This is nice and all but I'm much more interested in the inverse use case: managing Windows servers through ssh and bash. For example, it would make loading Let's Encrypt certs much easier.
ELI5: What does this mean to me, as a developer who's really familiar with (and prefers) Linux, but has to use Windows? What cool possibilities will this "Linux Subsystem For Windows" open up to me? What things will it specifically not open up to me?
I think the gist of it / watered down version is that you would be able to compile / run a good amount / most Linux (ubuntu) applications natively-ish in Windows.
For example, I read a blog where someone pulled down the redis source, compiled it, and was running it in wsl. That would be great for me, because it means I can just run redis in my dev environment, instead of turning on a vm to run redis (I know windows ports exist, but I've found it was faster and more reliable to just run the vm).
I also typically add ports of common applications (ls, grep, touch, etc) in my windows environments just to fill the gaps windows has. With this, I probably won't need to do that anymore.
What it opens up is the ability to use the commands and environment that are familiar to you, with as little change as possible. It runs the same bash as a Linux system, at a binary compatibility level. There will still be wrinkles to deal with, such as the case insensitivity of the Windows file systems.
I've been primarily a Linux user since 1995, and have been using Windows 10+WSL for the past week or so, since getting a new laptop.
The only thing I can think of that it'll open up for you is if you want to develop for Windows, while still having Linux available to you. Whether that's for developing games (most of the good game development tools run only on Windows, even if they can deploy to Linux/Mac/Android), or developing for Windows Phone (or whatever it's called this week), using Microsoft technology like Visual Studio and C# (though it's now possible to do most things on Linux via .Net Core, I think), or playing games that aren't available for Linux.
The modern Linux desktop is every bit as good as Windows (and Mac OS), IMHO, though I'd be willing to bet accessibility is still better on Windows. Hardware compatibility for modern devices is probably still better for Windows, too, though old devices certainly are more likely to run on Linux than Windows (a lot of my pro audio and music equipment hasn't run under Windows since like Windows Vista or something, but still works fine under Linux).
From my perspective after a week of tinkering with it, and using is exclusively for my development work, I don't feel like it has given me anything over my Linux environment. But, it does make the Windows development experience much closer to what I expect from Linux. If I developed Windows desktop applications, I would tell a different story, but I develop applications that run on Linux servers; developing on Windows provides no benefits, only negatives...but the negatives have gotten so small that they probably don't matter, anymore. A new developer just learning to build web apps, for example, could do it pretty much as easily on Windows as they could on Linux. Which has never been as true as it is today.
WSL is still very buggy and quirky, and the Windows build it runs on crashes several times a day (at least on my laptop, it does), so it's a net loss in productivity for me, as I'm spending some time waiting on reboots. But, in a few months, it'll probably be stable, widespread, well-documented, etc. And, at that point, maybe it'll even be an ideal option for some developers who are building for cross-platform deployment. Who knows, maybe in a few months, I'll find myself wanting to build something for Windows. I might come back from Linux for a while to spend time in Windows; I certainly don't hate it as much as I used to.
I agree with everything you've said. I also have been using WSL for the past few months. I can run everything in it that I run on Linux and OS X, which is great. I can now officially run any major operating system and have my zsh and vim setup and ready to go in 10 minutes flat.
As for advantages Windows gives me over Linux, there are two.
1) Visual Studio. Even with all of its warts, it's a better debugging tool than anything linux or mac has to offer.
2) Gaming. Games on linux and OS X are catching up, but they're not there yet. I develop as well as play and Unity3D and Unreal Engine, not to mention the graphics drivers, just work better on windows.
> The modern Linux desktop is every bit as good as Windows (and Mac OS), IMHO
I still adamantly oppose linux unless used through SSH.
I use windows for games and as a result just day to day stuff. This makes my life better because I don't have to use putty for SSH. At work being able to use linux tooling is wonderful. Just such a massive improvement over trying to figure out the command shell stuff.
> WSL is still very buggy and quirky, and the Windows build it runs on crashes several times a day (at least on my laptop, it does), so it's a net loss in productivity for me, as I'm spending some time waiting on reboots.
I had some weird video issues. In fairness, it's an insider build. Once they actually release the update maybe it will be better.
"I still adamantly oppose linux unless used through SSH."
Is it a political party? What's to oppose? Use it or don't. I'm surprised to learn there are still folks who feel so negatively about the Linux desktop; to me it just seems like it's gotten so good in the past couple of years. I have almost nothing to complain about (and the complaints I have are pretty specific to my odd use cases, and I have related complaints about Windows and Mac).
I'm curious if you've used a recent Gnome 3 desktop? I used Fedora 23 on my previous laptop and will be installing 24 on this new one. It's a truly excellent desktop experience, in my opinion.
