Hacker News Reader

Comment by bserge | Hacker News Reader

I don't really understand the use case for WSL. What are the advantages? You could run a full Linux VM and do everything on it on any machine these days. Which is what I do, but I'm just an old school hack. Anyone who uses WSL care to enlighten me? Thanks!

View on HN

When I launch my WSL2 Ubuntu shell, I have a CLI prompt in < 1 second. Clipboard works perfectly since it is a window similar to PowerShell, my windows file system is mounted in under /mnt, and generally everything is lighting fast.

Getting that same quality of experience under VirtualBox was nightmarish. Longer boot times, shared clipboard/filesystems were always breaking, random stuff related to hardening made update difficult.

>Clipboard works perfectly

Which one? Secondary? And the primary does not work at all?

The Windows clipboard just works, because WSL2 terminals are just regular Windows apps.

At this point - Can't you just SSH into the local linux VM or a remote linux server?

You can, but to get the same edit in the windows UI experience you would need sshfs, which is another nightmare thing to get running on Windows.

Primary and secondary selections are specific to X11 which is not involved at all when running WSL command line apps in a Windows terminal

Other than the startup time (mitigated by starting the VM right after startup) I have no issues with virtualbox. File mounts, copy & paste... everything works more or less fine.

WSL, on the other hand, has been nothing but trouble.

One downside there is the constantly huge RAM consumption of such a setup.

EDIT: At least compared to WSL1. However WSL2 also has some advantages in that it can reclaim unused memory and suspend the VM when not in use.

Why would VM-based WSL2 use any less RAM?

WSL VMs are gen2 (fully paravirtual) hyper-v VMs, so use dynamic memory allocation. VirtualBox, even with Hyper-V as its engine, doesn't do this. I don't think any standalone Linux distros offer a standalone Hyper-V paravirtual image that is not a WSL2 image.

The Ubuntu image you can select in the "New VM" dialog in the Hyper-V management tool (or whatever it's called) absolutely does work with dynamic memory allocation and I seem to recall getting it to work on an Arch VM as well.

I believe the "Linux Integration Services" for Hyper-V are actually mainlined at this point so I would expect most things to work. Setting up RDP for enhanced desktop sessions is the only painful thing I remember.

You don't need paravirtualization to do dynamic memory allocation -- all VMs do this.

No. Gen1 images use a fixed allocation pool instead of an oversubscription model. You need a gen2 image with a cooperative memory allocator on the guest in order for this to work properly.

Well, so barring ancient HyperV versions, all VMs do it. This is more of a guest thing than hypervisor thing, anyway. All the VM needs to support is balloning and this is completely unrelated to paravirtualization.

I use a mix of WSL2, and VMs. WSL2 has a few nice properties for some use cases:

a) It comes default with some network binding magic (ok, just configuration) that makes your WSL env almost actually behave like localhost. I know you can configure this yourself, but it's nice out of the box. (on the other hand, sometimes is this exactly the opposite of what you want...)

b) My main machine is a bit RAM constrained, and WSL2 will "magically" reclaim RAM when I close the consoles (unless I left something running...). Once again, not the biggest win, especially if you spec'd out your machine to run VMs, but still nice.

I think WSL2's biggest wins are not for 'full out development'. But it's nice when you need want to jump in and poke at something quickly in Linux land. The startup time for the console/VM is fast enough (~1-2 seconds), and the cleanup is good enough I can quickly jump between, without having a VM actually hanging around all the time.

I'm sure the file system integration is useful for some usecases... classic one would be if someone emails you data files or something. You download in your browser in Windows and then can access pretty seamlessly from WSL.

I think I would summarize WSL's advantages as being a bunch of little quality of life advantages that are very much suited for Windows dominant workflows that periodically jump into Linux.

