I've done hardly any game development in my life (making a simple Gamemaker game at high-school in 2016 or 17, & making a box fall in Unity a couple years back; so you can call me a complete noob. But I was just wondering: If I for whatever reason wanted to make my game work natively on a Bunch of different Windows versions, like 95, 98, 98SE, 2000, ME, XP, Vista, 7, 8, 8.1, 10 & 11; would that be possible without making separate versions of the game for different Windows versions? It sounds like a cool project for doing just for the fun of it, for learning about the different OS versions once I already have more experience with development on modern Windows. What if I made the game on Godot game engine? Can Godot games even run on such old operating systems? I heard that Windows 2000 and above are NT based, and major Windows versions prior to that ran on something else: would this greatly affect the development process at all?
Clarification:
Sorry, but I should have clarified that my development platform is Linux, and would be porting to Windows, which obviously should change the answer to my question drastically; I have no idea why I worded things to sound like I would develop the game on Windows first and foremost; but that was my mistake.
One of the big problems with getting hold of dev tools from that era is there wasn't a lot of stuff that was free. Pretty much all the windows developer tools had a price of entry. This means there's less chance of being able to find stuff because someone making it available for download would probably be piracy.
They'd probably want to develop with Visual C++ 6 on Windows 98SE for the best chance of success. I think I agree that DX8 is probably the one to target, though there might be better luck targeting an OpenGL (or maybe SDL) version from that era.
Getting decent docs might be hard too as even the MSDN library was a paid for product.
As for networking, I think I've seen some projects that port TLS 1.2 to 9x windows, so I wonder if you could bundle a library to enable this. I guess the OS TCP/IP stack is riddled with holes too though probably making this a bad idea regardless.
It would definitely be an interesting project, but I don't doubt it's going a pretty frustrating time throughout
I think using OpenGL will be even harder than DX8. Your point with the documentation is spot on: it will be hard to come by. As for TLS/SSL third party libs: I had the fun opportunity to implement/maintain the backwards compatibility for a piece of software for Win XP/Server 2003 ~3 years ago. We had to jump some hoops with a custom compilation of OpenSSL to get the parts of TLS 1.2 we needed in Windows 10. I can say with confidence, that that route is a dead end for even older Windows versions.