I think it's everything I want from a desktop environment: Reliable, gets out of the way, looks gorgeous and consistent across all of the apps I use (I mention consistency because my terminal in Windows, a variant of ConEmu, looks ridiculously out of place and has a clunky and crowded UI, to boot), and has the right amount of configurability (with options for getting at stuff that isn't visible). I can't think of anything that Windows 10 or Mac OS have that I miss in Linux. The fact that it has my preferred shell environment natively and the most sensible (to me) filesystem paths, and provides a near enough to identical environment as what my CentOS 7 servers run, is just icing on the cake.
In short: What don't you like about Gnome 3 that you do like about the Windows or Mac OS desktop experience?
> I'm curious if you've used a recent Gnome 3 desktop?
Yes. I have such unreasonable demands as thinking that, say, minimising windows might work the way I'd expect on every other OS, or that the extension API might remain stable across ostensibly minor releases, or that features don't disappear on a whim. I would like my desktop to feel like a desktop, and not like a tablet.
"I would like my desktop to feel like a desktop, and not like a tablet."
So, you don't like Windows 10, either, I guess?
I find it to be even more insistently tablet-like than Gnome (which I agree has some elements I previously only associated with Android, and I was slow to adapt, but I've come to be very comfortable with it). My new laptop has a touchscreen, which is a thing I didn't want...but it was the only way to get a 4k screen, so I have it, and I've even come to like it. My habits are evolving slowly, and it'll probably be a year or two before I even remember to touch the screen more than a few times a day, but I think the evolution is a positive one. That said, in the "cons" of having a tablet-like interface, Windows 10 has a "feature" that makes your start menu different every time you open it, which I find very difficult to use (I have to start typing to find anything always). That's just awful. But, I think I would hate it on a tablet, too.
For the other stuff, I kinda like the minimize behavior (but never use minimize...I use virtual desktops to determine what's in and out of view), I think the extensions API has been stabilized for a year or two now. Gnome 3 shipped in some distros when it was not ready (even though they were calling it stable, it wasn't), but it's firmed up a lot. Now, they're mostly doing incremental improvements. Each time I upgrade now (and I've been using Gnome 3 since it first shipped with Fedora) the surprises are small and pleasant (of the form "Oh, that looks nice!").
In short, I agree with some of your criticisms...but, only when applied to Gnome 3 of a year or two or three ago.
>The modern Linux desktop is every bit as good as Windows
Like resizing Windows in Xubuntu? They give you a 1 px border to grab. Bug reported years ago. No fix; happens in all built-in themes. This kind of lack of polish is everywhere.
It's good, it's functional, usable, I'd take it over Windows XP or something, but let's not say it's every bit as good.
What issues did you have? Also, what distribution and desktop environment did you use? I know a blind programmer who uses Fedora with GNOME 3 as his primary OS.
If I can figure out if W10 can turn off its spyware, it'll allow me to ditch VMware for most use cases. Since you can run a Windows X server, I can use graphical emacs. If perf is good enough, I'll run a whole WM like XMonad and use Firefox and stuff inside it.
I'd rather a linux sub sys for windows -- gamers on linux would flock to it, wine sux. There's some nice windows apps, and niche things built only on windows -- but I've staying on Ubuntu most likely.
That's what WINE is, to the nearest approximation. Windows' ABI to userspace is at the dll layer and much larger than Linux. WINE tries to handle that, but falls short in some areas.
I don't think that's fair to Wine. It is tackling a huge problem space, without a specification or source to work with. They're reverse engineering vast swaths of Windows. I think Wine works remarkably well given the size and scope of the project, and the nature of Windows being a huge moving target.
Steam has so many games for Linux now, and a lot of that is due to Wine having advanced to the point where porting to Linux has become relatively trivial. Wine basically has allowed me to go years between reboots into Windows on my previous laptop; and I expect once I get Linux installed on my new laptop, I won't need to reboot into Windows again for another year or two or three. My favorite games are all working on Linux, mostly because of Wine (I don't run them explicitly under Wine, but a lot of the ports on Steam are using Wine behind the scenes).
I'm just saying WOW is buggy on linux wine, etc.. -- I think if windows officially had their own wine-alternative with proprietary drivers, dlls, etc--it would open a whole new world of cross-platform use.
I tried out tmux starting from a week ago after a long time using screen, and after I changed the key-bindings to be more screen-like I have to say that I'm very impressed. I wasn't even close to a screen power user but with tmux I feel like I have screen plus a whole lot of extra.
Yep, works perfectly. Some terminal emulators on Windows suck up some command key combinations which might impede use of tmux or others that need those key combos, but that's usually configurable.
This is not a unique problem to Windows, btw. I use tmux with alt + tilde (`) as my modifier key, and most terminal emulators need config changes to allow this.
WSL is still not the highest quality right now, but they explicitly say that it will be still a beta in Win10 RS1, and it's impressive the progress they managed to do during the few months of the Insider period it has been published. To be clear, it's already largely usable for its advertised purpose (dev) and maybe even more, if you feel adventurous. I guess in RS2 it will be quite solid.
I just got a new laptop, and as it doesn't have an optical drive and I couldn't find a USB flash drive in order to install Linux, I've kinda been stuck on Windows for a few days.
