As far as I understand it, the whole personal computing revolution that Microsoft Windows did was not entirely by its own design. Is it true that the Windows OS at its core was originally designed to simply be the OS of the terminals of the Windows server architecture? Similar to how Fedora is at its core intended to be the OS of the Red Hat Enterprise Linux servers?
Comments are not for extended discussion; this conversation has been moved to chat.– Chenmunka ♦May 6, 2020 at 15:12
You removed the anecdote about Windows Update from your question, but if you post a separate question about that you might get some interesting answers...because the anecdote is almost completely backwards.– user3067860May 6, 2020 at 17:20
2It needs to be remembered that "DOS" was originally "QDOS" -- "Quick and Dirty Operating System" marketed by Seattle Computer Products, then later sold to Microsoft. And IBM had a windowing UI (can't recall if it was OS/2 or DOS based) before Windows became a thing, but IBM never pushed it.– Hot LicksMay 7, 2020 at 0:48
Game setup. Please select your sound card: Soundblaster, Gravis, AWE 32. It was that kind of pain and the fact that MS were in the right place at the right time.– FrankMay 7, 2020 at 6:45
@HotLicks: I'm somewhat confused – you're not referring to IBM's 1988 Presentation Manager for OS/2, are you? Because I always thought that this was IBM's first GUI for PCs. But Presentation Manager is basically a contemporary of Windows 2.0.– SchmuddiMay 7, 2020 at 10:17
The short version is that Windows became the de facto operating system thanks to Microsoft’s business acumen (or shenanigans, depending on your point of view), marketing, skilled developers, a strong focus on backwards-compatibility, and the success of MS-DOS.
The success of Windows in general can be traced back to the success of Windows 3.0, which has already been addressed here. I’ll try to explain how we ended up with a “server-oriented” operating system, Windows NT, as the de facto operating system on home PCs nowadays.
Early versions of Windows (in the first half of the eighties) were developed in the context of a perceived “rise of the GUIs” (see also the Apple Lisa and Macintosh). They weren’t all that successful, until some skilled developers implemented new features in Windows which took full advantage of newer CPUs (286 and up), leading to the surprise success of Windows 3.0. The latter gave rise to a large ecosystem of software, which helped cement the Windows advantage: users started using Windows not only as a graphical shell, or a GUI for a small number of applications, but for most if not all of their computing needs. This led PC manufacturers to include Windows by default with their PCs (where previously many would only ship DOS), helped later on by Microsoft’s contracts which strongly encouraged bundling. PCs changed from “DOS PCs” to “Windows PCs” by default.
In parallel, Microsoft had been working on a replacement operating system for DOS for a long time; not Xenix, but OS/2, and after the fall-out with IBM, NT. This ended up targeting high-end PCs of the time, taking full advantage of 32-bit CPUs, with a high-level architecture resulting in a more portable and maintainable operating system, but also a more resource-intensive operating system. As a result, it was marketed for servers and workstations initially, effectively as a competitor to Unix (and mini-computer operating systems).
The success of Windows 3.0 started the ball rolling on the unification process (which would take a long time): Windows NT switched from an OS/2-based API to a new 32-bit API, Win32, based on the existing 16-bit Windows API (this happened before the first release of NT, 3.1). But NT was still too big for home (and office) computers, and was really bad at doing one important thing for home users, playing games. The UI was very similar across all three operating system lines at the time — compare OS/2 1.3, Windows 3.0, Windows 3.1, and Windows NT 3.1.
The parallel development streams continued: Windows 3.0 was extended with multimedia features (MPC and then Windows 3.1) and, importantly for office use, networking features (Windows for Workgroups). Windows also started being bundled with computers, ultimately in combined MS-DOS 6.22 / Windows for Workgroups 3.11 packages.
At this point non-NT Windows still hosted 16-bit applications, which was increasingly limiting (and bad for publicity, for the relatively small number people who understood such matters, since pundits had definitely moved to a 32-bit future). It’s hard to imagine nowadays, but at the time, many in the computer press wondered what the future would bring: Windows 3 was stuck with its “16-bit GUI on top of DOS” roots, NT wasn’t really there yet, OS/2 was mis-marketed, and Unix could never be the next new thing because it was already old at the time (and more importantly, heavily fragmented). Even the hardware side was up in the air, up until Intel released the Pentium: many thought the next big thing would come from the RISC world.
