Ubuntu on Windows [LWN.net]

Ubuntu on Windows

Posted Mar 30, 2016 17:50 UTC (Wed) by kokada (guest, #92849) [Link] (1 responses)

Windows already have a next-gen filesystem, ReFS (https://en.wikipedia.org/wiki/ReFS). However, you can't use it as a boot disk yet, at least in desktop versions of Windows.

Ubuntu on Windows

Posted Mar 30, 2016 17:57 UTC (Wed) by rsidd (subscriber, #2582) [Link]

Interesting, but sounds like it's not production ready. Nor is Ubuntu-on-Windows, of course.

Ubuntu on Windows

Posted Mar 30, 2016 18:42 UTC (Wed) by einstein (subscriber, #2052) [Link] (20 responses)

This doesn't appear to be, ultimately, anything more than a re-hash of the old "MKS utilities for windows", the microsoft "unix services" or the "posix" layer in nt - in other words, not any sort of real, usable Linux environment. To get an idea of how much spin there is here, the same article repeats the marketing hype about linux containers running natively on windows - which is, of course, nonsense, as lxd is a function of the Linux kernel. The whole "lxd on windows" mirage really only amounts to running lxd on Linux, inside a VM on a windows pc.

So take everything here with a grain of salt. Technical details, not marketing speak, will tell the tale.

Ubuntu on Windows

Posted Mar 30, 2016 18:50 UTC (Wed) by rsidd (subscriber, #2582) [Link] (19 responses)

I'm not sure what article you read but the one linked here makes explicit that it's not a container on Windows. Nor is it a recompilation of Linux binaries. It is the unmodified Ubuntu binaries running on Windows, via an interface in the Windows kernel that maps linux syscalls to windows syscalls and with access to the unmodified Ubuntu userland (libraries etc). The closest analogue I can think of is the linuxulator on FreeBSD. But FreeBSD is much more like Linux than Windows is, so this is much more impressive. The most impressive part being that it's largely done in-house at Microsoft.

Ubuntu on Windows

Posted Mar 30, 2016 19:11 UTC (Wed) by ssmith32 (subscriber, #72404) [Link] (6 responses)

So the inverse of wine? (But in kernel space)

Trying to think of a catchy acronym...

wile ?

(Windows is not a linux emulator?)

I used my windows wiles to run wily werewolf on windows?

Ubuntu on Windows

Posted Mar 31, 2016 0:00 UTC (Thu) by asaz989 (guest, #67798) [Link] (4 responses)

Sounds like it.

And if I may digress, the fact that it's possible to write Wine is user space is a credit to the Windows ecosystem. They clearly thought long and hard about backwards compatibility before putting together the win32 API, and made userspace programs go through a shared library rather than directly making syscalls. The intention was to make it easier for Microsoft itself to swap out the kernel (which they have done), but the ability for third parties to swap in *nix has been a happy side effect.

Ubuntu on Windows

Posted Mar 31, 2016 5:20 UTC (Thu) by roc (subscriber, #30627) [Link] (3 responses)

WINE would have been just as easy to write, or easier, if Windows offered stable syscall APIs. Linux already supports a variety of syscall personalities and WINE would have implemented another one.

Not having a documented, stable syscall interface means that many tools (e.g. strace, rr) are difficult or impossible to support on Windows.

Ubuntu on Windows

Posted Mar 31, 2016 5:27 UTC (Thu) by Cyberax (✭ supporter ✭, #52523) [Link] (2 responses)

> WINE would have been just as easy to write, or easier, if Windows offered stable syscall APIs.
It does. It's called "native API" and is fairly compact and is extremely stable.

There were projects to add native Windows "syscalls" to Linux: https://en.wikipedia.org/wiki/Longene - not many people are interested in it, though.

> Not having a documented, stable syscall interface means that many tools (e.g. strace, rr) are difficult or impossible to support on Windows.
Way back before Docker, brave people from Parallels wrote container support for Windows: https://virtuozzo.com/support/pcw/ - without getting access to Windows source code, btw.

They did it by providing a thin layer above the native API and running multiple userspaces on top of it.

Windows kernel is pretty neat, although it's showing its age.

Ubuntu on Windows

Posted Mar 31, 2016 18:40 UTC (Thu) by xilun (guest, #50638) [Link] (1 responses)

I don't think the NT API is particularly stable. At least not at binary level http://j00ru.vexillium.org/ntapi_64/ and probably also not at source level.

