Will this be Open Source?

[sirinath]

Hopefully Bash On Windows will be opensource

[bitcrazed]

We have no firm plans as yet, but we're not averse to open-sourcing some of this tech.

[sirinath]

Hopefully you will be slanted towards open sourcing it with well documented internals to get other subsystems like Darwin, BSD, SmartOS or other systems running on Windows.

[BSalita]

There's lots of issues with terminal emulation in the first beta. This is particularly an area where open source availability would quicken the pace of bug fixing and compatibility.

[poizan42]

Can I vote for having lxcore.sys open? Just think of the possibilities in having a starting point for the community to implement other subsystems.

(Am I wrong in assuming that implementing a subsystem requires tight coupling with kernel internals that you don't want to have to keep that way for backwards compatibility in the future?)

[russalex]

Great conversation around open sourcing and we hear the requests. Might I suggest voting on this on our uservoice page?

[RaeesBhatti]

There is actually a very similar open-source project called flinux. Whic is based on the same idea of translating Linux Syscalls to Windows Syscalls. Mabye Windows team can get some inspiration from them 😄

[RaeesBhatti]

I've just created a UserVoice feedback for this idea. Anyone interested can vote for this there.

[ytrezq]

@RaeesBhatti : it convert Linux calls directly to ɴᴛ calls. This is different since the windows subsystem _(win32k.sys)_is completely bypassed in favor oflxcore.sys (yes, windows is subsystem itself).

[ssokolow]

@ytrezq: Yeah. A lot of people don't know that Windows NT was designed to be modular that way. It was originally designed to have three subsystems: Win32, POSIX, and OS/2.

Heck, drive letters are an artifact of the Win32 subsystem and NTFS can operate in a POSIX mode where the only characters disallowed in file names are NUL and /.

If it weren't for the whole "can't patch it or make derivative works and you have to pay license fees" part, we'd probably all be running some hybrid of Windows NT (for its modular kernel design and excellent UNC support) and Linux or BSD (for the elegant "hardlinks and open file handles are treated as equals in the reference counting" approach to allowing renaming and deleting of in-use files).

[ytrezq]

@ssokolow : as I’m currently looking for a way to compile a native ᴅʟʟ for the ᴘᴘᴄ architecture without original hardware, I perfectly know that :-)

[ssokolow]

I suspected as much. I probably should have been a bit more clear that my comment was for other people's benefit.

[slavanap]

@bitcrazed What about disclosing source code of /init binary? It has to use some of Ubuntu/Linux sources because it depends on libc.so.6. And it is non-Ubuntu binary for sure because it contains lxssmanager string inside.

[ssokolow]

@slavanap
All dynamically-linked Linux binaries depend on some version of libc.so and glibc's license specifically contains an exception to allow that for closed-source applications.

[russalex]

Long story short, we are still working on /init. What is out there was required to get WSL out the door, but there are some pretty low level changes we need to make for future features. We're just not in a position to discuss open sourcing it yet.

We have had the larger open source conversation a number of times in house but the same story goes there. We have some new features that still need to be made before we can truly discuss feasibility and what parts of the project we can / make sense to open source. None of us are opposed to doing it, we just need to do it when it makes sense and do it right.

[Kreijstal]

narrator: it was not open source