasing recognition and a certain amount of market penetration. 

Byte magazine also singled out Windows/386, presumably the 2.1 version, for praise, giving it a Byte Distinction award in the January 1989 issue. It was a short paragraph, and I am going to quote it in its entirety here,⁵⁹ ”Windows/386 was one of the first programs to take advantage of the advanced architecture of the 80386 microprocessor. As such, it is serious competition for OS/2. While OS/2 is held back because it was designed to run on 80286 machines, Windows/386 taps into the power of the 80386 to support multitasking of DOS applications and break the 640K-byte memory barrier. The graphic user interface of Windows/386 resembles that of Presentation Manager, a benefit for those considering making the switch to OS/2.” 

This award is slightly lessened when one considers that OS/2 got a Byte Excellence award and several paragraphs in the same issue, but still shows that Windows was no longer the laughingstock it had been just four years prior. Windows 2x, in all of its variants, had made steady progress and was starting to be taken seriously.

And it was far outstripping OS/2’s sluggish sales, both by being far more affordable, and by having lower system requirements. And Microsoft was just plain better than IBM was when it came to marketing to the home computer market. IBM seemed to prefer to primarily focus on the corporate market, and priced accordingly, with the OS/2 1.0 Extended Edition coming out at a painfully high retail price of 795 dollars⁶⁰ in 1987.

This, contrasted with Windows' far lower pricing and pursuit of bundled deals with OEMs, meant that while comparatively few people were seeking GUIs for their PCs, the bulk of the ones that did were primarily winding up with Windows. 

So what exactly was the plan for Microsoft, Windows, IBM, and OS/2 and how were these companies and products supposed to interact and work with each other? OS/2 was being jointly developed with Microsoft, but Microsoft also had Windows, which was competing directly against OS/2, even while Microsoft was supposedly working on the next generation of OS/2, with a team headed by recent acquisition Dave Cutler working on what was called “NT OS/2”. It was all very confusing, especially to a software developer trying to figure out which one to target for a given application.

And there was turmoil going on at IBM as well, as Bill Lowe left his position overseeing IBM’s PC division and was replaced by long time IBM veteran and actually pretty neat guy, Jim Cannavino. I did an overview of Cannavino in the Fall of OS/2 video, and I won’t repeat myself here, suffice it to say that Cannavino was a hard driving, intelligent man who was also a formidably talented programmer in his own right.

However Cannavino was not a PC guy, he came up through the ranks in IBM’s mainframes, and his whole programming style was the product of that particular environment. In other words, he came from a culture that valued carefully thought out, carefully implemented, carefully documented, and painstakingly methodical code over clever tricks and difficult, hard to maintain code that nevertheless accomplished remarkable things. 

Cannavino and Gates were complete opposites as programmers, and although neither one was in the trenches writing code anymore, they were both hard driving men who were bound to knock heads. Especially once Cannavino got his head around what exactly was going on with OS/2 and where exactly IBM stood in relation to Microsoft under the 1985 JDA.

HP NewWave

HP NewWave

And there was another problem on the horizon. And to understand what happened we need to briefly discuss an alternative GUI for Windows called NewWave, from HP. Yes you heard right, just as Windows gave MS-DOS a graphical interface, Hewlett Packard’s NewWave was a graphical interface for Windows, one that replaced Windows 2’s desktop interface with a new, object oriented one of its own. 

NewWave is itself a separate topic well worth discussing in its own video, as it had a surprisingly lengthy lifespan, lasting all the way up until 1995 as an active product. But for purposes of this video, I am just going to hit the bare minimum required to put it into context and why its release caused Apple’s Eye of Sauron to focus on Microsoft.

NewWave was designed to run under the retail version of Windows 2.0, version 2.03 and Microsoft had apparently licensed out elements of Windows, such as overlapping windows, and according to the NewWave product manager, Bill Crow, Microsoft had also assisted in its design and development. And Crow made the mistake of telling this to an Apple employee, a man named Rappaport.