Ubuntu on Windows

Posted Mar 31, 2016 19:26 UTC (Thu) by Cyberax (✭ supporter ✭, #52523) [Link]

It is stable. Windows still supports several implementations of the USB stack, just because some drivers depend on a 17-year-old first stab at USB support.

Things do break from time to time, but MS actually cares about old drivers and system software.

It's much easier in Windows, as it's built on top of a "message passing". Almost all operations involve sending a uniformly formatted "message" (IRP) which can pass through multiple layers that can filter and/or modify it. So as long as the message format is preserved, keeping the compatibility is doable (not simple, but doable).

Linux kernel API is completely ad-hoc so it simply can't achieve the same stability level. The flip side is much better performance and maintainability.

Ubuntu on Windows

Posted Mar 31, 2016 21:03 UTC (Thu) by Seegras (guest, #20463) [Link]

Yes, this is totally Wile. E. Coyote.

Ubuntu on Windows

Posted Mar 30, 2016 19:30 UTC (Wed) by adler187 (guest, #80400) [Link]

We have a thing like this on IBM i, where the IBM i kernel (SLIC) provides (a subset of) AIX system calls for applications to use. We call it PASE (for Portable Application Solutions Environment, previously Personal Address Space Enablement). Many times I wish we'd have implemented the Linux syscall interface instead, since it would make porting OSS to IBM i much easier. I'm sure if we were doing it now instead of 15+ years ago it would be based on Linux.

https://www.ibm.com/support/knowledgecenter/ssw_ibm_i_72/...

Ubuntu on Windows

Posted Mar 30, 2016 19:41 UTC (Wed) by SEJeff (guest, #51588) [Link]

Or perhaps binfmt_misc on Linux?

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

https://www.kernel.org/doc/Documentation/binfmt_misc.txt

ie: PE :D

Ubuntu on Windows

Posted Mar 30, 2016 20:21 UTC (Wed) by einstein (subscriber, #2052) [Link] (9 responses)

> I'm not sure what article you read but the one linked here makes explicit that it's not a container on Windows.

Right, we know this is not linux containers on windows, which doesn't exist anyway. Pardon the reference, I had been looking at SJVN's article on the announcement, and in that article he also mentioned the lxd thing. Let's be clear, nobody has said this has anything to do with containers, only that the same level of breathless hype appears in the two claims.

So, to recap, it's not really any different from what has been trotted out several times before. It might have a temporary niche, as a curiosity, but beyond that, it seems merely to be a solution searching for a problem. Linux users sometimes have a need to run some legacy windows app - but do windows users have some need to run some Linux app that's not available on windows? Just not feeling it.

Ubuntu on Windows

Posted Mar 30, 2016 20:35 UTC (Wed) by Cyberax (✭ supporter ✭, #52523) [Link]

> Right, we know this is not linux containers on windows, which doesn't exist anyway.
You're saying?...

http://www.colinux.org/

Ubuntu on Windows

Posted Mar 30, 2016 21:01 UTC (Wed) by rahulsundaram (subscriber, #21946) [Link] (5 responses)

> but do windows users have some need to run some Linux app that's not available on windows? Just not feeling it.

Yes they do but you are probably not "feeling it" because you are not a Windows user.

Ubuntu on Windows

Posted Mar 30, 2016 22:29 UTC (Wed) by donbarry (guest, #10485) [Link] (4 responses)

WINE was developed to help bridge users from the proprietary Windows OS as they converted to free software.

This is intended exactly the reverse way, to leverage the free software community to promote a proprietary OS.

That Ubuntu would sponsor this is simply more revelation that their agenda is not friendly towards free software.

And those who see this in purely opportunistic terms: that's a consequence of the "open source" movement and its "pragmatism" -- which is really another way of saying its adaptation to corporate interests.

Ubuntu on Windows

Posted Mar 30, 2016 22:51 UTC (Wed) by Cyberax (✭ supporter ✭, #52523) [Link] (3 responses)

After such posts, I'm starting to think that _anything_ that is "bad for free software" in the end is actually a really great stuff for everybody (except Stallman).

Ubuntu on Windows

Posted Mar 31, 2016 4:48 UTC (Thu) by floriansnow (guest, #107824) [Link] (2 responses)

Yes, great for everybody who does not care about computing freedom. I think the technical achievement here is neat, but all it does is let me run more Free Software on a non-free OS. And that additional Free Software is probably not all that useful to people who aren't familiar with it already, so people who could also run a completely free system. So if this is more than just a tech-demo, then it is a way to lure people back to non-free software.