There are all sorts of other disadvantages though - I would not recommend anyone who has a comfortable VM based workflow to ditch for WSL. An example of an uncomfortably stupid rough edge would be periodically time-desyncing: https://github.com/microsoft/WSL/issues/4149 . I shit you not, my workaround is to manually adjust my linux time forward and back with `sudo date --set "${INCREMENT} seconds" until it gets close enough that OAuth doesn't shit itself.

Have you tried WSL1? It sounds like it might fit your use cases better:

- WSL1 shares the same network interfaces and addresses as the host, no proxying like WSL2

- Significantly less RAM usage

- No time sync bug

Oh, I use both =P

I use WSL2 for some docker based projects.

Also, I think WSL2 is currently only way to access GPU through Cuda API from the guest linux.

Hyper-V can also do PCI pass-through if you have a compute card to assign to the guest. If you're just running the one card, though, you need WSL2 to handle the sharing.

Additionally Win Pro is needed for Hyper-V, while WSL2 can be used in lower level Win versions.

I used to do all work on Fedora, dual booting to Windows for the Windows apps. WSL was OK when I tried long back, so hadn't bothered. But for the last few months I was forced to try WSL (now WSL2) again due to some Windows work, and have been using the Ubuntu app. I've got to say I'm now a believer and don't scoff anymore. I can now do ALL my work without problems on WSL2: mainly Go/C++/Dart, containers (podman/buildah), Flutter, GRPC, etc. The added beauty is that you can go seamlessly between Windows and Linux apps: e.g., you can just run an .exe from the BASH command line in WSL2 just as you run an Linux command. It's insane how they did it, and you have to try it to experience it.

There are some issues still: GUI apps are still a pain to set up (but they're fixing this), there are some networking issues (e.g., accessing servers running on Linux), file path issues on some apps (like I guess git here), etc. I never personally experienced any of the git issues cited, and I use it regularly from the Ubuntu command line. Overall, I very highly recommend WSL2 now, Microsoft hit a home run with VSCode with devs and IMO WSL2 is turning out into another.

I have been teaching computer related classes to uni students for some time. Most of them run Windows. Running a virtual machine often means accessing the UEFI to be sure they have virtualization extensions enabled. And many of them have shitty personal laptops, which means running a VM will really slow the "Linux" experience. Sharing files between the host and the guest is also sometimes tricky, and I don't want to force them to install another OS on their personal laptops. WSL is nice in that it allows them to have a bit of Linux world without being very intrusive, and runs better than virtual machines on many of their laptops.

I'm what is called a "vacataire", which means I'm also not in the position of asking the school IT departement to enable stuff on the computers for the classes…

> Running a virtual machine often means accessing the UEFI to be sure they have virtualization extensions enabled.

Likewise for WSL2.

I hope modern computers with a capable processor are all shipped with virtualization enable at firmware level, because Hyper-V can be used for tons of things in recent Windows.

Enabling hardware virtualisation opens up a significant and deep attack surface. Considering the vanishingly small percentage of users which benefit from it, I hope it stays off by default.

Windows is moving to a model where Windows itself is run as a virtualized OS. I believe this is enabled by default in new installs.

So having a Linux VM in Hyper-V isn't opening up much new attack surface.

First I hear of this. Source? Googling for this predictably returned unhelpful results.

In VBS environments, the normal NT kernel runs in a virtualized environment called VTL0, while the secure kernel runs in a more secure and isolated environment called VTL1.

https://www.microsoft.com/security/blog/2020/07/08/introduci...

https://docs.microsoft.com/en-us/windows-hardware/design/dev...

It's not enabled by default. Enabling Hyper-V still causes a battery/performance hit that is going to be hard to get rid of.

Virtualization-based security -- a lighter mode of Hyper-V sans real VMs -- is enabled by default on new installs on recent-enough hardware:

https://techcommunity.microsoft.com/t5/virtualization/virtua...

The link you posted only contains one device and it happens to be an ARM device. Seeing the impact it still has on battery life at least on x86, I really doubt they have enabled it by default. It was not enabled by default on x86 in 2020 at least.

