Slashdot Mirror
Windows 10 Spring Update Improves Linux On WSL With Unix Sockets and More (anandtech.com)
Billly Gates writes: Windows 10 build 1803 has come out this month, but with some problems. AnandTech has a deep-dive with the review examing many new features including the much better support for Linux. WSL (Windows Subsystem for Linux) now has native Curt and Tar from the command prompt as well as a utility to convert Unix to Windows pathnames called WSLpath.exe which is documented here. In addition it was mentioned on Slashdot in the past about OpenSSH being ported natively to Win32 in certain early builds. It now seems the reason was for Linux interoperability with this Spring Update 2. Unix sockets mean you can run Kali Linux on Windows 10 for penetration testing or run an Apache server in the background with full Linux networking support. Deemons now run in the background even with the command prompt closed. [...]
They can get Linux to run as a layer or app under Windows, but can't figure out how to get it to run from a different drive. Some of us have a smaller SSD as C drive and would like to at least play with this without having to jump through hoops. Yes, I know about symlinks in Windows, but it's kind of a crappy fix, if it works at all.
All this work to be compatible with Linux is only being done because Microsoft is desperately trying to get a ticket on the cloud moneytrain its ignored for nearly a decade. So far bootstrapping things like Kinect, office, and exchange to their cloud offering has boosted its presence in much the same way that paying hosting providers to switch their park-web sites to IIS static pages improved their netcraft numbers.
Curt and Tar from the command prompt
i presume we mean curl but this is moot. Anyone who needed curl or tar "from the command prompt" (as if it came anywhere else?) has their macbook, or their linux system...and they have it for free in the amazon cloud as well as the 40 some other openstack players that exist.
it was mentioned on Slashdot in the past about OpenSSH being ported natively to Win32 in certain early builds.
Embrace extend extinguish only works when theres a product with a bottom line people are willing to choose. "becoming linux" isnt doing anything to Redmond but wasting programmer hours trying to catch a lizard. Larry Ellison learned this fact with MySQL. GPL is an armored license that prohibits the type of early nineties chicanery Microsoft was absolutely legendary for pulling on small companies and startup projects.
Deemons now run in the background even with the command prompt closed.
Daemons,Welcome to 1991. You could also just pick a cloud provider with a competent ecosystem that will natively run any of five or more major linux distributions that your programmers are already familiar with.
Good people go to bed earlier.
You may not but I do. My work requires that I use Windows. With WSL I can use all the classic Linux commands that I've been using for 15 years. The better they make it the happier I am.
Nope. On Windows you can type 'Curt' and it runs just fine.
Windows 10 introduces entire filesystem-insensitivity.
who sees Microsoft using all of their friendliness towards open source, and Linux in particular, to "soften" the stance of using Linux as a standalone solution in lieu of actually using Linux. In the last few years, Linux has been its own worst enemy. What with systemd, utter balkanization, no "usable" desktop to go mainstream with (save Red Hat using Gnome internally), getting into bed with Azure.
As a former Unix sysadmin who had to move cross country for family reasons, I have been forced into Windows jobs because I have to live in flyover country in a small town, which already has a dearth of remotely decent IT jobs. Windows is slowly killing the independence of Linux and in my mind, at least, no one seems to care. The "cloud" is the death of IT as we know it. Anymore, unless you're a programmer or hardware designer, you are working in and supporting the "cloud". I remember vast data centers teeming with talented people working on FreeBSD/OpenBSD Web servers with PostGreSQL backends. I remember running *nix-based firewalls and proxy servers, running CheckPoint atop Nokia IPSO boxes. I remember running and writing tons of sh and Perl scripts before the advent of stupidity like PHP and Word Press. We are seeing before our eyes, the dumbing down of *nix. I for one sorely miss pure Sun Solaris environments running atop Sun Sparc workstations. I'd happily take an FVWM WM over anything today. What the hell happened? Anymore, consumers buy machines that are only fit for consumption, not creation. We have given away the family silver.
Or you could just do it the sane way and run Linux, with Windows in a VM for running the one (or few) applications that you absolutely must have Windows to run.
Guns don't kill people; Physics kills people! - John Lithgow as Dick Solomon on Third Rock From The Sun
A big step towards running systemd on Windows!
Unix sockets mean you can run Kali Linux on Windows 10 for penetration testing or run an Apache server in the background with full Linux networking support.
You could just run Linux (and maybe Windows 10 in a VM).
It must have been something you assimilated. . . .
"There is nothing that microsoft has contributed for open source which doesnt help their competing operating system, windows in some way." Dude, once again, they are a company. They make profit out of it. You can have them contribute to open source and make profit from it, or make propietary code, and make profit for it. Plain and simple.
