Speaking as an OS engineer who's worked on both kernels a lot, Windows is way better under the hood for low-latency, responsive end user stuff than linux

Then again, WinMo 7 was way faster and more responsive than Android at the time, and look how that turned out..

The PREEMPT_RT patchset (sometimes confusingly referred to as RT-Linux) is slated to improve the kernel quite a bit wrt. "low-latency" and "responsive" workloads.

I'm not sure why this is such a problem in Linux, where latency is really not such a big problem in Darwin/MacOS/iOS.

Architected for throughput instead of latency.

> Speaking as an OS engineer who's worked on both kernels a lot, Windows is way better under the hood for low-latency, responsive end user stuff than linux

Still it's probably a lot cheaper to fix some latency issues on linux that to maintain your whole kernel. The main issue with linux is the GPL for drivers, a long with kernel API instability.

> The main issue with linux is the GPL for drivers

That's the main benefit with Linux. OEM drivers on MS Windows are a shit show, even on Windows 10.

I don't believe you: https://en.wikipedia.org/wiki/SCHED_DEADLINE

One costs a lot to maintain, another doesn't.

Windows is a pretty good kernel, GNU/Windows (aka WSL) effectively feels like Linux, with some drawbacks mainly NTFS being very slow (which Microsoft are aware of and working on - see https://github.com/Microsoft/WSL/issues/873#issuecomment-425...).

Yeah I remember reading this particular comment when it was new.

It mostly feels like “ok guys this is the best we get out of this” to me though. If they do not go with a different approach, the filesystem will always be an issue with WSL.

I planned ditching my mac in favor of WSL. Due to MBP hardware issues and the fact that a git status would take 10x time drove me away.

I wish it was good enough though.

I keep a Debian VM around with Samba exporting my ~/src directory to the Windows host. That way it’s possible to avoid file system issues while still being able to use editors and other tools (Adobe CC apps in my case).

git status maybe 10x slower on WSL than native Linux - I wonder how much slower it is than macOS - in my brief use macOS filesystem never felt anywhere near Linux fast to me.

Never been bothered by it in practice but good question. The newish filesystem (apfs) feels faster in daily use so your experience might be outdated by now.

Anyone tried ext3 or 4 on Windows?

They already have their Linux distribution, it just doesn't run on the desktop.

https://azure.microsoft.com/en-us/services/azure-sphere/

However on what concerns the desktop, I am betting more on GNU/Windows.

Why would it need to be Linux? I love linux, but mostly because it works well and is open source. I'd love for a big tech company to put their weight behind SEL4. That would go a long way for Microsoft: formally proven security, tiny footprint, good performance, etc. It's something that has massive potential for super computers all the way down to IOT. It just needs a proper ecosystem behind it.

Or maybe buy QNX off of Blackberry and make it open source again. That would also be nice.

I think as long as Microsoft holds onto Windows tightly the better bet in the future will always be a Linux or UNIX-like OS. Like many other things, the thing that works more like evolutionary systems tend to prevail. Linux might be the dominant race of the UNIX ancestry tree at the moment, but it may not retain that dominance and the thing that will overtake it will probably come from the same UNIX evolutionary roots, e.g. the BSD side where FreeBSD, Darwin, MacOS and iOS came from or the other Linux side branches like Android and Chrome, or somewhere less unexpected. The only way Windows tree OSes will prevail over this is if they allow more branches to evolve, even open source parts of it. User interfaces tend to do better if owned by an organisation with a concentrated purpose and commercial vision, but the stuff under the bonnet seems to do a lot better from being more open and evolutionary.

Looking at the embedded market OSes, everyone is pretty keen in moving away from Linux into RTOS, mbed, NuttX, Tizen IoT, RIOT, eventually Fuchsia, ....

Who knows which ones will survive, however they all have two things in common, they aren't GPL based and just enough POSIX to keep C and C++ happy. Additionally some of them do have stable ABI for drivers.

Linux's long term victory might be constrained to the server room, and even there it is debatable, given the increase in managed runtimes for micro-services, which could even be running bare metal for what I care.

Perhaps Google Fuchsia[0]?

I hope not.

[0]: https://en.wikipedia.org/wiki/Google_Fuchsia

Because of the market share and ecosystem.

As you said, it works well and it's open source. But that's not enough, as they could then chose BSD. But Linux is what's used by most, not BSD.

Given Microsoft new strategy to just adopt what everyone else is using, they will eventually use the Linux kernel.

Microsoft already has a lot invested in Linux support. They're not going to switch to something with a fraction of the community and ecosystem even if it has a lot of good qualities.

It would be make me so happy to see WINE supported as the first-class way of running Windows software.

I would fully expect them to buy Codeweavers should they do that.

Win32 binaries running on Wine is one of the best option for desktop software.

It's funny how it is like the polar opposite of UWP, but it just work and I guess that matters more than technical beauty.

I would be surprised if that was the approach Microsoft took, but I wouldn't be surprised to see a supported way of running Windows Apps emerge.

This could be used as a way to maintain backward compatibility in Windows as it evolves in the future.

Valve (Steam) are doing this.

At the rate things are going, may be Microsoft will even open up their Windows kernel, leaving many of the Windows user land and UI proprietary.

I just hope one day they add regular expressions to Excel.

WSL came out of Project Astoria.