In Windows, there were a couple of stop-gap measures (Win32s and WinG), but the full switch to 32-bit applications really started with the release of Windows 95. The latter was an enormous event, not only in the computing world, but in the (Western) world at large: Microsoft had been promoting “Chicago” (the code name for what was supposed to be Windows 4.0 initially, then 95) for years, and the release event was coordinated around the world with massive media presence (including, in the UK, a specific edition of The Times), and a whipped-up frenzy which resulted in long queues in front of computer stores...
By this point the home market was a done deal. Windows 95 and 98 were the operating system for home and office PCs, but they were still a support nightmare (and Windows Me even more so). NT was getting better all the time, and Microsoft pushed to unify the two streams in Windows 2000 and then XP — not by changing NT all that much, but by adding good support for games (thanks to DirectX and a re-vamped graphics driver architecture) and for older software (with Sound Blaster emulation for old DOS games in particular). Windows 2000 was a great gaming platform at the time, at least with Windows games (Counter Strike, Unreal Tournament, Age of Empires II...).
There were of course a number of questionable practices in all this, the two major ones being pre-announcements (pre-announcing MS-DOS 5 and 6 helped counter DR DOS, and pre-announcing Windows 4 helped counter OS/2), and bundling agreements with PC manufacturers (who could include DOS and Windows at reduced cost, but only if they paid for the bundle on all the PCs they sold).
Is it true that the Windows OS, at it's core was originally designed to simply be the OS of the terminals of the windows server architecture. Similar as to how Fedora is at it's core intended to be the OS of the REL servers.
Both of these statements are incorrect, but it would take another answer to address them...
6I would add two nitpicky details: First, Microsoft made printer compatibility a priority for Win 3.1, which help seal the deal for third-party vendors thinking of porting their office/productivity packages from DOS. Second, the dream (by some) that the Mac would overtake Windows was decimated with the release of Win 95. Of all the major alternatives to Windows, the Mac was the only one you could truly say targeted consumers, and Apple failed to capitalize on it until Jobs returned. But this is small stuff compared to the broader picture. Great answer. May 5, 2020 at 18:04
2I think an important point in the historical evolution of the NT kernel becoming the standard home desktop kernel was the catastrophically bad Windows ME that drove Microsoft to pushing out a replacement ahead of what had been originally planned. If ME had been a decent upgrade to 98SE, then XP probably would not have happened (at least, not when it did). May 5, 2020 at 18:36
13Maybe some mention of Windows 2000, which introduced many of the 9x features into NT for the first time (like FAT32, WDM, and DirectX 7). XP was the spit-and-polish release, but 2000 is really where the streams merged, and some of those "in the know" were running Win2k at home in 2000-2002.– hobbsMay 5, 2020 at 21:20
6@hobbs I remember preferring 2000 over XP for quite some time, if my memories are correct it was more stable and less of a resource hog. And I considered XP "friendly" GUI to be ugly :] If I recall correctly, it took some service packs to make XP usable for me. It pretty much began scheme of server and consumer Win versions built on same kernel.– PTwrMay 5, 2020 at 23:00
3@Thorbjørn a beta version of Windows 3.1 printed a warning on DR DOS, that might be what you’re thinking of; that led to lots of FUD afterwards, but Windows itself worked fine. There was a “business update” to DR DOS 6.0 which circumvented the check, but it wasn’t needed for the retail version of Windows 3.1. May 6, 2020 at 8:06
The other answers include a lot of sound historical information about how Windows evolved into its dominant role on PC's in both the home and business environment. But I think the most fundamental, simplest, "Occam's razor" answer is that consumers never had to make a choice. It was PC manufacturers that chose Windows as the default OS, not users and consumers.