Apple was not happy to hear that Microsoft was licensing out various interface elements that Apple viewed as its own intellectual property, and thus Apple decided to go all-out on the offensive. Hence the lawsuit charging Microsoft and HP with infringing on Apple’s intellectual property.. 

Apple versus Microsoft

Apple’s reaction to Windows 2.03 was a rather significant cloud hanging over Microsoft, one that had the potential to destroy Windows right as it was gaining momentum.

But this lawsuit wasn’t coming from a vengeful Steve Jobs, as he was no longer with Apple and in fact had sold off all of his Apple stock except for a single share. No, this was the John Sculley era of Apple, the CEO who had signed a rather unwise agreement with Microsoft in which he gave them permission to use certain aspects of the Macintosh interface in exchange for Microsoft's continued work on applications for the Macintosh, especially Excel.

But how far did this agreement stretch and what did it cover? Apple felt that it only covered the original version of Windows, version 1 whereas Microsoft felt that the agreement covered all versions of Windows going forward. Or at least those were basically the two companies’ official stated positions as the legal battle kicked off.

Yes HP was part of the suit as well, however it was pretty much Apple versus Microsoft and the NewWave issue was a side issue compared to Windows 2.03’s overlapping windows or so that’s how Apple framed it.

In 1988 Apple Computer formally started the ball rolling by charging that Windows 2.03 infringed in over 180 different ways⁶¹, one major way being its shift from tiled windows to overlapping windows. 

Windows wasn’t Apple’s first target, as they had previously warmed up by successfully  attacking the long-suffering Digital Research in 1985, charging that their GEM graphical user interface was too similar in look-and-feel to the Macintosh operating system. Before things went very far, Digital Research capitulated,⁶² agreed to modify a number of things in GEM, and forked over a certain amount of money to Apple. 

So Apple was coming off of a win…somewhat, when they decided to drop the hammer on Windows a few years later, and they were determined to hit Microsoft as hard as they could. Of course Microsoft was a far more powerful company than Digital Research was…and Bill Gates was far more ruthless than the gentlemanly Gary Kildall, one of the all-time class acts of the computing world.

After almost two years of going back and forth, those 180 plus claims were reduced down to just ten. Basically,  on July 25th 1989, Judge William Schwarzer ruled⁶³ that all but ten of the claimed infringements were covered under the license that Apple CEO John Sculley had signed with Microsoft in 1985, and the remaining ten were ideas that were too broad to be copyrighted. 

Given Tandy Trower’s oversight of Windows 1 and 2.0’s development, and the major role he played in changing Windows from a tile based interface to one that used overlapping windows, he was also kept quite busy by the trial. In Trower’s words, “Having been a principal manager in charge during development of Windows 2.0, I was now caught up in the maelstrom and over the next year I got a thorough education on the US legal process as I briefed the Microsoft legal team, created exhibits for them, and was grilled by deposition by the other side. To me the allegation clearly had no merit as I had never intended to copy the Macintosh interface, was never given any directive to do that, and never directed my team to do that.”⁶⁴

Trower also rather dryly states that “the allegation was almost insulting. If I wanted to copy the Macintosh, I could have done a much better job.” 

Expanding on this, he told me that⁶⁵ “Apple's lawyers probed whether I intentionally "copied" from the Apple GUI during my days of managing Windows. But that had never been my goal and I was somewhat insulted by the implication that similarities were based on copying. Further, I had never been made aware of the license that Gates had negotiated that actually granted Microsoft rights to do so, if we had so chosen. From my perspective, if there was inspiration, it was more on the various forms of GUI that I had seen first on the Xerox Star and other GUIs.”

Thanks to appeals, the case would drag on for six more years, until February 21, 1995, when the Supreme Court refused to hear it and the saga finally came to a close. By 1995 of course, the computing world looked very different from 1988, and a flailing Apple was struggling to stay in business, while IBM was busy putting the finishing touches on its total destruction of OS/2 as a viable product.

But that resolution was years in the future. In 1989, Microsoft was still very much under the cloud of what the end result of the lawsuit could be, while Gates was getting more and more determined to keep pushing forward with Windows. The first small taste of success he had seen was whetting his appetite for more. Additionally, if Gates could make Windows into an industry standard via extensive sales into the consumer market, it seemed unlikely that the courts would feel able to order it removed, even if the Apple lawsuit started going against Microsoft.