Eventually the Linux subsystem will be all that's left of Windows. Maybe a legacy support module on the side.
You'd have to call it the Linux supersystem by then, I think. Also "mission accomplished"!
Now you understand why the SystemD is on such a big phagocytosis spree :
Swallowing the whole Microsoft Windows into "system-msctl" was Lennart's secret end goal all along !
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]
The differences are :
- On Windows, the use of .ini files has completely disappeared. Instead the registry hives being an opaque format that can be only access (in theory) with Windows' API of regedit. This would make it impossible to hand access them manually, say from a boot stick. (Well, in practice, there are 3rd party Linux tools able to access the hive format, so fixing from a boot stick is possible, but you got the idea).
- In gnome with gconfig the configuration is still stored in sets of plain XML files. Only they are now stored in an organized fashion in a specific set of subdirectories, and there's a centralized API and tool set to access them. But they are still human readable (you could still edit them with your favorite editor - emacs/vim/nano/ed) and easily machine readable (e.g. with your favorite Perl module such as XML::Twig).
The windows equivalent would have been if the .ini were kept, but now Microsoft defined a specific path to store them (e.g.: in a specifc subfolder tree within %USER_PROFILE% or whatever, instead of all over the place like in good old Win 3.x days) and mandated a specific API to manipulate them.
The closest thing to Windows' registry in Linux-land would be journald's internal database format, except that it has a very well documented format and journald forwards messages to any of your favorite system logger as soon as that deamon startsup - and it is configured to do so by default on virtually all the GNU/Linux distributions except for the most storage-starved ones in embed systems (e.g.: mercore/Sailfish OS doesn't have a syslog forwarding setup by default because it has to run on your smartphone limited internal flash. But Raspbian forwards to syslog by default).
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]
but I fear the ONLY reason ms is doing this is to further advance azure's penetration and to get more linux admins on windows systems.
(and Linux devs to switch away from Mac OS X systems)
Well that, and also the initial idea was to offer enough of the Linux API exposed by the Windows kernel so Windows Mobile could eventually run Android applications so that their OS wouldn't have been the irrelevant joke without any significant app ecosystem.
Except that they didn't manage that even by far. They are light-years away from even running the simplest Android apps, so WSL is what they pivoted to in order to salvage the invested work.
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]
Linux and all GPL software should move to GPL4, which should be identical to GPL2 with an added clause stating that Microsoft and Microsoft employees may not use any of it.
No need, and it would go against the GPL spirit.
The fundamental idea of GPL is that you should be able to do whatever you want with GPLed code, as long as you make sure that anyone you forward the code can still enjoy the same "whatever you want" freedom that your received. (that last part being the key difference with BSD-like persmissive license).
Restricting an imaginary GPLv4 against microsoft would go against the "to whatever you want" part. (And wouldn't be of any use, Microsoft would simply spin off a separate company to handle such GPLv4 code).
Also, by making mandatory to keep the same freedom to the next in line, GPL is pretty robust against EEE : you can't leverage extensions much if you have to publish them due to GPL, and you can't extinguish something that's freely available.
There's a reason why the older microsoft guard were shitting their pants and calling GPL "cancer" : RMS had designed something that incidentally happens to be completely EEE-proof.
The modifications of further GPL version were just about patching circumventions that some companies have found around the "keep the same freedom to the next in line" part.
GPLv1 made it mandatory to make source available together with the software.
Companies: "here's the code, but you can't legally do anything with it, because it's patent covered and you're violating our IP"
GPLv2 made it mandatory to grant access to the patents, without royalties.
Companies (e.g. TiVo): "here's the code, but in practice you can't really modify it because uploading your mods requires our secret cryptographic key"
GPLv3 made mandatory to provide a way (e.g.: key provided, unlockable bootloader, etc.) to actually be able to use modification in practice.
Currently there's no apparent need for a GPLv4 : no company has invented a way to give you the code, the patents license and the cryptographic keys, but still prevent you from actually modifying the code.
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]
i presume we mean curl but this is moot. Anyone who needed curl or tar "from the command prompt" (as if it came anywhere else?) {...}
Yes, curl comes anywhere else.
It has also a library (libcurl) and that library is used for web interface by lots of modules.
Probably lots of GUI application use curl as their peculiar backend to download stuff.
Except that in WSL's specific case, support for GUI isn't stellar (basically, you need to X-over-network to a Windows native X-Server), so probably nobody is using GUI, and in practice, yes, curl is mostly only used on the command line in Windows-land.
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]
Nae king! Nae laird! Nae yurrupiean pressedent! We willna be fooled again!
