Microsoft Continued Advancing WSL2, Mesa & Its In-House Linux Distro In 2023

[phoronix]

Phoronix: Microsoft Continued Advancing WSL2, Mesa & Its In-House Linux Distro In 2023

While years ago it was hard envisioning Microsoft contributing significantly to the Linux kernel or Mesa 3D graphics driver stack, maintaining its own in-house Linux distribution, or publishing so much open-source software, these days it's par for the course thanks to Windows Subsystem for Linux (WSL), Linux dominating Azure instances, etc. Over the course of 2023, Microsoft continued investing in various Linux/open-source efforts that benefit their cause...

https://www.phoronix.com/news/Microsoft-Linux-2023

[AlanTuring69]

It is rare to be able to run Linux in the corporate world, and WSL is truly amazing to be able to bridge that gap when macOS isn't an option. Long gone are the days of cygwin...

[anarki2]

They need to fix IPv6.

And firewalld being broken in Docker containers for whatever random reasons.

[paulocoghi]

I really want to believe that, this time, Microsoft won't do the EEE, aka "Embrace, extend, and extinguish".

But half of me doesn't believe they won't.

[highball]

I really want to believe that, this time, Microsoft won't do the EEE, aka "Embrace, extend, and extinguish".

But half of me doesn't believe they won't.

I think MSFT is just moving to have their own Windows userland on top of the Linux kernel. CBL-Mariner has been moving in that direction. If they do, it will be much cheaper for them to spin off cloud Windows environments for their Office 365 service. Others think it's going to be MSFT's way of shutting down Linux adoption by developers in the corporate sphere. Why let your developers have Linux installed when WSL is available and works good enough by IT standards. Personally my money is on MSFT ditching their own kernel. They'll save money in the cloud and they will save money on engineers and development costs. Win-Win for MSFT.

[ehansin]

I absolutely am rooting for Linux in the desktop space and of course will keep using it where I can. But at work I am on Windows for my desktop PC (and macOS via an older MacBook Air laptop.) I am thinking of running WSL2 on my Windows 11 desktop PC (I bit the bullet and jumped from 10, and customized to how I like things.)

I have the means to dual-boot at work, but that doesn't work well if I want to be in both environments at the same time. I already have a pretty clean install of Windows 11 going. I'd just need to enable Hyper-V stuff and install WSL2 (if I recall correctly.) Longer-term I might toss in a second drive, and from there figure out a way to run Windows (already on the primary drive) via QEMU. But that is more down the line.

I know there are people here that do not like the whole WSL thing and totally get it. But if it would allow me a Linux environment at work without much hassle, could be a good thing. I am not sure if there are issue enabling Hyper-V (in which case, again if I understand correctly, is a Type-1 hypervisor and therefore Windows runs on top of this as well.) If that doesn't matter, might be easy. Enabling Hyper-V use to conflict with VirtualBox, but I believe no longer an issue.

I understand the issues people have with this. But from a practical point of view, it gets me a Linux environment at work. I would never run a server like this, nor a more dedicated Linux based desktop. But in this case... I assume it is good enough, but I suppose I'd need to jump in a figure for myself.

[kpedersen]

CBL-Mariner is interesting in that they are attempting a "common base". I don't really know of any Linux distro that has been disciplined enough to have a proper base as seen in most of the BSDs.

Because Linux distros are really just a collection of 3rd party software, it perhaps isn't possible to have so much control to keep a tight base maintained. This is probably why "Linux Standard Base" (LSB) kind of failed. No-one could agree. No-one wanted to adhere.

The closest we have is probably "EL" which is really just "do what RHEL does" and call it a standard.

[fafreeman]

I think MSFT is just moving to have their own Windows userland on top of the Linux kernel. CBL-Mariner has been moving in that direction. If they do, it will be much cheaper for them to spin off cloud Windows environments for their Office 365 service. Others think it's going to be MSFT's way of shutting down Linux adoption by developers in the corporate sphere. Why let your developers have Linux installed when WSL is available and works good enough by IT standards. Personally my money is on MSFT ditching their own kernel. They'll save money in the cloud and they will save money on engineers and development costs. Win-Win for MSFT.

all the free labor open source provides i don't understand why its taken them this long to start getting their feet wet with open source. if some guy in canada wants to spend time sending in patches to improve their networking stack, why deny that free labor? let alone another billion dollar company wanting to send in patches to improve something for them, for free. i think they have finally realized this. the new edge based off chromium has been a big success in this department. all that stuff done by google + others for the rendering engine and all that, with microsoft just caring about the stuff of a web browser they want to care about, mostly the UI. i can see them doing this with the kernel and future file systems. rather than trying to maintain all that complexity in house, just adopt the linux kernel + one of the many linux, open source file systems to replace aging ntfs, and focus most of their efforts on things that help sell more windows versions. UI updates.

i know some people will go on saying "well they need backwards compatibility, they won't adopt the linux kernel." microsoft doesn't care about backwards compatibility. they care about people running windows. backwards compatibility keeps people running on windows. as long as people can run their software, they don't care what version it is, or what operating system it is. when you look at wine, proton, and the like, and seeing the massive success over the past ten years in running windows software on linux, knowing its nearly all reverse engineering efforts, not only is the cat out of the bag with "we need windows solely for compatibility in general" but there is absolute no reason why microsoft can't do their own "wine" to make windows stuff run on a microsoft branded linux kernel with microsoft patches for windows nt compatibility. microsoft doesn't need reverse engineering. they have full access to their api's and loads of internal documents / contacts to original microsoft engineers. i see them doing wsl as a way to get their feet wet for something like this later down the line. i really can't see them continuing their own in house kernel anymore. it has to be a pretty sizable net drain on them these days. i wouldn't be surprised at all if cbl-mariner is cheaper for them to develop and maintain than the nt kernel and windows in general. so much of the work for cbl-mariner is done for them by others.

[oleid]

I know there are people here that do not like the whole WSL thing and totally get it. But if it would allow me a Linux environment at work without much hassle, could be a good thing.

I used it for a few months in my new job, but at some point its graphical output stopped working correctly. It drew something, but it wasn't readable anymore. On top of that, my IDE Clion randomly misbehaved when using its WSL based "remote development".

So I ended up installing Debian and never looked back.

[rabcor]

This is not microsoft contributing to the linux kernel, this is microsoft extending windows functionality through the linux kernel.

You can't really call it a contribution if it doesn't benefit anyone who isn't a windows user. Honestly I'm not even sure we should accept these so called contributions in the first place, they add complexity to the kernel in a way that's meaningless to most of us and only increases potential for bugs and security holes. This shit shouldn't be mainline.

There should just be a patchset for WSL and people who want to use it should have a WSL compatible kernel, and people who don't want it shouldn't have to have it in theirs.

EDIT: Seems my point's already being proved... https://www.phoronix.com/news/Linux-...-RNDIS-Drivers
This is probably just the start of a trend, I say we remove all of it from mainline and be done with it, it helps literally nobody who isn't on windows, so it shouldn't be in the mainline linux kernel it doesn't make sense.