IBM and OS/2 were still hanging over Microsoft’s head as well, but OS/2 had so far failed to catch any traction, and Windows was definitely outselling it with ease. More importantly, according to the book, The Making of Microsoft, 1989 also saw Windows reaching the surprising heights of becoming Microsoft’s best selling single product.⁶⁶ Customers were snapping up applications such as Excel, SuperBase, Designer, Omnis, and CorelDraw in ever increasing numbers, all of which required Windows. 

And Windows’ market penetration was increasing even more than these 50,000 copies sold a month would indicate, thanks to the increasing number of Windows applications that took advantage of the free run-time version of Windows 2. Plus piracy of course, as these early versions of Windows were essentially unprotected from being spread around as widely as possible.

A fact that I have little doubt was deliberate, as Gates sought to disseminate Windows as widely as possible. My guess is that to Gates, someone running a pirated version of Windows was better than someone not running Windows at all.

This was a lot of money flooding into Microsoft's coffers, and more importantly it was a revenue stream that was only increasing. Microsoft was also hard at work on the mammoth task of converting Microsoft Word to run under Windows, something that could hopefully be expected to give a further boost to Windows sales. 

And the number of application developers starting to climb onto the Windows bandwagon, while still not anywhere near the number of developers creating MS-DOS applications, was continuing to grow. In fact, more applications were being developed for Windows than for any other graphical environment. Or so Bill Gates claimed⁶⁷ at a January 1989 meeting in Paris that he spoke at. Hard to say if he was telling the truth or not, but I have no reason to disbelieve him, especially not in light of what was about to happen with Windows.

For balance, I will point out that the GUI competition was quite fragmented at this point, with platforms such as the Commodore Amiga, the Atari ST, the Macintosh, GEM, and of course OS/2 plus several others all competing for the attention of developers wishing to create applications that ran under a graphical environment. And the free runtime version of Windows undoubtedly played a huge role in helping Windows come out on top. IBM would never dream of giving away a free version of OS2 in any form.

Given Windows' steadily growing success, even with the reduced resources that had been allocated to it in comparison with OS/2 and Presentation Manager, Bill Gates was starting to see the outline of a future where he was no longer IBM’s junior partner, but rather was completely independent from them. But how to get off the back of the IBM tiger without getting bitten?

As the release of Windows 2.1 passed, and it became clear that demand for Windows itself as well as Windows based applications was growing, Gates decided to refocus his attention on the next version of Windows and actually make it a Microsoft priority. Its development would be given a lot more resources than had been lavished on Windows since the height of the development of the original version of Windows. This newest release, Windows 3.0, would become the biggest release of Windows yet and the first version of Windows to become an overwhelming success, even though it tends to be overlooked in light of the following release, Windows 3.1. 

Undeterred by the Apple lawsuit, or the uneasy alliance with IBM and the continuing joint development of OS/2 and Presentation Manager, which most people still expected to be the eventual winner and standard setter for the 1990s, Gates chose to aggressively push ahead with Windows development, and make 3.0 far and away the biggest and fanciest version of Windows yet. 

However, ever the careful gambler, Gates also decided to continue putting considerable resources into OS/2 and Presentation Manager. He wanted to try to find an accommodation with IBM that would hopefully let Windows and OS/2 coexist, presumably so that if OS/2 suddenly took off and blew past Windows, Gates could sideline Windows and continue his relationship with IBM and the continued joint development of OS/2. 

It was a high wire balancing act, and the tensions between IBM and Microsoft were only ratcheting higher and higher as the 1980s began to draw to a close and the newest version of Windows began to shape up for release. Windows 3.0 would be a release that would be unlike anything that had been seen before for Windows and the industry. One that would shape the company’s direction in a radical new direction, destroy the IBM relationship, and put Windows firmly on its path to industry domination. And it started, because of a fortuitous hire that Microsoft had made when it purchased DSR several years prior.