Aside from the one-time media event that accompanied the release of Windows 95, home PC buyers have been accustomed to simply buying PC's with the assumption their purchase would include software. After all, a PC without software (the OS, in this discussion) is not serviceable for any task. In the mind of a consumer, the PC and its OS are inseparable parts of the product. It would be like buying the chassis of your car separately from an engine - not likely - and few buyers even bother to "look under the hood".
Based on this analogy, it is easy to understand that the behind-the-scenes dealings between Microsoft and PC manufacturers dictated this outcome much more than any technical merits or detractors associated with Windows - and its long history provides many examples of both. I'm not claiming that Windows either succeeded because of its merits or despite them. Simply that such concerns pale in comparison to the fact that almost all PC purchases included Windows, as a default. Don't bother consumers with deciding what software is included. As long as it is sufficiently serviceable and compatible with the applications, the buyers simply don't care.
Much has been written on Microsoft's business approach to bundling their software with consumer hardware. It really began with Microsoft Basic, continued with MS-DOS, and reached its peak with Windows.
14it would be remiss to mention this facet of the spread of MS Windows and operating systems without mentioning the monopolistic behaviors of Microsoft (for which is was sanctioned) that curtailed the potential growth of competitive environments. en.wikipedia.org/wiki/United_States_v._Microsoft_Corp. May 5, 2020 at 13:37
5@WillHartung I think you are right. But even Microsoft's behavior vs. competitors is a small contribution compared to their superior understanding of consumer behavior.– Brian HMay 5, 2020 at 13:48
9This answer is also missing the critical fact that in order to get OEM licensing for Windows, OEMs had to pay for an OEM license for every PC shipped, regardless of whether Windows was included, which slaughtered any competitive market in a landscape when every other component could be customized. May 5, 2020 at 18:14
1@chrylis-onstrike- It's a good point that goes to the heart of some competitors' legal complaints against MS bundling. Those complaints did not resonate with typical PC buyers. We techies may have loved a "Choose your OS" option at purchase, but typical users would just be confused by it...– Brian HMay 5, 2020 at 18:48
2If the "typical users" weren't confused by specific options for RAM and hard drives, or for that matter several different versions of Windows, a "no Windows" option wouldn't have been any problem. For that matter, the typical user is probably substantially less aware of such matters today, and "no OS" options are available from most manufacturers that make configurable PCs. May 5, 2020 at 19:04
How exactly did Windows become the OS of the home PC?
Is it true that the Windows OS, at it's core was originally designed to simply be the OS of the terminals of the windows server architecture.
No. Windows started out as a GUI component of DOS - eventually hiding DOS beneath. Anything like a windows server architecture was only devloped way later. In fact, much later than the time Windows became a stand alone OS (wich is, depending on your PoV OS/2 or Windows NT)
So if this is true how exactly did a OS intended to support a companies server offerings become the defacto standard of home computing?
Because it is not true. Windows became a thing for home users way before any server architecture and mostly thru several factors:
It offered a GUI to existing PC users. At that point it was in fact rivalled by GEM, which was, for some time quite ahead in sales of Windows, at least in Europe that is, or more integrated solutions like DeskMate or late comer GEOS.
PC-Hardware became rather cheap during the late 1980s, early 1990s, eventually undercutting equal powerful high end home computers like Atari ST, Amiga or Acorn. Windows was the default counterpart to their offerings for a GUI
During the mid to late 1990 the PC became a serious game platform, even more rivalling existing home computer platforms. With evolving graphics and sound capabilities, as well as (at the time) almost endless memory, 486 machines draw in game development - not to mention the ability of DOS for CD support, easing the distributing of ever growing games. This again cemented the PC as base hardware and while many of these games were DOS based using various extenders (DOS/4GW made this race due being included with Watcom C), Windows was a simple choice for all other tasks - even more so when game companies offered kind of integration to the desktop.
The most important, single turning point by Microsoft was the introduction of DirectX in 1995/96, with Windows 95. While it wasn't complete new (there was WinG before), it was the dedication to a unified set of APIs to cover all (well, most) hardware out there - this came at the same time when the 32 Bit API was stabilized, making DOS-Extenders obsolete, creating the perfect wave for game developers to switch - which in turn made even hard core DOS users install Windows - and the remaining home computer users switch.