That's just a blog post about the feature being deployed, of course it won't have many examples. Take any PC from the past few years and install Windows 10 x64 on it. It will have VBS enabled and hypervisors that do not support Windows Hypervisor Platform won't work. That's been my experience since at least 2017.

If you click through "capable hardware" to here[1], you'll see the list of requirements for VBS, including:

> Virtualization-based security (VBS) requires the Windows hypervisor, which is only supported on 64-bit IA processors with virtualization extensions, including Intel VT-X and AMD-v.

So it will never be the case on x86/IA32

1: https://docs.microsoft.com/en-us/windows-hardware/design/dev...

Well, I'm asking because my experience is exactly the opposite. I've installed Windows countless times on systems with all the requirements and HyperV is not enabled. The day it starts being enabled, I don't even want to imagine the number of support calls.

Do you have a reference to the increased power consumption that running a hypervisor causes? This thread is the first I've heard about it, and I would like to learn more.

No, and I also would like to find some academical test. Try it -- you can even dual boot hypervisor on/off. It's not a small effect. Around 1h extra one on "almost idle" scenarios.

Erm, explain? What attack can you do with hardware virtualization enabled that you cannot otherwise do?

I've read[1] that it can make Meltdown/Spectre attacks possible and thought I'd seen many more reports and discussions about it, but seems I was mistaken.

[1] https://nvd.nist.gov/vuln/detail/CVE-2018-3646

Are there significant risks from running virtualization locally like this? If so, can you provide any links or elaborate a bit so I can follow up? Most of what I've seen on such vulnerabilities refer to server infrastructure.

Having searched through my browsing history and some web searches it seems you're right. I do wonder if the move to more virtualisation outside of the server world will open up additional vulnerabilities but it does look like that's where most of the trouble is at the moment.

In Windows, enabling virtualization actually reduces attack surface. E.g. it is used to protect against kernel-level malware:

https://www.techrepublic.com/article/how-virtualisation-is-c...

Yes quite correct. I know some anti-virus engines have been using it for behavioural analysis. I wonder if this will mean more exploits against anti-virus emerge, as with their unpacker routines.

Well, android simulators do use hardware virtualization. And there is actually many people use them outside of developments.

unning a virtual machine often means accessing the UEFI to be sure they have virtualization extensions enabled

Note that this benefit applies to the deprecated WSL 1 only, WSL2 actually runs a VM underneath so it requires the same hoops that a VM would require.

To sell Windows laptops to the same crowd that buys Apple laptops to develop GNU/Linux software, instead of supporting Linux OEMs, and are unhappy that Apple only cares about developers on Apple ecosystem.

Microsoft understood that they only care about having some kind of POSIX support, and nowadays being Linux compatible is more relevant than straight POSIX, as the BSDs and IllumniOS also found out with their compatibility layers.

Or, alternatively, a crowd whose needs aren’t completely met by Linux.

Then why they develop GNU/Linux software to start with?

Because their needs aren't met by developing Windows or macOS software?

Then again why they don't support GNU/Linux vendors?

So if someone installs Linux themselves they don't meet this bar you're setting? That seems pretty far from the normal open source ethos.

Ah, so using evil commercial OSes is part of that so called ethos?

No, installing your own OS is very open source.

Which isn't a thing with macOS and Windows, so where is that ethos again?

"Yes, you see, you're only free if you do exactly what we tell you."

Also, it's a thing on Windows/PCs.

We are talking about GNU/Linux here.

We are not. You are.

That was the whole point of this thread, naturally some rather move goal posts to avoid the subject that Linux users rather give money to proprietary desktop platforms than help desktop Linux ever become a reality.

Yes. Because Linux is not competitive in some cases.

You were confused why people would do things with GNU/Linux without "supporting GNU/Linux vendors".

But that's restricting it to people that use the preinstalled OS, which is strange, because of how fitting it is to install your own copy of GNU/Linux.

