From a purely technical point of view, what would you have wanted next? Windows XP running "on" DOS? And the irony is that it was not even that much different o

From a purely technical point of view, what would you have wanted next? Windows XP running "on" DOS? And the irony is that it was not even that much different on that point: Windows 9x did not really run "on" DOS either. It had some form of "cooperation" (for legacy drivers only) / backward compat. And it was booted form there. And you could still run DOS bare-metal. For most users, the DOS parts were perceived as deprecated anyway; aligned with the fact they officially were... And well even in the Windows 3 era DOS was merely used as a bootloader on modern machines. Of course Windows 9x took the concept even further. The contextual difference was simply that there was tons of DOS programs and games even when 3.x was out (esp. games) and that even if that continued a little bit for games, it quickly migrated as Windows 9x became a comprehensive platform with the needed APIs (plus more and more people having it). And the default boot UX that was kept as-is by most people. So from a market perspective, 3rd party DOS + Windows 9x would probably have died well for 98 anyway, and probably become ultra-niche way before (well, was probably niche already just before the 95 launch...) Not to say MS did not have anti-competitive behaviors during that era (or attempts e.g. the AARD incident), and of course even just strategies to compete. They are actually well-documented thanks to various trials. But there is also the usual taking of random tech dependencies for granted until the main vendor suddenly changes its stack: and OMG you can't run booted from another DOS. Hell even when parts of the tech stacks were actually documented, licensed and supported, sometimes it stopped to be the case on the next iteration: and then you can't use a Cyrix processor on a MB with an Intel chipset anymore. It is not necessarily an abuse of dominant position all the times. And here the point of the original post was to say there could have been a Win9x capable DR-DOS, so I don't even see why you are talking about "the next major version of Windows wasn't able to be "integrated" with any non-Microsoft DOS" when it clearly could: simply the compatible 3rd party DOS capable of that was not commercialized. Should MS have been forced of cooperating with other companies to have parts of their OS interchangeable? I'm not sure. Apple surely did not have to e.g. thoroughly document and then advertise their PowerPC nanokernel as reimplementable by 3rd parties (not an exact equivalent, but it is impossible to come with exactly the same scope given the two systems are so different, so I'm just taking a random component that can be distinguished from well-known and previously existing one) -- but then if some found an interest in doing that without support, they also could have done so.