I installed WSL a few days ago. Installation is a ridiculous process, and takes about 24 hours for the service to finally appear after you follow the steps to make it happen, which is not well-documented and just feels idiotic to a Linux user, who's used to being able to change kernels in about three minutes, and switch back and forth at will.
But, I mostly love it. I still plan to install Linux and I'm unlikely to ever boot back into Windows except maybe for testing (I used to boot into Windows for games, but these days Steam has more than enough Linux games to keep me happy, as I don't game much), but Windows is suddenly not the worst of all possible worlds for developing software for Linux, anymore.
It has some weird/scary quirks: Using an editor installed on the "Windows" side of the system to edit files on the "Linux" side of the system led to files disappearing completely. I also saw the usual Windows CR/LF issues in sharing files between Linux and Windows partitions, but also an additional weird thing that might be Unicode related: I piped output from a command into a file on the Windows side in a Powershell session, opened it on the Linux side, and found it had spaces after every character and some other bizarre stuff inserted into it; kinda like maybe it was created in a UTF encoding but then loaded in a non-Unicode encoding or something. These are probably things that have some reasonable explanation and some way to work around them, but they made me scared enough of inter-operation between the two sides that I now isolate the two completely. I don't try to work "back and forth". If it's a Linux thing, I work only with Linux tools, and vice versa. This doesn't seem to be as big of an issue when using the MSYS bash shell that comes with git, which is my other bash shell in this system. At least, using the git bash shell hasn't led to any files disappearing.
Another interesting issue: Permissions are super lax, by default, in the Linux side. Owned by root:root and mode 777 across the board in your home directory, which causes a lot of software to complain. ssh, Ruby bundler (downgrade to 1.12.3 if you're having problems running bundle on Windows, either under WSL or in Powershell; same solution works for different bugs), and a bunch of other stuff has weird quirks when given a directory that is 777. It's also just plain nuts. I guess Windows has a different security model and this may not be a crazy as it seems at first glance, but it does cause software to freak out. But, as I understand it, WSL didn't even have a user account until recently, so they seem to be making it better on that front. In the meantime, I wouldn't want to rely on it for anything important.
That said, it gets a lot right. It feels like magic to run nearly any Linux application and have it Just Work. I, admittedly, am only running text mode apps at the moment: vim, bash, git, Perl, Ruby, Node. I haven't tried any graphical apps, so far. But, that's enough to replicate most of my workflow for developing server apps, and I don't need a VM to do it.
I don't plan to stick with Windows, even now, but it's been fun playing with it, and seeing what Windows is like these days. It's pretty nice, honestly. If I were just starting out with computers, I'd probably have no strong need to move to Linux in order to be productive. I'd learn how to use Powershell effectively (though it's gotten a lot easier/nicer in recent years, and accepts many of the common commands a UNIX user would expect, like ls, cd, cp, mv, etc...last time I used it many years ago, every command was some "Stupid-Long-Ass-Thing WithBIzarreLY cApitalized-OptioNS"), and probably be happy with Windows as a developmen...
You are talking about enabling insider preview which takes around a day to kick in and offer you the update. Since bash on windows is not included on the stable "branch" you have to go through the hoops to get it. This has nothing to do with the feature itself, which takes even less to install than pulling the Ubuntu image off docker hub.
I found it really clunky. I wanted a thing, and it made me wait nearly two days to get that thing; and it was confusing, to boot; I kept following the instructions and not seeing the service I expected to see. I now understand the process and what it was doing during all that time (er, actually, I still don't know what it was doing during all that time, but I understand it was working as designed in making me wait), I just think it is a clunky process that exhibits one of the bigger weaknesses of Windows. Windows simply handles updates and packages extremely poorly compared to Linux.
It's interesting that the answers I've gotten about it being a poor user experience are all basically of the form, "No, that's how it is supposed to work. It isn't a poor user experience." Have y'all not used Linux before? I mean, apt-get and yum are really something special, if you're coming from systems (like Windows or Mac OS) that don't have good package management. Having one universal method to install and update everything on your system, including OS updates, is just wonderful.
Nonetheless, Windows 10 is the best Windows I've used (ignoring the privacy concerns among other things), and I like WSL. It's cool technology and fun to play with. I'm probably installing Linux tonight or tomorrow, but it was a fun week of tinkering with something new to me.
> I installed WSL a few days ago. Installation is a ridiculous process, and takes about 24 hours for the service to finally appear after you follow the steps to make it happen, which is not well-documented and just feels idiotic to a Linux user, who's used to being able to change kernels in about three minutes, and switch back and forth at will.
That is not my experience.
edit: I already had insider builds.
I don't know if it is fair to include the OS update time, but I can see why it is frustrating.
I grew up through some of the more interesting business practices Microsoft employed, and it's several attempts to harm the Linux ecosystem. I would love to see this move by Microsoft in a positive light, but can't help but be suspicious ? For now the possibilities it opens are great, however I wish I understood Microsoft's strategic view of this and how it might evolve over time for Linux.