People might get paid money to develop (proprietary!) software that deploys to Linux. That doesn't mean they enjoy using Linux.

Indeed, I have been doing that for years and am yet to install WSL.

One reason might be, is that they might need to in part work with a server-side system that runs on Linux and it’s easier this way.

But maybe they also need to work with stuff thats Windows only. Say they need to produce media assets with the Adobe suite.

Not everything is vimmable.

SSH and X Windows servers also exist for Windows.

It takes me 2 minutes to install WSL Ubuntu. Or I could spend half a day on figuring out a hacky solution to a Windows/SSH/XWindows workflow that offers even a comparable level of integration. It will grow into weeks of obsessive tweaking until I feel compelled to write a blog post for HN where I show my sick setup and hours and hours I dumped into this, while meaningful work piled up in my TODOs.

... Or I could just install WSL.

Or just install Humming Bird, working just fine for me since 2000.

Or just install WSL?

What's the problem with Microsoft having an answer to this?

In what concerns Microsoft, I praise them for having acknowledged the strategy error that was not giving the UNIX subsystem the same love as Win32.

Because as proven by macOS and now WSL adoption, GNU/Linux would never have taken off if PCs already had a mature set of POSIX toys, given that its users care 0% about GNU/Linux, and would be deploying to HP-UX, Solaris, Aix, Irix, Tru64 just as well.

Cool. I'll use WSL though.

I guess you'll have to use it to "believe" ;)

Setup is much less hassle, startup is much faster (one or two seconds) and resource usage is much lower than running Linux in a VM.

Also little things like VScode (running as Windows UI application) automatically detecting WSL and setting itself up to connect to the WSL side. To do the same for a VM requires at least to configure the SSL ports on the VM.

Being light-weight won it for me. And the hassle free setup to use the same disk on both environments.

Docker for Windows using WSL2 as its engine is also nice. The Docker commands somehow (magically) get added to your Linux distro and you can mix and match how you like.

It's very useful for projects that require (or work better in) a linux environment. I don't want to start up a full linux VM, that sounds like a lot of overhead and I'm more likely to not bother at all. WSL is much lighter than that (we're not talking virtualbox here), if I open windows terminal its up and running, there's almost-zero start up cost.

With VSCode I can switch between working in WSL and windows easily, without needing a desktop GUI to write code in. There's no startup cost, I'm immediately in the correct environment for that project. I've found it very useful since it came out.

You've not explained why you would not run a native Linux system in the first place…

Office, which doesn’t have a good replacement on the web or on Linux.

What is inadequate about Libre Office ?

Calc is really not an adequate substitute for Excel if you need to use VBA macros - which might be coming from your customers! And in the wild I see plenty of older Excel documents (before 97) which I don’t think are compatible with LibreOffice. Obviously this isn’t an ideal situation but historical data around prices/etc from the 90s are often stored as old Excel binaries.

In general I think using Calc (or Google Sheets) instead of Excel is a bad idea for a business. Word processing and spreadsheets - sure (although there might be some things specific to Word/PPT that I am not aware of). But Excel is a very specific piece of software and should not be thought of as a general “spreadsheet tool.”

Calc has several options for languages to use for macros, for instance Python – I'm not certain about VBA support, but it would seem that many industries are slowly migrating to Python anyway :

https://news.ycombinator.com/item?id=25588720

The point is that VBA macros are portable across Excel but not between Excel and Calc. Calc does not support VBA macros. As many businesses need to consume spreadsheets with VBA macros, and since porting VBA to Python isn’t sustainable, Calc is simply not a good enough replacement for enterprise.

I am not sure that one anecdotal blogpost supports the idea that industries are moving from Excel to Python. It is more likely that people are productionizing Excel spreadsheets with Python these days, instead of C++ or Java.

Not sure about Calc, but 'modern' Excel doesn't even seem to be able to deal with text (UTF-8) in CSV properly :

https://news.ycombinator.com/item?id=25015679

(And I know from experience that 2003- versions of Excel don't.)