Last, but for sure not least, the DOS/Windows-Tax, that forced system manufacturers to either supply some alternate OS (and still pay to some degree) or simply deliver their machines with DOS/Windows. This was heavily enforced by MS during the late 1990s, thus making next to every brand machine being delivered with Windows installed by default. Users had to active change this - something most never even think of.
So in conclusion, Windows 3.1 became usable in ~1993 as (application) Desktop, Windows 95 made 1996 the turning year, were DirectX was able to pull developers over from DOS. And it proved to be a cornerstone up till today. All with a more or less gentle nudging by lawyers waving their version of a 2x4.
"Anything like a windows server architecture was only devloped way later." Windows NT was contemporary with Windows 3.1, and pre-3.1 Windows weren't exactly popular. And even early Windows were almost standalone - while they used DOS for file management, all the other OS stuff was Windows (unsurprisingly, given how thin of an OS DOS was). Calling Windows a "GUI component for DOS" is doing it a disservice - it was a full-blown OS with optional fallback to DOS drivers and using DOS as a pseudo-bootloader. May 6, 2020 at 11:31
1The big milestone for Windows was 2.1 - which allowed multi-tasking with DOS applications running in Windows. In other words, just like all successful MS products, it pushed application compatibility to ridiculous levels. While their competitors (in other ways superior) had only a couple applications, Windows gave you access to pretty much the whole software market on the PC. And when you developed software, you chose the ultra-compatible platform that runs everything. That's the biggie, and that's why there have been very few threats to Windows over the decades (e.g. Java). May 6, 2020 at 11:34
1#2 is entirely correct. As a un-enlightening detail for those interested, I'd add that those alternatives were actually much more powerful initially. The problem was that the relatively open platform (and larger installed base) of the PC meant that every hardware firm in the world was free to work on improving it, and had the financial incentive to do so, so those alternatives didn't stay more powerful for very long. (Which is was vital to getting to point #3...)– T.E.D.May 6, 2020 at 16:22
1@T.E.D. Not really true. The first PC had a 4.77 MHz CPU and could address 1 MiB. Mainboard RAM was limited to 64 KiB - 256 KiB in the second, 1982 Version. And that wasn't 'theoretical' - in both RAM was by default expandable to 1 MiB (minus 48 KiB for ROM). IBM offered RAM cards from day 1 for up to 640 Ki. Third party (or with some fiddling) could fill the whole address space. Also, the 68k was ready (and used) long before the PC was designed. Last but not least, it wasn't superior in any way, in fact, at the same time the x86 always had an advantage (we already had such a question on RC).– RaffzahnMay 7, 2020 at 13:28
1@rcgldr Oh yeah, NT 4 could run quite a few games, and Windows 2000 was pretty much on par with 95 (especially with VDMS to give full VESA and sound capabilities in the console). Sadly, many games checked for version instead of DirectX, and rejected all NT systems regardless of actual capability. 2000 came with compatibility mode that let those games think they're running on 98 and everything worked swimmingly. May 10, 2020 at 12:56
The history of Windows goes back a long way. Windows 1.0 was released in 1985 and was simply a graphical interface for MS-DOS. This was neither revolutionary nor uniquely Microsoft, but a trend at that time. For example, GEOS appeared in 1986 and was the same thing for the C64. There was also GEM and a couple others. The development times on these projects make it highly likely that they were parallel developments instead of copies of each other. Instead, they are both answers to Apple's Mac OS from 1984 (and of course all of them are based on the Xerox PARC research).
Windows 1.0 was a dud. Nobody cared. It was technologically inferior to pretty much every competitor, and didn't offer any actual advantages for the user.
Windows 2.0 (1987) became famous mostly because Apple took Microsoft to court for a number of license agreement and copyright violations. In the marketplace it was another failure, its technology started to catch up to competitors but the added value for the user was minimal, about on the level of the "multitasking" of the iPad today. Like 1.0 it was a program for MS-DOS, so you had to start your computer into the MS-DOS command line and then load Windows up separately.