I don't care what you want, but I use it so that I can use some nice Linux commands in Windows. For example wget, cloc etc. Those can be installed in native windows, but it's more quicker to install them in the wsl. It's a better version of "command line". Unfortunately I don't get to choose my work OS.
This means that I will be able to run Linux as a bag on the side of your wonderful Windows 10 product with two big advantages over a normal Linux installation: I will have the enormous overhead and slow boot time of Windows to deal with daily. And, Windows 10 will continue to spy on my every move and report it to you without telling me. I can hardly wait to get on the bandwagon with this one.
They do it because they have an SOE, and people stepping outside the SOE costs them money, time and frustration.
This. It would be petty of me to change jobs because I can't use the OS of my choice. Windows works and does the job fine. The company I work for is very particular about treating everyone fairly. For example, I did make a stink about additional monitors because I work better with three monitors. Now, everyone on my floor has three monitors, the two standard 21" screens and one 34" curved 3440 x 1440 monitor.
But I couldn't make a honest compelling argument to use Linux. Sure I prefer it but tools like Cygwin, MobaXterm, and now WSL allow me to do what I need. Integrating it with Active Directory and setting up Remote Desktop so that I can remote in to their Windows Servers would take time for me and that of the system administrators.
Comment removed based on user account deletion
I had always been dubious of this, not being a Microsoft fan, but if this works then that's a good thing.
In the past, the POSIX layer in the Windows kernel was highly dubious, it had basic file support but no access to anything in Win32 where all the UI, services, networking, etc. It seems like it existed solely to get past some DoD contract requirements (even mainframe MVS had some POSIX just get some deals).
If there's a standard tool to easily convert to/from Unix filenames then that is very useful also. That's a headache I have with Cygwin trying to work with some Windows command line tools, and it's always been painful with makefiles.
Been a bit nervous as we were acquired by a heavily Windows and Cloud oriented mothership, and someday I expect they'll want us to start using Windows so having a unix layer would come in handy. (engineers from the mothership have expressed envy that we don't have to use VSTS)
Some unix software has abandoned this, but quite a lot has not. Sure the big name distributions want to go to a single sysadmin tool that does everything, but the real benefit from unix comes from have the small tools that do one or few jobs that are easily composable. Ie, I'm using rep every day, and I use find at least once a week for something, and things like "find . -name \*.xyz | xargs grep mystring" aren't uncommon.
In engineering usually you get to choose the best tools. The whole company doesn't get these tools though as they're expensive, but if a company can't make an exception to allow someone a matlab license then it's not really an engineering company. Sure, they may want a site license if more than one user wants the same application, but that's no reason to force a department to be unproductive just because Microsoft isn't the publisher of application they want.
This isn't even about regulations. I've worked in highly regulated medical equipment companies, and we could download applications we needed. The problem usually starts with IT staff wanting to have cookie cutter Windows PCs everywhere with only basic office applications. Especially in large corporations where all the IT staff are cookie cutter certificate holders.
The main issue :
Emulation - of course, by definition - would require to run a full blown virtual machine emulating a whole freaking Android smartphone.
On top of the Windows 10 Mobile running smartphone.
Not many smartphones have enough resource to play at this games. And again the whole point of the effort is to make Android apps available on Windows 10 Mobile to as many users as possible to make it attractive by tapping into the dominant ecosystem. It would be counter productive to advertise "Windows 10 Mobile can now run your favourite Android App - (*only on select few high range phones)".
That's why in Linux land "Android-in-a-box" efforts are shifting toward "andbox" (lightweight containers, no full blown emulation).
Past effort have also been running straight a top of the main Linux kernel (e.g.: Alien-dalvik by Myriad, runs the "I can't believe it's not JAva(tm)" JIT simply as another user-space program in a chroot).
ChromeOS too is relying on containerization.
But of course that's much easier when you main kernel is having nearly identical API that your target (save for a few android specific things like its peculiar IPC, that you can compile and load as modules anyway). Microsoft are having a much upstream battle. That they are apparently losing (Hey, how does it feel to have a taste of your own medicine ? Ask the wine guys what they are thinking!...)
Also some minor other issues:
Also another thing is that you'd have to install the Android VM image.
Which might bring some licensing issues (Google services are licensed vs. the free AOSP misses pieces that some application might require)
And make Android app convenient to install (in a VM setting, there should be some android app store available inside the VM - e.g.: aptoid is a popular one).
Having android as just another userspace layer makes it easier to install apps "from the outside", e.g.: install from the Microsoft app store (real-world example: see how aptoid apps are integrated into the main Jolla Sailfish store).
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]