Not much in there require the DOS version to be precisely MS-DOS; on the contrary it was done mainly for compat with 3rd party code which e.g. "knew" where the DOS structures were, and surprise 3rd party DOS obviously had to be compatible on those points.

Now remains the "suck the brains out" problem, and although it could have been easily solved cooperatively IMO, it is illustrative of why people would not have needed 3rd party DOS when Windows 9x was running anyway: the kernel did nothing, the libs were mostly irrelevant (and MS-DOS probably gave that small compat edge even there, simply because its prevalence even before 9x), and the cmd line tools, well a) who cares about the tools b) actually if you do I suppose that was the easiest things to run (and if not: the irony; oh it is scandalous that windows relies on internals of MSDOS -- but not that e.g. Caldera DOS tools relies on internals of their DOS kernel)?

Caldera would have needed a Windows clone to compete. DOS was just becoming irrelevant.

Once you postulate a reality where Windows 95 was an unbundled product, a few things change:

* It becomes much easier to make legal noise about Microsoft pricing. If they're saying "Buy Windows 4.0 and we'll give you MS-DOS 7.0 free, or for some token sum" it's pretty obviously market dumping. If you had to pay, hypothetically $50 for MS-DOS 7 and $50 for "Windows 4", instead of a single $100 package, Caldera can compete with a $30 DR-DOS product even without having a Windows clone.

* There could still be a territory war over the quality of the underlying DOS. This was still a time when you'd have a lot of games, in particular, that required a drop down to DOS mode.

* There was also the opportunity to sell features that still had DOS components involved more modularly. I'm thinking about FAT32 in particular, because I can recall radio advertisements boasting about it to try to sell Win95 users on upgrading to 98 or 98SE. What if you could pay $10 and get a DR-DOS competitive upgrade that supported it instead of a full $100 upgrade with a bunch of features you didn't care about?

In such an environment, there could have still been 5 or so years of DR-DOS as a viable commercial product until the much less unbundlable NT-based Windows went mainstream. That's a significant amount of runway. Their pivot from there is not obvious-- I can imagine some attempt at a Linux distribution with a tightly integrated WINE setup, but that might have been too much to ask for at the time.

I also wonder how a Microsoft that had been clearly slapped and told "you must play well with others" in 1995 or so would have behaved over the next 20 years. The trajectory of IE would have likely been very different, and I wonder if they would have handled some of the other Office competitors with kid gloves to avoid legal scrutiny. Maybe even some healing of the rift with IBM on OS/2, so they could point and say "look, we're interoperable, here's another compatible product."

The trouble with FAT32 being a feature to sell DOS is that Windows 9x doesn't use DOS to access the filesystem in normal circumstances. AFAIK it had a compatability path that could use DOS, but then you'd be slowing stuff down.

I can only imagine bad things happening if Windows and DOS disagree on filesystem aspects.

That's why the Windows 95 CD-ROM version shipped with the "LFNB" long filename backup utility https://jeffpar.github.io/kbarchive/kb/135/Q135315/

> Caldera would have needed a Windows clone to compete. DOS was just becoming irrelevant.

Well, Digital Research had GEM[0] since mid-80's. Development was probably halted after Novell or even Caldera acquisitions.

[0]: https://en.wikipedia.org/wiki/GEM_(desktop_environment)

This does not seem like a reliable source about Microsoft's actions in the 90s. What are they gonna do? Blog about how they're guilty of anticompetitive behavior?

I don't remember Raymond Chen ever lying. If there weren't legitimate reasons, Raymond would just not talk about them. Especially since this would be a very elaborate, multi-faceted lie that anybody with an intermediate knowledge of the inner workings of Windows 9x would've been able to call out.

Raymond Chen may not intentionally lie, but he can be wrong and has been wrong in the past. One obvious example that I remember is when he said that the reason Pinball.exe disappeared from Windows Vista was because they couldn't port it to 64 bits. And yet as I read every single clickbait newssite parrot what he was saying, I was literally playing 64-bit Pinball.exe on a standard Windows XP x64 edition ...

The comments in his blog disappear every time Microsoft shuffles it around, so these discussions are lost forever.

Not lost, for now: <http://bytepointer.com/resources/old_new_thing/index.htm>. (Thankfully, MSDN blog comments didn’t require JavaScript, or they would now be lost...) Here’s the Pinball piece, for example: <http://bytepointer.com/resources/old_new_thing/20121218_294_...>.

It was probably the 32 bit version of pinball, running on 64 bit windows via WoW64

No. Easily distinguishable by the * (or lack of it) in task manager.

I guess I'm expecting lies of omission here. My guess is that content on Microsoft's blogs, especially about this period is cleared with legal before anything goes up.