Windows 3.0 (1990) was the first that was successful and, while still technologically behind the competition, had caught up enough to become popular with users, especially in the commercial sector. For the private sector, a few early games tried but largely failed (here's a few games from those days, and most of them are for 3.1) Again, Windows 3.0 was an MS-DOS program, not an operating system.
Windows 3.1 (1992) and 3.11 (1994) were the popular ones that really started Windows as a thing. Wikipedia reminded me that they finally added drag & drop in that version. So I'll refrain from mentioning again that it was still playing catch-up with other GUI systems.
And, guess what, it still run on top of MS-DOS. Windows 95 was the first Windows that was not an MS-DOS executable, but a bootable operating system.
Sorry for the long introduction, but it is necessary to clearly show that the success of Windows cannot be viewed independently from the success of MS-DOS. With MS-DOS being the dominant operating system at the time (for reasons discussed in various lawsuits and a near unlimited amount of articles and comments) the Windows GUI was riding on the dominance of the MS-DOS OS for a very long time, and only became an actual operating system when it had already achieved dominance of the market. For the vast majority of people, both privately and commercially, the decision at that time was not between Windows 95 and another OS, but whether to move to 95 or stick with 3.11
This history also answers your other question. No, Windows wasn't a server-terminal system. In fact, Windows NT 3.1 - the first server OS from Microsoft - appeared in 1993 and was basically Microsoft's fork of OS/2. It had a so-so reception in the market, unlike its successor NT 4.0 (1996) which as I remember (I started my IT career around that time) was the first server-side Windows OS that professional IT people took seriously.
By that time, Windows was already dominating the client-side market, and I would guess that it was rather the other way around: With NT, Microsoft leveraged its near-monopoly on the desktop market to gain entry to the server market.
1I used W2.11 and we successfully depolyed it with applications on 6 sites. This was a work thing. 2.11 was too clunky to use at home.– cupMay 6, 2020 at 10:57
1as far as I remember back NT 3.5(1) was the first usable of the NT streak I met ... while NT4 basically adopted the Cairo / 95 frontend on the NT core– eagle275May 6, 2020 at 11:18
Yes, 3.5 was a huge step up from 3.1 - I just couldn't list every iteration or it would've all been twice as long. :-)– TomMay 6, 2020 at 12:33
1You make it look like Microsoft stole NT from IBM - in fact, they were developing OS/2 together. OS/2 3.0 was when the cooperation broke down. But it wasn't Microsoft's first server OS, even ignoring the joint OS/2 development - that would be Xenix, which (originally licensed from Bell/AT&T) was by far the most common UNIX of the 80s. They started from UNIX 7, but quickly expanded it and ported it to new hardware. MS-DOS had a huge advantage in hardware requirements, compared to both UNIXes, OS/2 and (much later) NT. But the OS MS pushed was Xenix, until they switched to NT. May 7, 2020 at 9:17
@Luaan yes, it was initially a joint-venture. I said "fork", not "steal" or "copy", hoping that would be clear enough without going into the fine details.– TomMay 7, 2020 at 13:06
As far as I understand it, the whole personal computing revolution that Microsoft Windows did was not entirely by its own design.
As several earlier posters have noted, it was almost entirely by Microsoft's design. We are talking about the age where MS Executives were comfortable talking about "cutting off [their competitors'] air supply". The marketing and competitive warfare was brutal, but largely invisible to the end user population. It was not much different from the IBM at the time, so MS-DOS on the IBM PC was probably a match made in heaven (added to the fact that it saved the IBM executives from having to deal with a woman, which they would if they had gone with CP/M, the dominant DOS at the time).
Is it true that the Windows OS at its core was originally designed to simply be the OS of the terminals of the Windows server architecture?
The first version of Windows I saw was 2.0, running (if you can call it that) on an IBM AT. The experience was horrible - even by the standards of the time. In the mid 80s, I remember working for a company whose corporate policy was a desktop computer was an MS-DOS machine (well, they were making them...). We actually had a Macintosh in the PC room (this was long before the days when every desk had a PC on it), but it had been officially described on the purchase order as a "Graphical Design Aid". I was officially writing code on ISPF on an IBM 370 Mainframe, but it took over a year for me to obtain a 3270 terminal on my desk. A couple of us engineers were playing around with early Amigas, marvelling at the elegance of the core software.