MS is now a services business, they want to sell you their cloud. They can't do that if the linux ecosystem is shut out of Windows as it was before. Simply supporting Linux VMs is not enough, because long-term they still want to maintain some control on the OS market. So they open enough doors that you (or your corporate boss) will be enticed to consider their offerings and move to their cloud. You have some sweet cloud-management tool that runs on Linux? "You can now run it in the same way on Windows! Oh and with Windows in our cloud you also get all these other goodies..." - et voilà, locked back in.
Most "advanced" cloud services are built with the aim of locking you in (AWS, Oracle, Heroku, etc etc): easy to adopt and easy to extend, which means "hard to leave". Microsoft is no exception.
Microsoft still seems very divided. On one hand we have cool developer initiatives, but on the other we still have anti-user behavior like forced W10 upgrades, encrypted black-box telemetry that cannot be disabled, Android patent extortion, and lawsuits over FAT support in other operating systems.
I would love to see MS get their crap together, but it still seems like they can't break their addiction to abusive behavior.
Strategic View is simple I believe, they want developers back in there fold. MS sells Windows and most devs have no problem with its UI but want to use linux CLI, also some new tools/languages are developed only/first for Linux. Now OSX being compatible has captured this lucrative marketshare.
Having good developers using Windows will be a great boost for ecosystem, a percentage may choose to experiment with UWP, a react-native developer may try a Windows version of app.
Only thing to loose is people may not build cli tools for Windows, but that is already a lost cause. Linux won't pose a threat to GUI, so no chance of non developers getting comfortable and migrating to Linux. So win-win and later some more win.
And WSL could potentially be a much better experience than OS X's "real Unix" with Homebrew and etc.
The paths and base configuration will match your production environment, no worries about version mismatches or missing extensions. Ubuntu packages generally come with a sane and ready-to-run configuration, with service scripts and the like. (I use pkgsrc which seems more polished than Homebrew, but the installing experience is still not as nice as Ubuntu. Maybe someone will make a "Linux subsystem for Mac"...)
If anything this helps Linux because it cements popular Linux distros as the default *nix environments.
I think "Linux subsystem for Mac" is stuff like Vagrant, where you can still work on your dev projects locally/natively but they're shared into a Linux environment that can be set up to mimic your production environments..
This can be done today on Windows too of course, even without this new fancy thing.
Sure, the serious use case is a VM or Docker or etcetera-virtual based solution. However I must have a couple dozen maintenance-mode projects on my system and there's a certain appeal to alt-tabbing from MS Excel or Photoshop to a console window and typing `service start foo` (or the launchd equivalent).
There are still a gazillion people using WAMP (and so on) as a bad impersonation of their production environment, so this fills a role.
I haven't tried W10 at all, but it's likely they didn't implement all of the terminal features required by screen, or that they did, but $TERM is set to something not recognized by [the terminal library that presumably underpins] screen.
For any new programmers who want to learn more, search for "ANSI escape codes", VT100, "terminal emulation", termcap, and terminfo. Maybe add curses and pseudoterminal for extra credit.
You can certainly run MariaDB. All you need to do is "apt-get install mariadb-server". The Linux subsystem is the entire Ubuntu 14.04 system. I love my Surface Pro 4 with latest Windows 10 release, only 1 week or so to go before its outside preview program. Look forward to see what Devs come up with. Some of the cooler stuff I've seen is using the Linux shell with Visual Studio Code as the shell on a windows system.
The strategic view is not hard to understand. They got their clock cleaned on the server side. They are irrelevant in the IoT. Machine learning, docker, apache spark run best on linux. The direct-x gaming advantage will erode as Vulkan opens up the industry. They failed spectacularly in mobile as the vast majority of connected devices run Android and iOS. ChromeOS is about to run native Android apps which marks the end of charging for windows desktop, which is why they're so desperate for win10 upgrades to monetize user tracking. Some see their embracing of open-source as the new MS but they have no alternative.
I am really looking forward to this for my Surface Book. Historically I've loaded up a virtual machine running Linux, this was more complete and had fewer hiccups than a Cygwin type of setup for me. Of course it would be REALLY useful if they also support NFS for mounting stuff from my NAS appliance and a compositor (Wayland/X11/Etc.) that would allow things that create windows to work as well.
It's on the slow ring now, too, which is where I'm using it. I'm quite impressed so far. There's some performance overhead in some cases (e.g. Hugo for Linux is about half the speed in the lxrun environment than Hugo for Windows on the same machine), but it's very pleasant.
It's nice and what I use day-in/day-out... WSL on Surface Book.
The first iterations were a bit dodgy in my experience, but it's shaken out nicely. The only real, very nearly show stopper for me is that the console still stinks. That's not to say that Microsoft hasn't been working on it and making progress... but it's not as quality an experience as you'd have using one the Linux terminal programs in a GUI. I find emacs with not a whole lot of custom key mapping to be especially annoying; I swear sometimes that half the key bindings I want to use in org-mode are mapped first as short-cuts to the Windows console program itself.
