Comment by anonymous982347
3 months ago
Developing a new consumer-grade OS is literally not possible. I don't mean it would take a herculean effort like the software ecosystem issue takes to address, I mean actually not possible regardless of how much effort any development team put in. Virtually all hardware on the open market is made for Windows, largely powered by proprietary, closed-source drivers. Linux gets some afterthought from a percentage of vendors, but even for it, hardware support is in an absolutely atrocious state. Hardware vendors will obviously not give the time of day to any uppity new OS. This relegates any attempt to a hobbyist project targeting virtual machines or obsolete hardware. The only way a new player could enter the game is by using Apple-level money to develop their hardware in-house, but any kind of corporation fronting Apple money to do that would certainly not be aiming to produce a user-driven experience.
Drivers are a lot of work. IMHO, do some core stuff, and then build in driver adapters. NDIS wrapper, linuxkpi, etc.
If you want to work hard to make things easy, I bet you could build a hypervisor that does pci passthrough for each device to a guest that runs a different OS driver and rexports the device as a virtio device, and then the main OS guest can just have virtio drivers for everything. It can't be that hard to take documentation for writing Windows drivers and use that to build a minimal guest kernel to run windows drivers in.
That indirection will cost performance and latency, but windows 11 feels like more latency than windows 10 too, so eh. You can also build native drivers for important stuff as needed / over time.
> It can't be that hard to take documentation for writing Windows drivers and use that to build a minimal guest kernel to run windows drivers in.
ReactOS has been working on that for like, what, the last 20 years and still is far from generally usable.