Ubuntu on Windows

Posted Mar 31, 2016 5:44 UTC (Thu) by jspaleta (subscriber, #50639) [Link]

Uhm...
There's a large collection of workers in enterprise environments who are using computing resources dictated by corporate it policy.

Any technology advancement which makes it easier for them to use free software to do their job, even if they are forced by corporate policy to run a mix of free and non-free software, is a win...

Its called progress.

Hats off to MS and Canonical for getting things this far. It's an interesting technical achievement, with a lot of potential.

The work is far from complete from the sounds of it, even for getting a usable strictly command-line linux developer's environment working. But there is obvious potential here to make it easier for people to pick up traditional linux user-space and make use of it even if they are forced to use windows (or choose to use windows) compared to having to run a VM or container.

I skeptical of the claim that Kirkland made that "most" of the ubuntu repository packages are going to work... but i give him the benefit of the doubt... its the first public demo of this..and if MS continue to put engineer resources into this then the claim might become reality. The only way to really know is to test packages and see what works and what doesn't.

My biggest concern is if MS takes the effort just far enough to serve their own purposes..but not far enough to serve actual use cases very well. I'd hate to see users asking Ubuntu and Debian packagers to service bugs specific to the windows 10 specific implementation of the linux ABI. At the end of the day, a lot of the magic here comes down to MS willing to spend the engineering resources to service the linux compatibility layer because they are the only ones who can do the work. The lack of open development, with a clear contribution friendly development model might keep this from being able to jump from interesting potential to real solution for real users. If people can't reasonably expect to be able to use a random ubuntu package successfully.. then developers, the stated target group, are just going to use a VM as a more reliable alternative, even if its a performance hit.

Ubuntu on Windows

Posted Mar 31, 2016 6:51 UTC (Thu) by ncm (guest, #165) [Link]

This is Microsoft's response to the decreasing fraction of competent developers willing to shackle their careers to the Windows ecosystem. Not long ago it was considered perfectly sensible to lock yourself into MS-only languages, libraries, and tools, but it's finally recognized as extremely short-sighted. Offering the ability to work like a normal developer helps to keep them closer to competitive.

Part of this is that developers who still are locked into MS are competing with each other in a shrinking market at cut-rate prices. So, there are still plenty of them, but they are the ones who work for less because they must.

Ubuntu on Windows

Posted Mar 31, 2016 0:19 UTC (Thu) by aryonoco (guest, #55563) [Link]

If you think the ability to run Linux binaries unmodified on Windows, for Windows to provide Linux API compatibility, is "the same as has been trotted before" and "only a niche", you're not living in the real world.

This is a colossal improvement over Cygwin and everything else that's been available for Windows. It's even an improvement over OS X, I don't have to do gymnastics to replace all the ancient BSD utilities that Apple ships with OS X with their GNU counterparts.

And those of us who teach/train people in Python/Ruby/Go don't have to fret anymore when someone shows up with a Windows laptop.

Ubuntu on Windows

Posted Mar 31, 2016 12:03 UTC (Thu) by nye (subscriber, #51576) [Link]

>Linux users sometimes have a need to run some legacy windows app - but do windows users have some need to run some Linux app that's not available on windows?

Fuck yes. If this is even half as good as it sounds, this is the best product announcement I've heard from *anybody* in *years*.

Ubuntu on Windows

Posted Mar 31, 2016 10:29 UTC (Thu) by khim (subscriber, #9252) [Link] (1 responses)

I still fail to understand why people never try to apply logic to announces. Microsoft have failed - and now it's time to change the strategy. By now it's obvious that Linux will not die any time soon. It's future on the servers it's secure (heck, HPC is almost entirely Linux nowadays) and it's winning on mobile.

These successes mean that Linux desktop could survive on coattails of server/mobile development. What Microsoft needs to do to protect itself? Well, the answer is obvious: make sure developers could develop all these mobile and HPC applications on Window. And it's also entirely obvious that development of desktop Linux applications should be as hard as possible. That's not hard to achieve, really: just make sue there's no GUI.

Ubuntu on Windows

Posted Mar 31, 2016 12:10 UTC (Thu) by nye (subscriber, #51576) [Link]

>And it's also entirely obvious that development of desktop Linux applications should be as hard as possible. That's not hard to achieve, really: just make sue there's no GUI.

Just use Cygwin's X server via TCP. Performance won't be amazing, but outside of games it's unlikely to matter.