Any reason to prefer nfsv3 to samba/cifs? As far I've been able to figure out nfsv4 is about as popular and well-supported as ipv6 was 10 years ago... And If you're stuck with nfsv3 - is that really a happy place?
Why not just use the good old python windows installer? I am using it for many years on various systems for both pytgon 2x and 3x and never had any problem or thought that it was difficult.
On the other hand, installing a different than the default version of python on a linux system is not something that a newbie can do (since it involves getting the sjc, compiling it and making an altinstall).
My experience with WSL has been excellent so far, I see it making Cygwin obsolete almost entirely. Cygwin was always a pain to set up and occasionally very buggy.
And slow. For some reason all commands I execute through cygwin seem to be taking a little more than if executed through a normal windows command prompt.
This really breaks the cygwin experience for me :/
Appreciate this is a minor point but the name Windows Subsystem for Linux confuses me. Am I missing how to read it properly? It seems more like a Linux subsystem for Windows to me :-)
How stable is it? I've been playing around with the idea installing it on my only (dev)machine. I dont expect the ubuntu subsystem to work perfectly but my old workflow to be stable.
Edit: I am regularly suprised that nano filename does not open quick edit in my machine. And then I remeber I am working in my local machine.
It's pretty stable. But also pretty useless. Fun to play around and use a bit of byobu. I had a repo that wouldn't check out in windows because of some file names. But bash on Ubuntu on Windows worked like a champ.
But as I said, the utility is limited. If you really need the Linux kernel, this is useless. If you don't, then everything is already available on Windows natively.
THis makes no sense.. Most people don't comprehend that it's not just bash, its the entire Ubuntu user space. You can run Ubuntu apps just by running apt-get. VI, Ruby, Golang, Rust. You can set Bash as your debug shell in VS Code, you can have all your familiar SSH keys/forwarding/scp tools.. Of course there is a ton of windows native stuff, but in essence this makes Linux native so now you have just that much more.
I think many of us who use Windows take for granted the number of tools we use to cover over our pain points.
I use Msys2, Cmder, a bunch of custom tool builds, and some Powershell scripts. This setup works great; I'm productive with it and my command line feels fantastic. However, I have been tweaking my setup for years. WSL makes all of that learning, setup, and tweaking unnecessary.
All of your tools are available out of the box, and you don't have to configure anything, hunt down binaries, or make your own builds.
This isn't about making you and I more comfortable - we already are. This is about making it easier for other devs to find a happy setup on Windows.
I found that blog post to be very disappointing. It didn't really demonstrate a meaningful use case for WSL (i.e. programming in C by compiling with gcc). I think that blog post is mostly made for tech journalists who aren't actually very techy. For example, the author used nano to copy and paste in a Python script. There's a two things wrong with that: 1) he used nano for "programming" (regardless of which text editor you use, I think we can all agree that nano definitely shouldn't be used for programming) and 2) Python already has native support on Windows.
"It’s not about replacing other Windows-native tools such as Python, PowerShell, C# and more. It’s about removing friction when you just need to get stuff done quickly and efficiently in a modern development workflow."
My example was meant to support my assertion that:
> I think that blog post is mostly made for tech journalists who aren't actually very techy.
---
This statement:
> This is definitely a “by developers, for developers” Windows 10 feature, specifically designed to remove a bit of friction from developers’ daily workflow.
reads like a press release. I agree that implementing WSL will reduce "friction from developers' daily workflow", but the article did a fairly poor job at demonstrating that.
I think you're right though I wasn't disappointed by the fact. I saw quickly that the audience wasn't me and kept reading for curiosity's sake... to test who they're trying to reach.
There's probably a fair number of Windows only people that the whole Linux ecosystem is something new and foreign; especially sysadmin and corporate IT types that have heard about this, but may not know anything about it... I expect there may be some gentle introduction for those types here, too.
That name bothers me. I understand that people don't want to use the name GNU/Linux because it's stupid and overcomplicated, but in this case there is no Linux at all. This is quite literally GNU running on top of the NT kernel, so why use the Linux name at all?
On a side note, the irony of GNU/NT must be killing Stallman.
Cygwin is GNU on top of the NT kernel. WSL is an implementation of a Linux ABI. In theory, you could run a non-GNU userland environment compiled for Linux on top of WSL.
98 comments
[ 3.8 ms ] story [ 173 ms ] threadFor example, I read a blog where someone pulled down the redis source, compiled it, and was running it in wsl. That would be great for me, because it means I can just run redis in my dev environment, instead of turning on a vm to run redis (I know windows ports exist, but I've found it was faster and more reliable to just run the vm).
I also typically add ports of common applications (ls, grep, touch, etc) in my windows environments just to fill the gaps windows has. With this, I probably won't need to do that anymore.