And that is where we will pick up the story in the next installment of this series. I want to give a huge shout out to Tandy Trower for being kind enough to exchange multiple lengthy emails with me, filled with lots of interesting details about his time with Microsoft and overseeing Windows’ development.

Thanks for reading. If you reading this lengthy script, please consider subscribing to the Substack.

1. Ichbiah, D., & Knepper, S. L. (1991). The Making of Microsoft: How Bill Gates and his team created the world’s most successful software company. Page 190

2. "Palantir Readies Windows Filer 3.0", InfoWorld July 21, 1986
https://books.google.com/books?id=Yi8EAAAAMBAJ&q=Trower#v=onepage&q&f=false

3

Nicholson, M. (2014). When computing got personal: A History of the Desktop Computer. Page 144

4

Manes, S., & Andrews, P. (1994). Gates: How Microsoft’s Mogul Reinvented an Industry--and Made Himself the Richest Man in America. Touchstone. Page 319

5. Ibid

6. - Trower, Tandy. Re: "Tandy Trower role in Windows 2.1", received 10/8/2023

7

Eller, M. (1998). Barbarians Led by Bill Gates: Microsoft From the Inside: How the world’s richest corporation wields its power. Page 69

8. - Trower, Tandy. Re: "Tandy Trower role in Windows 2.1", received 10/8/2023

9

Wallace, J. M., & Erickson, J. E. (1992). Hard drive: Bill Gates and the making of the Microsoft empire. Page 346

10. Manes, S., & Andrews, P. (1994). Gates: How Microsoft’s Mogul Reinvented an Industry--and Made Himself the Richest Man in America. Touchstone. Page 285

11. Wallace, J. M., & Erickson, J. E. (1992). Hard drive: Bill Gates and the making of the Microsoft empire. Page 346

12. Manes, S., & Andrews, P. (1994). Gates: How Microsoft’s Mogul Reinvented an Industry--and Made Himself the Richest Man in America. Touchstone. Page 318

13. Manes, S., & Andrews, P. (1994). Gates: How Microsoft’s Mogul Reinvented an Industry--and Made Himself the Richest Man in America. Touchstone. Page 346

14. Ibid

15. Ibid

16. Wallace, J. M., & Erickson, J. E. (1992). Hard drive: Bill Gates and the making of the Microsoft empire. Page 341

17. Manes, S., & Andrews, P. (1994). Gates: How Microsoft’s Mogul Reinvented an Industry--and Made Himself the Richest Man in America. Touchstone. Page 311

18. Wallace, J. M., & Erickson, J. E. (1992). Hard drive: Bill Gates and the making of the Microsoft empire. Page 342

19. Manes, S., & Andrews, P. (1994). Gates: How Microsoft’s Mogul Reinvented an Industry--and Made Himself the Richest Man in America. Touchstone. Page 311

20. Wallace, J. M., & Erickson, J. E. (1992). Hard drive: Bill Gates and the making of the Microsoft empire. Page 342

21. Eller, M. (1998). Barbarians Led by Bill Gates: Microsoft From the Inside: How the world’s richest corporation wields its power. Page 77

22. Manes, S., & Andrews, P. (1994). Gates: How Microsoft’s Mogul Reinvented an Industry--and Made Himself the Richest Man in America. Touchstone. Page 321

23. Manes, S., & Andrews, P. (1994). Gates: How Microsoft’s Mogul Reinvented an Industry--and Made Himself the Richest Man in America. Touchstone. Page 323

24. Eller, M. (1998). Barbarians Led by Bill Gates: Microsoft From the Inside: How the world’s richest corporation wields its power. Page 85

25. - Trower, Tandy. "Re: Tandy Trower role in Windows 2.1", received October 3, 2023

26. Ibid

27. Eller, M. (1998). Barbarians Led by Bill Gates: Microsoft From the Inside: How the world’s richest corporation wields its power. Page 85

28. - Trower, Tandy. "Re: Tandy Trower role in Windows 2.1", received September 13, 2023