Similar to how Fedora is at its core intended to be the OS of the Red Hat Enterprise Linux servers?
RedHat don't make servers. They distribute Linux-based OSs and other software. Fedora is a community distribution of Linux (GNU/Linux, if you must) by RedHat. RedHat use it, in part, for road testing their new technology. RHEL is designed as an enterprise operating system, with long term stability and paid-for support. Whilst most of the tech and architecture in RHEL has passed through Fedora, it's not fair to say "Fedora is the core of RHEL". They are kind of forks of the same codebase, rather, with the Fedora fork always supposed to be a bit more bleeding edge. Incidentally, you can get the full source to both RHEL and Fedora (the Centos project, among others, is a community rebuild of RHEL).
So if this is true how exactly did an OS intended to support a company's server offerings become the de facto standard of home computing?
Well, your premise is not true, but the "server" core did replace the old desktop core.
I don't remember any Windows Server Architecture with Windows 3.1 or before. The concept of Windows as a server only comes along with NT, IIRC. By then I was managing a set of HP-UX servers. At the time, if you weren't in the IBM camp (mainframe or midrange), you had a choice of DEC VMS or various proprietary Unix implementations. DEC had emasculated their next generation GUI OS (Project Mica) before it saw the light of day, and MS happily cherry picked many of the best (and disgruntled) Mica engineers to work on what became NT.
The Intel-based desktop machines were hamstrung by their 16-bit software model, and the early versions of Windows had to exist within those limitations. When NT came along, MS' pitch was Windows NT Server and Workstation in the enterprise space, and Windows 3.1 at home/small business.
The dichotomy between the two versions became a thorn in MS' side over the course of many years. The low-end version evolved from 3.1 (successful) to Windows 95 (very successful) to Windows 98 (disastrous), but the pressures on the core, like poor 32-bit support, and the lack of any inherent concept of user identity/permissions, meant, especially in the age of the ubiquitous Internet, massive problems. The ultimate fix was to rehome the graphical shell on top of the NT core, creating a more or less single Windows architecture, in the several variants which we see today.
No-one foresaw the elimination of the original desktop Windows codebase - it just happened because it was the most pragmatic solution to the engineering morass MS were falling into at the time (it had happened to IBM, too, in the 1960's: in that case it led to the 360 mainframe system).
I remember a grey-beard after one IT moaned at the auto update feature of Windows 10 just mention in passing that this quirk for the home user is probably a design feature that has something to do with servers as a terminal of a server would probably never want to skip an update.
Actually, it's got more to do with life on an ever-hostile Internet. Before Windows Update came along, just keeping Windows current with security patches was a massive exercise, and a lot (maybe even the vast majority) of owners did not keep their machines patched. The result was massive problems, both in the enterprise and in the home. And when the easiest target of malware is also the predominant platform, you have the makings of a perfect storm. The way Windows is built means a distribution mechanism for updates like you get with RedHat or Ubuntu is a lot harder (the package management systems in most Linux distros provide pretty good decoupling and dependency management) to implement, which is why I think it took MS so long to get it to work properly. Even then, it was originally something you had to take responsibility for by running it yourself. It took quite a while for the realization to dawn that the only proper solution was to make the process automatic, at least for the critical stuff.
So this greybeard (who worked servers, networks and security in the nineties and the noughties) has no compunction in telling your aforementioned greybeard to pull his head in - the alternative is so much worse.
i may have thought Ms-Servers a bit more ubiquitous then it really is/was May 6, 2020 at 13:12
haha, I think the younger bearded fellow was getting a bit of stick from management for the increased bandwith usage the update system was making a reality. May 6, 2020 at 13:19
@NeilMeyer back then, there were more commercial server options that could also serve Windows clients (eg Novell NetWare). Jan 13, 2022 at 14:45
When IBM went into the Personal computer business their design became a standard for the industry and IBM compatible computers quickly became a majority of personal computers produced. MS-DOS from Microsoft was the operating system that IBM picked and the other manufacturers used it as well. When Microsoft came out with Windows they already had a big advantage. As Software programmers were writing software they concentrated on the operating system that had the largest share of the market which was Windows. When people were trained to use computers the same thing happened many were trained on what was most common. So Windows market share advantage became self reinforcing.