This subsystem runs the exact binaries that you would run in your native boot system.
It's as simple as "apt-get install redis-server redis-tools"
The only thing I can think of that it'll open up for you is if you want to develop for Windows, while still having Linux available to you. Whether that's for developing games (most of the good game development tools run only on Windows, even if they can deploy to Linux/Mac/Android), or developing for Windows Phone (or whatever it's called this week), using Microsoft technology like Visual Studio and C# (though it's now possible to do most things on Linux via .Net Core, I think), or playing games that aren't available for Linux.
The modern Linux desktop is every bit as good as Windows (and Mac OS), IMHO, though I'd be willing to bet accessibility is still better on Windows. Hardware compatibility for modern devices is probably still better for Windows, too, though old devices certainly are more likely to run on Linux than Windows (a lot of my pro audio and music equipment hasn't run under Windows since like Windows Vista or something, but still works fine under Linux).
From my perspective after a week of tinkering with it, and using is exclusively for my development work, I don't feel like it has given me anything over my Linux environment. But, it does make the Windows development experience much closer to what I expect from Linux. If I developed Windows desktop applications, I would tell a different story, but I develop applications that run on Linux servers; developing on Windows provides no benefits, only negatives...but the negatives have gotten so small that they probably don't matter, anymore. A new developer just learning to build web apps, for example, could do it pretty much as easily on Windows as they could on Linux. Which has never been as true as it is today.
WSL is still very buggy and quirky, and the Windows build it runs on crashes several times a day (at least on my laptop, it does), so it's a net loss in productivity for me, as I'm spending some time waiting on reboots. But, in a few months, it'll probably be stable, widespread, well-documented, etc. And, at that point, maybe it'll even be an ideal option for some developers who are building for cross-platform deployment. Who knows, maybe in a few months, I'll find myself wanting to build something for Windows. I might come back from Linux for a while to spend time in Windows; I certainly don't hate it as much as I used to.
As for advantages Windows gives me over Linux, there are two.
1) Visual Studio. Even with all of its warts, it's a better debugging tool than anything linux or mac has to offer.
2) Gaming. Games on linux and OS X are catching up, but they're not there yet. I develop as well as play and Unity3D and Unreal Engine, not to mention the graphics drivers, just work better on windows.
I still adamantly oppose linux unless used through SSH.
I use windows for games and as a result just day to day stuff. This makes my life better because I don't have to use putty for SSH. At work being able to use linux tooling is wonderful. Just such a massive improvement over trying to figure out the command shell stuff.
> WSL is still very buggy and quirky, and the Windows build it runs on crashes several times a day (at least on my laptop, it does), so it's a net loss in productivity for me, as I'm spending some time waiting on reboots.
I had some weird video issues. In fairness, it's an insider build. Once they actually release the update maybe it will be better.
Is it a political party? What's to oppose? Use it or don't. I'm surprised to learn there are still folks who feel so negatively about the Linux desktop; to me it just seems like it's gotten so good in the past couple of years. I have almost nothing to complain about (and the complaints I have are pretty specific to my odd use cases, and I have related complaints about Windows and Mac).
The politest thing that I can say about that is "there's no accounting for taste".
I think it's everything I want from a desktop environment: Reliable, gets out of the way, looks gorgeous and consistent across all of the apps I use (I mention consistency because my terminal in Windows, a variant of ConEmu, looks ridiculously out of place and has a clunky and crowded UI, to boot), and has the right amount of configurability (with options for getting at stuff that isn't visible). I can't think of anything that Windows 10 or Mac OS have that I miss in Linux. The fact that it has my preferred shell environment natively and the most sensible (to me) filesystem paths, and provides a near enough to identical environment as what my CentOS 7 servers run, is just icing on the cake.
In short: What don't you like about Gnome 3 that you do like about the Windows or Mac OS desktop experience?
Yes. I have such unreasonable demands as thinking that, say, minimising windows might work the way I'd expect on every other OS, or that the extension API might remain stable across ostensibly minor releases, or that features don't disappear on a whim. I would like my desktop to feel like a desktop, and not like a tablet.
So, you don't like Windows 10, either, I guess?
I find it to be even more insistently tablet-like than Gnome (which I agree has some elements I previously only associated with Android, and I was slow to adapt, but I've come to be very comfortable with it). My new laptop has a touchscreen, which is a thing I didn't want...but it was the only way to get a 4k screen, so I have it, and I've even come to like it. My habits are evolving slowly, and it'll probably be a year or two before I even remember to touch the screen more than a few times a day, but I think the evolution is a positive one. That said, in the "cons" of having a tablet-like interface, Windows 10 has a "feature" that makes your start menu different every time you open it, which I find very difficult to use (I have to start typing to find anything always). That's just awful. But, I think I would hate it on a tablet, too.
