POSIX layer

Lucractius · Post by **Lucractius** » Thu Apr 06, 2006 5:51 am

http://www.reactos.org/forum/viewtopic. ... highlight=

here, discussion about cygwin mentions the significant improvement that a subsystem has over any translation layer, the question that i have is simply will it be done? will ReactOS implement a Real, (cygwin just isnt complete enough, and Interix is good but still doesnt complete the full enviroment) POSIX layer that works properly in a seamless native fashion. This would be a huge benifit to Server useage, enabling software that simply requires POSIX compliance, to work nearly immediatly (after a recompile to the right binary form obviously)

Also with a full POSIX layer, key parts of the existing Linux/BSD open source systems can be implememnted in a productive and benificial way.

I doubt any work has been done on any sort of RDP functionality in ROS, and realy its probably better to avoid it, MS have always been arcane in their treatment of thin clients and remote application serving and im sure theyd fight brutaly anything that started making it easy for admins to just do it. As an alternative to RDP, with a translation layer that lets an X11 layer see the WGF calls made by normal programs, NX could be implemented in some fashion that would make this useful to admins.

that example is probably a long way from easy to do, but possible, and im sure people can think of better ways a full Posix layer could be done, but its just an example. POSIX is one of the underpinnings of the majority of open souce applications and proper compatibility is needed. X11 is the other big one, and unfortunatly i doubt that anyone here cares too much for any work on integrating an X11 internal loopback & integrated X server calling to either the graphics hardware or the native ROS GUI layer. *shrugs*

But im still interested in any thoughts from people or the devs on a full POSIX layer, and even any discussion on how it relates to X11 compatibity or anything else that could benifit ROS in some way. ROS can do what windows does, but can it do what windows cant?

Matthias · Post by **Matthias** » Thu Apr 06, 2006 1:38 pm

The problem with a POSIX layer is that hardly anyone will make use of it. Firstly, because there are good, free POSIX implementations available (GNU/Linux, OpenSolaris, Free/Net/Open/Dragonfly-BSD), secondly because many programs run well with cygwin already. Finally, most remaining UNIX/Linux programs aren't geared towars Windows users anyway, or they have good Win32 equivalents. Although I think that a POSIX Layer is a good idea, i also think that there are *far* more important Tasks to do - proper file systems, device driver compatibility, stability etc. pp. ReactOS shouldn't try to make the second step before the first, IMO, so let's focus on making windows applications run for now

linuxgx · Post by **linuxgx** » Thu Apr 06, 2006 7:40 pm

I agree, we arnt at a step where we should start distracting ourselfs with small projects, that will require alot of work. Stay on focus and get this thing done. This Alpha Should be a Beta by now.

Post by **Ged** » Thu Apr 06, 2006 7:44 pm

kjk started one but scrapped it.

colinux will probably be used for any *nix needs

vperez69 · Post by **vperez69** » Fri Apr 07, 2006 12:51 am

why posix, why X11? This is a Win32 clone, not another linux clone.
There is plenty of work to be done just to get the basic windows compatibility working. Sure ... for Reactos 4.0 or so, it might be nice, but not for 0.3

linuxgx · Post by **linuxgx** » Fri Apr 07, 2006 4:11 am

Although think about it, how nice would it be to combine a Win32 OS, with Linux a Kernel, Linux apps running under a windows like os would be nice.

And Sure would get us support form the linux comunity

cmoibenlepro · Post by **cmoibenlepro** » Fri Apr 07, 2006 4:26 am

why re-invent the wheel, since there is the existing colinux project that could do it?
I don't think that a posix layer would be useful.

Post by **ThePhysicist** » Fri Apr 07, 2006 5:05 am

A possible future scenario:
1 year: ROS will become somehow usable
2 years: ROS has a huge community, a lot developers and is in many things as good as MS Windows
4 years: ROS is a great competitor to MS Windows
7 years: ROS (and maybe some ROS derivates) and Linux share 70% of the OS market. ROS already runs most Unix/Linux apps and Linux/Wine runs most W32/64 (nobody uses the word Windows anymore) Apps.
10 years: ROS and Linux both run either ROS or Linux/Unix or OS2 or DOS or whatever (Windows? Oh yes, Windows can run ROS Apps, too). And the devs start to standardize the API. MS doesn't define any standards anymore.
15 years: A very sophisticated, advanced and standardized API will be present that allows future programs to run on whatever OS. Future enhancements will be something the OSS community will discuss and enforce.
Finally the great companys have to accommodate to what millions of users and devs all around the world decide together in a global democracy.
OK, just a stupid idea. It's pretty late (or early in the morning) here in Germany and I have already had some beer