Yes this is an important point. Windows became prominent because MS-DOS was already widespread. Microsoft already had their foot in the door.– KingsleyMay 6, 2020 at 22:19
I feel like all the answers miss the key point that catapulted Windows above the others:
Microsoft made it very easy for developers to use.
This has been a Microsoft strength for a long time. The success of any platform is going to hinge largely on developers coming in and writing programs that use it. Unix suffered (and its descendants continue to suffer, but it was a huge problem in the 80s and 90s) by being fractured — a program for one flavor of Unix or Linux does not necessarily work on another flavor and must be compiled separately. Apple suffered by being too much of a closed system. Meanwhile, Microsoft begins developing concepts like DirectX, and IDEs like Visual Studio, and powerful debugging tools like WinDbg. These really cemented Windows as the developer's platform of choice.
Maybe someone else could elaborate about the earlier days (8088–80386 era in particular) but I believe it all holds true: Microsoft (and Intel) were more interested in creating an open platform, available to developers, while Apple wanted to keep a firm proprietary grip and Unix simply could not settle on a standard, leaving Intel/DOS/Windows architecture as a clear choice for any developer who wanted to have a broad audience with a single executable.
("Single executable" was also more important in the 80s and 90s when you couldn't simply expect people to go download the one they need! Or worse, do their own compiles.)
1This is a good part of it that is underrated, thanks! Especially consider that for Windows the Windows group prevailed upon the Languages group and got the entire toolchain (compilers + SDKs) released to anyone for free!. Free as in no charge at all! Free as in $0 compared to $>250 for MS compilers up to then! Compare to Apple and the Mac and their pricy (but deluxe! hardcover manuals!) SDK! (I don't remember what the Apple SDK subscription cost but I had a friend who was an independent Mac developer and I remember he thought they were expensive. (And he was used to HP's prices!)– davidbakNov 18, 2022 at 17:17
@davidbak when was the Windows SDK free? I remember it costing significantly less than, say, the OS/2 SDKs, but it still cost money (whether acquired separately, or bundled with Microsoft or Borland’s compilers, or later through MSDN). Nov 18, 2022 at 17:29
you know, i think you might be right, @StephenKitt - but the cost must have been quite low. on the order of $100, not $300-500. The MSDN eventually cost real money - but that gave you access to this unending supply of CDs with tools and operating systems and stuff. I have to think about this ...– davidbakNov 18, 2022 at 18:00
1@JamieB on top of the decent tools (well, MS C was a pain to use, but Visual C++ was quite nice even in 1993), there was also Charles Petzold’s great Programming Windows books which helped lots of developers get started. Nov 18, 2022 at 18:27
@davidbak Programmer’s Shop catalogs indicate $500 retail, around $350 in practice. (BC++ was $495 retail.) The OS/2 SDK cost $3,000… Nov 18, 2022 at 18:28
By the nature of network effects, there was always going to be a single dominant OS. Nobody at that time was able to pull off 100% compatibility with a complex GUI and its ABI. (DOS was still simple enough that there were at least three.)
It wasn’t going to run on proprietary hardware based on the Motorola 68K (like the Mac or Amiga), because that was more expensive and incompatible with existing software and third-party hardware. That was even more true of UNIX workstations. It wasn’t DesqView because that had no GUI and the company was too small to make a big splash.
Many people at the time expected it to be IBM’s OS/2, but IBM turned out to be so bad at selling it that its own personal computers shipped with Windows. It also did a poor job of prioritizing the features customers actually wanted. Most egregiously, early versions could not print anything and had no GUI. By the time IBM got OS/2 into a usable state, they had to bundle a copy of Windows with it to make it compatible. By then, Microsoft had won.