For the other stuff, I kinda like the minimize behavior (but never use minimize...I use virtual desktops to determine what's in and out of view), I think the extensions API has been stabilized for a year or two now. Gnome 3 shipped in some distros when it was not ready (even though they were calling it stable, it wasn't), but it's firmed up a lot. Now, they're mostly doing incremental improvements. Each time I upgrade now (and I've been using Gnome 3 since it first shipped with Fedora) the surprises are small and pleasant (of the form "Oh, that looks nice!").
In short, I agree with some of your criticisms...but, only when applied to Gnome 3 of a year or two or three ago.
Like resizing Windows in Xubuntu? They give you a 1 px border to grab. Bug reported years ago. No fix; happens in all built-in themes. This kind of lack of polish is everywhere.
It's good, it's functional, usable, I'd take it over Windows XP or something, but let's not say it's every bit as good.
You honestly have no idea how much. Desktop accessibility in Linux, at least for me as a screen reader user, is a joke.
What advantages would this give you over running Firefox natively on Windows?
Steam has so many games for Linux now, and a lot of that is due to Wine having advanced to the point where porting to Linux has become relatively trivial. Wine basically has allowed me to go years between reboots into Windows on my previous laptop; and I expect once I get Linux installed on my new laptop, I won't need to reboot into Windows again for another year or two or three. My favorite games are all working on Linux, mostly because of Wine (I don't run them explicitly under Wine, but a lot of the ports on Steam are using Wine behind the scenes).
Once the authors discovers GNUScreen/tmux, xargs and other tools, his life will change :)
WSL is still not the highest quality right now, but they explicitly say that it will be still a beta in Win10 RS1, and it's impressive the progress they managed to do during the few months of the Insider period it has been published. To be clear, it's already largely usable for its advertised purpose (dev) and maybe even more, if you feel adventurous. I guess in RS2 it will be quite solid.
I installed WSL a few days ago. Installation is a ridiculous process, and takes about 24 hours for the service to finally appear after you follow the steps to make it happen, which is not well-documented and just feels idiotic to a Linux user, who's used to being able to change kernels in about three minutes, and switch back and forth at will.
But, I mostly love it. I still plan to install Linux and I'm unlikely to ever boot back into Windows except maybe for testing (I used to boot into Windows for games, but these days Steam has more than enough Linux games to keep me happy, as I don't game much), but Windows is suddenly not the worst of all possible worlds for developing software for Linux, anymore.
It has some weird/scary quirks: Using an editor installed on the "Windows" side of the system to edit files on the "Linux" side of the system led to files disappearing completely. I also saw the usual Windows CR/LF issues in sharing files between Linux and Windows partitions, but also an additional weird thing that might be Unicode related: I piped output from a command into a file on the Windows side in a Powershell session, opened it on the Linux side, and found it had spaces after every character and some other bizarre stuff inserted into it; kinda like maybe it was created in a UTF encoding but then loaded in a non-Unicode encoding or something. These are probably things that have some reasonable explanation and some way to work around them, but they made me scared enough of inter-operation between the two sides that I now isolate the two completely. I don't try to work "back and forth". If it's a Linux thing, I work only with Linux tools, and vice versa. This doesn't seem to be as big of an issue when using the MSYS bash shell that comes with git, which is my other bash shell in this system. At least, using the git bash shell hasn't led to any files disappearing.
Another interesting issue: Permissions are super lax, by default, in the Linux side. Owned by root:root and mode 777 across the board in your home directory, which causes a lot of software to complain. ssh, Ruby bundler (downgrade to 1.12.3 if you're having problems running bundle on Windows, either under WSL or in Powershell; same solution works for different bugs), and a bunch of other stuff has weird quirks when given a directory that is 777. It's also just plain nuts. I guess Windows has a different security model and this may not be a crazy as it seems at first glance, but it does cause software to freak out. But, as I understand it, WSL didn't even have a user account until recently, so they seem to be making it better on that front. In the meantime, I wouldn't want to rely on it for anything important.
That said, it gets a lot right. It feels like magic to run nearly any Linux application and have it Just Work. I, admittedly, am only running text mode apps at the moment: vim, bash, git, Perl, Ruby, Node. I haven't tried any graphical apps, so far. But, that's enough to replicate most of my workflow for developing server apps, and I don't need a VM to do it.
I don't plan to stick with Windows, even now, but it's been fun playing with it, and seeing what Windows is like these days. It's pretty nice, honestly. If I were just starting out with computers, I'd probably have no strong need to move to Linux in order to be productive. I'd learn how to use Powershell effectively (though it's gotten a lot easier/nicer in recent years, and accepts many of the common commands a UNIX user would expect, like ls, cd, cp, mv, etc...last time I used it many years ago, every command was some "Stupid-Long-Ass-Thing WithBIzarreLY cApitalized-OptioNS"), and probably be happy with Windows as a developmen...
It's interesting that the answers I've gotten about it being a poor user experience are all basically of the form, "No, that's how it is supposed to work. It isn't a poor user experience." Have y'all not used Linux before? I mean, apt-get and yum are really something special, if you're coming from systems (like Windows or Mac OS) that don't have good package management. Having one universal method to install and update everything on your system, including OS updates, is just wonderful.