I am not arguing in favor of Excel! I hate Excel. I don’t have it installed on my personal devices. I never use Excel unless I have to. I am saying that Excel does specific things that are widely used in industry (especially finance and industry) which are not supported by Calc and that Calc is not a good replacement for many businesses.

Because corporate only allows their sanctioned images.

I write C# for .Net Core and .Net Framework professionally.

I'm open to suggestions but it simply seems crazy difficult to do this outside of Windows. WSL lets me jump into *nix quickly when I need it for whatever reason. I don't need a desktop machine. Why not just use WSL?

I just like being able to work in a Windows IDE and compile in a (99%) Linux environment without messing with a VM or mounting shares. Like QMK[1] -- develop in Visual Studio, compile in WSL, flash in QMK Toolbox (Windows). Or in some very specific corner cases, have a Linux-based workflow that suddenly calls into a Win32 utility.

In some cases I need to test something that's Linux-only -- an idea or a GitHub project that just doesn't run on Windows -- and it's easier to jump into a WSL console than boot a VM. But on the other hand, I don't do that frequently enough to just keep a VM running all the time. (Plus there's VM idle RAM usage, and I can't use Hyper-V [which can reduce VM RAM] due to host performance concerns)

[1] https://github.com/qmk/qmk_firmware

To chime in. The company I work for has (after being acquired by Accenture) implemented "Endpoint Management" for all machines (read monitoring and spyware).

EPM prohibits us from creating virtual machines, but WSL2 is possible. So a lot of devs at our shop (at least the 5% or so who are using WIN) use this setup.

Personally - I do this on my private machine - I switched, because with the EPM software I wanted to separate my work machine from my private efforts, as I just do not like Accenture to be able to read all mails, read all files and install arbitrary software on a machine that I have private data on. Before we got acquired my employer allowed private use, we were admins on the machines and there was no spyware installed - so only using one machine for work and private stuff was feasible.

I do not use WSL but see a good reason for WSL. The benefits for the users are superficial but the really big thing is on the business side for Microsoft. With WSL they try to trap more and more Linux developers into their ecosystem. It is all about getting more control of the most talented and fruitful developer minds. In the end Microsoft and its shareholder will benefit if it keeps its hands on this crowd.

Docker. Docker can use WSL2, so it works much nicer and less problematic than when running under standard windows emulation.

Business. For better or worse, there are plenty of places where the policy and tech allow WSL but installing a hypervisor on your machine is much more challenging.

WSL2 is basically just a Linux VM, though (I think it actually uses Hyper-V containers which seem to occupy some weird space between Linux container and VMs)

Hyper-V is VMs not containers. When you enable Hyper-V it's actually the 'operating system' and Windows is a guest VM believe it or not.

This is how every type 1 hypervisor works.

https://en.wikipedia.org/wiki/Hypervisor

But it's not how type 2 hypervisors work, and traditionally when you used virtual machines on a desktop it was type 2.

That was long time ago, most people using VMware also use ESXi, and mainframes use type 1 as well.

Just those using Virtual Box for free not.

> most people using VMware also use ESXi

Most people using it are also using servers. Is ESXi on desktops/workstations anything other than very niche?

Yeah and? The whole point was about knowing what type 1 hypervisors are all about.

Where they are used is orthogonal to having knowledge.

Managing VMware installations requires having that knowledge.

Sure but, afaik, the docs originally called the "light weight" isolation mode Hyper-V Containers vs the traditional Hyper-V VM with more features exposed

WSL2 is "just a full Linux VM", just easy to install and integrated nicely with other parts of the OS

Lower overhead, and full integration with the host filesystem I believe are the 2 primary reasons many use it

I can pop open an Ubuntu tab in Microsoft Terminal and use *nix utilities as though they were made for Windows. For example: I could chain some scripts and commands together to process and produce reports on my Ableton Live sets without fussing with host extensions in a VM.

Reply on news.ycombinator.com