WSL2 magically solving its severe performance penalty due to the 9P protocol

Never unless you can bring the Linux kernel to the same address space as the Windows kernel, either as a userland component and invoking file syscalls (technically impossible due to ABI restrictions) or as a Windows driver (possible through lxmonika and some source changes to Linux).

Never ...

lol been subscribed and following the thread for this issue since like forever, thinking that the "open" status means still a chance and issues that are out of scope (e.g. asking for the impossible) would just get closed... but alas this was only false hope 🥲

issues that are out of scope (e.g. asking for the impossible) would just get closed

No they won't, they'll be left open until the Microsoft bot automatically closes after a year of inactivity.

If activity remains, it will become basically a discussion.

If activity remains, it will become basically a discussion.

TIL

i'm just now imagining how funny it would be to still see this being open another 6 years from now with an active discussion still going on despite not really going anywhere, due to people necro posting asking stuff like why it takes hours for their git clone ... or npm install commands to finish

This so much!!! I agree... It feels like support for 1 is lacking (despite not seeming to be the intention?), which would be a waste and could be alleviated with open sourcing. Never have I seen a project status to be so ambiguous and up in the air, with future support looking grim / being anyone's guess... yet also unambiguous at the same time with official docs from MS themselves clearly and explicitly already stating that 1 is not abandoned nor that 2 was ever supposed to deprecate 1, despite the naming which implies that 2 must be superceding the "previous" version of WSL 1.

It's like the status for WSL1 is weirdly stuck in some kind of paradox, like schrodinger's updates or something, where WSL1 is both still being officially supported and not really being supported at the same time... like depending on whether it is being observed or not. Feels like we WSL1 users out here are in some kind of weird minority with a very niche or odd specific use-case, with the current status of WSL1, when observed, could officially be answered with "We currently have no plans to deprecate WSL 1" for the few who do use it and question the future of WSL1, while just ignoring its existence (when not observed). With WSL2 becoming more widespread it has also just become synonymous with just WSL by itself, potentially obscuring the fact that for some, 1 may actually just be what they are looking for, espcially for those only looking for something like a "linux 4 windows, but without a vm". Not everyone really has the actual need for complete 100% syscall compatibility alongside the full Linux kernel, and/or looking to run a full on KDE Plasma Desktop environment or some native Linux GUI apps w/ gpu-acceleration running seamlessly alongside Windows applications on their Windows desktop (although yes, this does look very neat and a huge accomplishment of itself for those that need it).

People just see 2 > 1 and be recommending 2 by default and calling it a day or something... hence this chicken & egg kinda situation with WSL2 becoming the more popular option and receiving the majority of attention & support and this 2nd most commented of all time issue regarding WSL2, still open from 6 years ago, with the current solution still being to just use the linux filesystem, similarly to how one would work within a virtual container, or "downgrade" to 1 for those that do not need native Linux GUI app support nor wish to deal with the limitation of not being able to work on Windows (i.e. on the Windows filesystem itself) and the extra cognitive (small yet still dealbreaking) overhead of managing the equivalent of a seperate, but not really, somewhat kind of isolated container + having to swap filesystems for a working dev environment and/or usage of any linux tooling... hence people slowly realizing they are essentially just using a better integrated virtual machine that still costs CPU usage and consumes your RAM.

As a developer and technical user often just working on the command-line in general, one can only dream for "first-class citizen" support (well... as close as possible) while working on Windows with the Windows filesystem, using a UNIX shell with support for UNIX scripts / tooling, alongside Windows executables, with as little overhead as possible; not a whole seperate working environment and/or container for the "developer/linux-y stuff". I wanna be able to use zsh as my primary shell along with the ability to use any common unix utils like sed, grep, find, rm, or rsync, on any of my Windows files (yes I know some of these have .exe ports).

On the other hand I would completely switch to an Arch Linux distro as my primary OS and full-time daily driver... if it weren't for the fact that I am also a gamer. Not looking otherwise to have my expensive a$$ gaming PC, which costs more than my car, being demoted from "first-class citizen" in regards to gaming support and run some compatibility layer either. WINE could be amazing and provide near-native compatibility / performance as close as it wants to... but for gaming to me, emulation is still emulation, no matter how neglible or perceivable the overhead may be. Plus some games will simply just straight up refuse to run on anything other than Windows due to reasons such as implementing a proprietary kernel-level anti-cheat. (cough Riot Games cough cough Valorant)

Feels like I always gotta sacrifice something 🥲🥲🥲 (in this case the developer experience for the gaming experience). But I guess on the bright side this has taught me the valuable importance of inclusivity as a developer. Otherwise I would have just called it a day by only using linux while assuming and targeting a unix environment everywhere I go, and not have to deal with stuff like a common escape character also being the path seperator nor maintain a bunch of miscellaneous windows .cmd batch scripts I use, writing nonsense like for /f "usebackq tokens=* delims=" %x in ('command-with-output.exe') do set "stdout=%%x" or setlocal EnableDelayedExpansion and goto statements (as opposed to having just basic support for simple command substitution and/or functions).

sorry for long post/rant, just realized after typing... heheheh :P

I don't have much to add to the discussion, other than to emphatically thank you for the rant. You have captured exactly my frustrations and sentiments towards WSL and the entire Linux/Windows situation. I actually only very recently found out the differences between WSL1 and WSL2 in depth, and that WSL1 sounds like a much better option for my use cases, because WSL1 gets buried under 'WSL2 is just better!' talk everywhere you look. If only there was a perfect solution... I'm certain there are more like us with the same use case!

it's a opportunity to the bold to reimplement it

Stupid, I'd rather say.

Might serve some purpose to re-implement \Device\lxss, since this repo includes the client code that interacts with it (and can become a good test suite). With the same kernel interface, we can reuse the CLI to manage our new subsystem.

But not the full lxcore.sys. Microsoft failed to keep LxCore up-to-date with Linux, so I doubt random open-source devs can.

The best approach is to port a full Linux kernel on top of lxmonika.

@trungnt2910 linux was built by random open source devs.

@JakeSays Not really random, but rather elite open-source devs.

While the role of the wider community within Linux is undeniable, it requires a legend like Linus Torvalds to start the project, and hundreds of devs financially supported by large companies, including Microsoft itself, to maintain.

Have you progressed on it?