linuxgx · Post by **linuxgx** » Fri Apr 07, 2006 5:55 am

ThePhysicist wrote:A possible future scenario:
1 year: ROS will become somehow usable
2 years: ROS has a huge community, a lot developers and is in many things as good as MS Windows
4 years: ROS is a great competitor to MS Windows
7 years: ROS (and maybe some ROS derivates) and Linux share 70% of the OS market. ROS already runs most Unix/Linux apps and Linux/Wine runs most W32/64 (nobody uses the word Windows anymore) Apps.
10 years: ROS and Linux both run either ROS or Linux/Unix or OS2 or DOS or whatever (Windows? Oh yes, Windows can run ROS Apps, too). And the devs start to standardize the API. MS doesn't define any standards anymore.
15 years: A very sophisticated, advanced and standardized API will be present that allows future programs to run on whatever OS. Future enhancements will be something the OSS community will discuss and enforce.
Finally the great companys have to accommodate to what millions of users and devs all around the world decide together in a global democracy.
OK, just a stupid idea. It's pretty late (or early in the morning) here in Germany and I have already had some beer

Dude That is some optomistic views there!!!!!!

Linux Apps being the standerd.

The reality it that if we add linux support, we would become a windows killer. Creating a bridge between windows and linux.

But i assure you MS wont go softly. But Rember IBM was once a standerd. And IBM no longer has a pc division

Matthias · Post by **Matthias** » Fri Apr 07, 2006 10:39 am

cmoibenlepro wrote:why re-invent the wheel, since there is the existing colinux project that could do it?
I don't think that a posix layer would be useful.

The WinNT Kernel is designed so that you can easily add subsystems, in fact there used to be a primitive POSIX Subsystem in Windows (iirc up to Win2k). It'd be a pity not to use these capabilities. Also, a POSIX Subsystem would be much faster than coLinux. However, for the reasons i explained before, i believe that the time for a POSIX Layer hasn't yet come.

vperez69 · Post by **vperez69** » Fri Apr 07, 2006 2:48 pm

i got nothing against linux-compatibility: IN TIME. The focus should still be win32 for now.

logan_V8 · Post by **logan_V8** » Fri Apr 07, 2006 6:45 pm

Hi!,
i don't know very much related to the POSIX subsystem but i found a project time ago that i think that could be useful. That project is a POSIX API implemented in a driver. The source can be found at http://sourceforge.net/projects/winntposix and a description in the Project homepage located at http://winntposix.sourceforge.net/.

Althrough i agree with the people that said that its too early to think in a POSIX layer now. There are lots of more things to do.

regards,
Logan_V8

linuxgx · Post by **linuxgx** » Sat Apr 08, 2006 1:55 am

But if its open and predeveloped, why not impliment it.

Blazkowicz · Post by **Blazkowicz** » Sat Apr 08, 2006 4:26 am

when reactOS becomes usable, maybe people who did bother to create cygwin and colinux will be interested in helping with a POSIX subsystem?

(sorry, don't get offended if I'm often starting my post with "when reactOS becomes usable" )

ReactOS could become a bit like what's said on Mac OSX : an easy OS, but with the unix/GNU stuff here (esp. dev tools), X11 used when needed, well integrated.

Well, I tried cygwin, painful and long to install (all those packages to choose..), requiring lots of bandwith (and I think I never managed to install from local cache, or add packages without reinstalling everything :p). I could compile a hello world, and could ssh and run an X session over internet. but I never managed to do my little poor student stuff with C++, opengl and glut (installed, sort of), which was done on linux at uni.
I then put mingw, kind of added glut libraries/headers and tried to compile. then I put "msys", then microsoft Interix. and I also had got Java SDK for something else.
Big mess

, I felt windows was cluttered a bit, and I never compiled my program.

(that was monthes, or even a year ago)

I'm not an unix guru, though I can use the shell, compile stuff, edit makefiles and run configure scripts, tweak the options, run it again, shit it doesn't work, get and compile this library first, then this other one, configure again, edit the makefile.. (used solaris a lot at uni on terminals, with as a desktop environment, only the Motif WM and xterms, and wonderful apps like netscape 4 :p. in year 2004-2005)

I managed to do stuff in that solaris environment.

well, all of that to say that as a "skilled novice", quite geeky and not afraid to spend time guessing, doing boring stuff and trying to make sense of cygwin and mingw, I was discouraged a bit. (and aren't cygwin and mingw about the only free C/C++ developing tools for windows? why the pseudo unix stuff in mingw?)

maybe I'm polluting the thread a bit. (but I have some free time now, I should try to really get into developing.. but doesn't know exactly how)

linuxgx · Post by **linuxgx** » Sat Apr 08, 2006 5:52 am

You want a coding project ok.

Get us an upto-date linux kernel that will run under the Ros eviroment, and alow the use of linux apps under Explore, and have it run seamlessly with out emulation.

There is your project.

Should only take you a few years, but it is what we all realy want any way.

We all want an OS that runs Win32, and Kernel seemlessly with out any tricks or extra work.

Sort of the whole point of the wine project wasnt it?

POSIX layer

POSIX layer

don't have a beef against linux, but ....

exctly...

Who is online