I understand what you mean but one of the powers of Microsoft is, in fact, their extreme backwards compatibility and it would be very damaging if they chose to throw that out of the window.

They could do what Apple did and provide a Windows compatibility layer which gets progressively more buried as old software gets replaced until it's just kept around for the weird stuff that never gets properly updated because it's maintained by a rotating set of undergrads.

They tried that with the App Store stuff. That was supposed to be the new world. But it wasn’t compelling, the store sucked and still does today and it was and still is like wading in excrement working on that side of things.

Then it seems like a problem with execution rather than a fundamental problem with the concept.

Especially with Apple as a largely successful counterexample.

Oh yes exactly that. It's a terrible execution of a good idea.

That’s why I qualified with a comment on an LTS version of Windows.

Keep something around for the next decade (or two) that keeps old hardware running. Allow virtualisation for everyone else.

It does seem that Apple’s approach of aggressively throwing out support for features while designing their own silicon with none of the legacy support that Intel keep around for x86 has paid off. Rosetta 2 ‘just works’.

For enterprise, some version of Windows will probably always be needed. WSL has shown that MS can respond to the developer community and shows that they can do it well. For home users and gamers a fast and lean OS will help shift hardware. The former who will be tempted by Macs and the latter who don’t want bloat will want as little OS overhead as possible.

They could build a completely new OS and keep the compatibility with a compatibility layer like Wine.

Somehow I doubt the New Microsoft will do it right though. They don't strike me as the kind of people who give a shit about compatibility and user experience the way Old Microsoft did.

Apple had no issue with virtualizing Mac OS Classic, and people were in an uproar when this went away.

What do you mean by "a completely new OS"?

Do you think the kernel needs to be re-written from scratch? Yeah, the current one has issues, but that seems like an unnecessarily large undertaking, and one that is also not likely to produce bug-free code. I'd have thought some kind of testing/fuzzing/formal analysis would be a better approach.

Do you think the Win32/WPF/whatever APIs need to be thrown out and some other way of interacting with the core OS services be put in place (e.g. porting the GTKx.y/Glib API?) with some kind of shim for backwards-compatibility?

I mean start from scratch as much as possible and rethink how a computer is used by most people.

By all means keep Windows around for enterprise but build a new OS that is forward looking and eschews backwards compatibility when virtualisation can be used instead.

Aggressively simplify the OS. Make it consistent with a UX that keeps the OS out of the users way. Contain software from vomiting itself across your system. Make things discoverable and consistent, don’t impose restrictions such as locking open files.

I’m coming from this as a user not an OS developer. I use a mix of MacOS, Linux and Windows from day to day. Windows is often a dinosaur and constantly gets in the way. Each new version of Windows is another layer of paint over an existing peeling layer. Certainly Windows can be performant and it would be foolish to throw it all out.

I understand a lot of this is hot air, and your question is asking about specifics. I’m coming from this based on my observations on how students and elderly relatives use their computers.

I’ve moved my parents to iPads. They’re simple and they can do whatever they want to get done. They can browse the web, watch video, FaceTime and see photos of their grandkids etc. At the worst I have to run an OS update for them. They always forget the passwords but TouchID takes care of that.

Meanwhile my father in law has his ongoing war on technology with lists of passwords, frequent manual backups, and lists of things that need to be fixed whenever we visit. The computer is a ticking time bomb to be feared.

My students frequently have no idea what a file system is. Files are scattered everywhere on disk and the cloud. They have no idea what a USB drive does and some think simply just plugging one in magically copies the relevant files somewhere.

Usually by the time I’ve finished with them they’re writing code or running WSL or playing with a raspberry pi or whatever.

So we have this gap. There is the older generation where computers are scary and they’re happy with simplified but robust devices. There is the younger generation where most of them live on their phone and have no clue how computers actually work. When they end up in a job they’ll plod along with excel and whatever enterprise software they need. It really doesn’t matter what version of Windows or Office.

In the middle there are some people (of all ages) who actually have some degree of knowledge.

My solution is to make the desktop OS easier to use while still providing the complexity that is required as one gains experience. Right now Windows is intimidating and outdated.

And what about drivers? It isn't like Apple where there are a handful of supported devices from the last three years the new OS would need to support. Drivers would either be buggy through whatever compatibility mode they create, or they'll drop support for a lot of hardware.

Just to be accurate, support for the newest macOS will go back about six years, not three.

I mentioned an LTS version of Windows.

So I'm a musician, I use a lot of pro-audio hardware and it's a miracle that some gear requiring 15 year old drivers are still willing to run on W10. I'd like to keep it that way if possible.

Now I understand the idea behind technical debt, however backward compatibility is one of Window's biggest strength, so if a new OS has to be designed, it should be considered carefully.

Obviously open source drivers when needed or strict class compliance should be what manufacturers strive for today.

"Windows Classic Subsystem for Windows"

Microsoft already has "Windows on Windows" (aka WoW): WoW16 provides 16-bit app support on Win32 and WoW64 provides 32-bit app support on Win64 OS.

I think they did that twice and both were complete flops. The first one, I think, was the Windows RT. RT had none of the cruft, but many people complained that you couldn't install random windows apps and it just couldn't take off. I think Windows 10 S might be the other but my memory is failing me.

I think in both cases the legacy was still there, just hidden from users and developers.