29. Ibid

30. Manes, S., & Andrews, P. (1994). Gates: How Microsoft’s Mogul Reinvented an Industry--and Made Himself the Richest Man in America. Touchstone. Page 347

31. Wallace, J. M., & Erickson, J. E. (1992). Hard Drive: Bill Gates and the making of the Microsoft Empire. Page 347

32. Manes, S., & Andrews, P. (1994). Gates: How Microsoft’s Mogul Reinvented an Industry--and Made Himself the Richest Man in America. Touchstone. Page 348

33. Nicholson, M. (2014). When computing got personal: A History of the Desktop Computer. Page 94

34. Wallace, J. M., & Erickson, J. E. (1992). Hard Drive: Bill Gates and the making of the Microsoft Empire. Page 347

35. Wallace, J. M., & Erickson, J. E. (1992). Hard Drive: Bill Gates and the making of the Microsoft Empire. Page 348

36. Wallace, J. M., & Erickson, J. E. (1992). Hard Drive: Bill Gates and the making of the Microsoft empire. Page 349

37. Ibid

38. Manes, S., & Andrews, P. (1994). Gates: How Microsoft’s Mogul Reinvented an Industry--and Made Himself the Richest Man in America. Touchstone. Page 331

39. Manes, S., & Andrews, P. (1994). Gates: How Microsoft’s Mogul Reinvented an Industry--and Made Himself the Richest Man in America. Touchstone. Page 347

40

Ferguson, C., & Morris, C. (1993). Computer Wars: the fall of IBM and the future of global technology. page 81

41. Wallace, J. M., & Erickson, J. E. (1992). Hard drive: Bill Gates and the making of the Microsoft Empire. Page 357

42. Eller, M. (1998). Barbarians Led by Bill Gates: Microsoft From the Inside: How the world’s richest corporation wields its power. Page 86

43. Ibid

44. Nicholson, M. (2014). When computing got personal: A History of the Desktop Computer. Page 145

45. Nicholson, M. (2014). When computing got personal: A History of the Desktop Computer. Page 146

46. Byte, May 1988, "Microsoft Windows 2.03 and Windows/386", pages 153-157

47. Nicholson, M. (2014). When computing got personal: A History of the Desktop Computer. Page 146

48. Byte, May 1988, "Microsoft Windows 2.03 and Windows/386", pages 153-157

49. Nicholson, M. (2014). When computing got personal: A History of the Desktop Computer. Page 147

50. Manes, S., & Andrews, P. (1994). Gates: How Microsoft’s Mogul Reinvented an Industry--and Made Himself the Richest Man in America. Touchstone. Page 377

51

Buck, J. (2009). Timeline Analog 4. Tablo Publishing. Page 77

52. Buck, J. (2009). Timeline Analog 4. Tablo Publishing. Page 147

53. https://technologizer.com/2010/11/20/the-secret-origin-of-windows-2/3/

54. - Trower, Tandy. Re: "Tandy Trower role in Windows 2.1", received September 18, 2023

55. - Trower, Tandy. Re: "Tandy Trower role in Windows 2.1", received September 14, 2023

56. Nicholson, M. (2014). When computing got personal: A History of the Desktop Computer. Page 146

57. InfoWorld, May 28, 1990, page S46

58

From airline reservations to Sonic the Hedgehog: a history of the software industry. (2003). Choice Reviews Online, 41(02), 41–0968. Page 250

59. Byte, January 1989, Page 383

60. Nicholson, M. (2014). When computing got personal: A History of the Desktop Computer. Page 146

61. The Computer Chronicles, Windows 3.0, Aired October 4, 1990

62. https://www.nytimes.com/1985/10/01/business/digital-research-to-modify-gem.html

63. https://legal.thomsonreuters.com/blog/1988-apple-sues-microsoft/

64. https://technologizer.com/2010/11/20/the-secret-origin-of-windows-2/3/

65. - Trower, Tandy. Re:"Tandy Trower role in Windows 2.1". received September 13, 2023

66

Ichbiah, D., & Knepper, S. L. (1991b). The Making of Microsoft: How Bill Gates and his team created the world’s most successful software company. Page 238

67. Ibid