Nonetheless, Windows 10 is the best Windows I've used (ignoring the privacy concerns among other things), and I like WSL. It's cool technology and fun to play with. I'm probably installing Linux tonight or tomorrow, but it was a fun week of tinkering with something new to me.
That "thing" isn't officially released yet, hence the hoops.
That is not my experience.
edit: I already had insider builds.
I don't know if it is fair to include the OS update time, but I can see why it is frustrating.
Weird. It worked instantly for me. It did need a reboot though.
Most "advanced" cloud services are built with the aim of locking you in (AWS, Oracle, Heroku, etc etc): easy to adopt and easy to extend, which means "hard to leave". Microsoft is no exception.
This microsoft are no longer the ridiculously dominant force they once were.
Linux is a core part of their cloud business now.
They have a new CEO, and a lot of time has passed.
They are actively trying to change a lot of their culture and approach to things.
I would love to see MS get their crap together, but it still seems like they can't break their addiction to abusive behavior.
Having good developers using Windows will be a great boost for ecosystem, a percentage may choose to experiment with UWP, a react-native developer may try a Windows version of app.
Only thing to loose is people may not build cli tools for Windows, but that is already a lost cause. Linux won't pose a threat to GUI, so no chance of non developers getting comfortable and migrating to Linux. So win-win and later some more win.
The paths and base configuration will match your production environment, no worries about version mismatches or missing extensions. Ubuntu packages generally come with a sane and ready-to-run configuration, with service scripts and the like. (I use pkgsrc which seems more polished than Homebrew, but the installing experience is still not as nice as Ubuntu. Maybe someone will make a "Linux subsystem for Mac"...)
If anything this helps Linux because it cements popular Linux distros as the default *nix environments.
This can be done today on Windows too of course, even without this new fancy thing.
There are still a gazillion people using WAMP (and so on) as a bad impersonation of their production environment, so this fills a role.
Here's a question though that I seem to not see answered (but maybe I'm not reading enough):
Can I run MariaDB or nginx or other "servers" in this environment? Or does that stuff not work?
For any new programmers who want to learn more, search for "ANSI escape codes", VT100, "terminal emulation", termcap, and terminfo. Maybe add curses and pseudoterminal for extra credit.
you can run an X server from windows and connect to it from linux. Not bug free but it does work.
https://github.com/billziss-gh/winfsp
The first iterations were a bit dodgy in my experience, but it's shaken out nicely. The only real, very nearly show stopper for me is that the console still stinks. That's not to say that Microsoft hasn't been working on it and making progress... but it's not as quality an experience as you'd have using one the Linux terminal programs in a GUI. I find emacs with not a whole lot of custom key mapping to be especially annoying; I swear sometimes that half the key bindings I want to use in org-mode are mapped first as short-cuts to the Windows console program itself.
If you just want command line tools, there's Cygwin.
On the other hand, installing a different than the default version of python on a linux system is not something that a newbie can do (since it involves getting the sjc, compiling it and making an altinstall).
This really breaks the cygwin experience for me :/
Edit: I am regularly suprised that nano filename does not open quick edit in my machine. And then I remeber I am working in my local machine.
But as I said, the utility is limited. If you really need the Linux kernel, this is useless. If you don't, then everything is already available on Windows natively.
But I already have vim, ruby, golang and rust natively on Windows. I agree it's cool. But still pointless.
>You can set Bash as your debug shell in VS Code, you can have all your familiar SSH keys/forwarding/scp tools.
I can install vscode on Linux directly. And again, ssh, scp, etc is already there on Windows.
I use Msys2, Cmder, a bunch of custom tool builds, and some Powershell scripts. This setup works great; I'm productive with it and my command line feels fantastic. However, I have been tweaking my setup for years. WSL makes all of that learning, setup, and tweaking unnecessary.
All of your tools are available out of the box, and you don't have to configure anything, hunt down binaries, or make your own builds.
This isn't about making you and I more comfortable - we already are. This is about making it easier for other devs to find a happy setup on Windows.
https://www.nano-editor.org/dist/v2.5/NT/
"It’s not about replacing other Windows-native tools such as Python, PowerShell, C# and more. It’s about removing friction when you just need to get stuff done quickly and efficiently in a modern development workflow."
> I think that blog post is mostly made for tech journalists who aren't actually very techy.
---
This statement:
> This is definitely a “by developers, for developers” Windows 10 feature, specifically designed to remove a bit of friction from developers’ daily workflow.
reads like a press release. I agree that implementing WSL will reduce "friction from developers' daily workflow", but the article did a fairly poor job at demonstrating that.
There's probably a fair number of Windows only people that the whole Linux ecosystem is something new and foreign; especially sysadmin and corporate IT types that have heard about this, but may not know anything about it... I expect there may be some gentle introduction for those types here, too.
On a side note, the irony of GNU/NT must be killing Stallman.