I am not sure that replacing X with a proprietary Microsoft extension of an ITU standard is a great idea. If you think, RDP offers a good experience, feel free to use it. But my experience with open source RDP clients was always horrible. My experience with X over network is excellent. I use it everyday over LAN and WAN.
Also there is the issue of compatibility. Even if I wanted to (and I don't - because I don't see any technical reason) switching of X would break compatibilty with old and new applications. This would be a nightmare.
Everything consists of small things described by quantum mechanics. Still we usually see classical behaviour on a macroscopic scale... It is really hard to make large systems show quantum effects, because you have to isolate it from the environment. So yes, it makes sense to be skeptical.
I wasn't really unhappy with Linux 10 years ago and a few years ago Ubuntu and other even started to polish it up to be really nice (remember project 100 paper cuts?). I don't know what happend then, but at some point it all went downhill.
They started to constantly break my user interface, by randomly changing things, removing features, or just creating new bugs. Now I am even scared to upgrade, because some programm I rely on might not work anymore (or just disappear because it was coded against some obsolete freedesktop standard), or might miss an important feature which was previously available.
I am perfectly fine with re-designing things. I am not happy about breaking stuff and especially breaking backwards compatibility. The kernel does not do it, libc does not (usually) do it, the C compiler does not do it, network protocols do not do this. Why do the freedesktop/GUI/graphics people do it all the time?
How is it a complaint when I say that it works for me? The precise technical explanation is that direct rendering does not work over the network. For 99% of all applications this is irrelevant.
X will certainly not be gone. Maybe there will be a schism in the Linux community. This might actually not be bad thing. The freedesktop crowd who break my desktop (or some applications) basically every year (I think this started in 2008 or so) can go on with their misguided attempts to redesign everything over and over again in different ways. And people who want a UNIX-like system with backwards compatibility, stability, configurability, and powerful features could have there own distribution.... This would be nice.
Yes, it is already doppler shifted because we are moving relative to it. That is why you see a large dipole in images of the background radiation (google: CMB dipole). It has been removed in many images to show the absolute values.
I am not sure what you mean by "relativistic lens" and "current". Something like a wrap drive we local areas of space carry stuff around? While there are such solutions of general relativity, that is not how our universe looks. It is described by a FLTW metric, which means that it is very simply and homogenous on a large scale. Everything expands slowly and it is possible to have global coordinate system based on the CMB (a rest frame with expanding coordinates).
Due to the Doppler effect, you see the frequency shift if you move relative to the microwave background, which would otherwise be (almost) the same blackbody radiation of temperature 2.725 K from all direction.
Hm. You are right. Debian/Ubuntu seem to activate the security extension by default and Fedora/Red Hat might use XACE and selinux (I don't know). But it is a bit disappointing to learn how much this is neglected upstream.
X clients tunnelled over X are untrusted X clients and do not have access to everything by default. See X security extension and the -Y option of ssh.
True, but that would apply only to the clients going through the tunnel. Well, assuming it works (on my distro ssh X11 forwarding without -Y doesn't work at all).
I don't understand. Local programs have access to everything anyway (for this user) - even without X. So why is this a problem? (Well it *is* a problem - but unrelated to X)
What do you mean that X forwarding is not working? It is usually deactivated by default and you have to turn it on with -X or in a config file. If you use '-Y' you turn off security. You have no reason to complain, if you turned security off yourself.
A bit offtopic, but what I find sad is that there is no GUI following the Unix philosophy and no GUIs which integrate well with the command line. I mean things like GUIs for combining software components which to do useful stuff (cli completion is nice, but why is there no scrollbox to select from? Why no GUI with checkboxes which I can open with right-click when I forgot an obscure option?), good graphical shells which combine a file manager with a CLI shell (there was norton commander), or terminals with graphical features. I remember seeing some interesting projects, but nothing seems to have developed to a state to be really useful. I maybe wrong, I you know such things, let me know.
This is where I would have hoped to see innovation. Instead we get dumbed down GUIs with fancy annimations.
I use CUDA for HPC. I also use X over the network everyday. Please don't educate me about bandwidth. Your point seems to be that sending OpenGL over the network is not a good idea and that it is better to send the rendering result. This is certainly true for many 3D applications such as games. Guess what. Both can be done with the X protocol. No need to break compatibility.
Left 4 Dead 2 on Windows 7 with Direct3D drivers, we get 270.6 FPS as OpenGL implementation on Windows. Left 4 Dead 2 is now running at 303.4 FPS Left 4 Dead 2 is running at 315 FPS on Linux.
Network transparency means that applications can render graphics to a local terminal or over a network with zero changes in code path and zero need to know anything about the underlying rendering model. Therefore, any remotely modern version of X.org is by definition not network transparent since every since modern local rendering technique such as DRI and compositing is completely incompatible with the fallback socket-based path that is used for networking remote X programs.
No, you are confusing various things. DRI and compositing are different concepts. XRender is perfectly network transparent. OpenGL can also be used over the network via GLX. DRI allows direct access to graphics hardware and is not network transparent, but this is an optimization for local use.
Consequently, modern X is not network transparent and people who can't understand that just because it is still possible to send X pixmaps over a network socket in a kludgy manner does not mean X is "transparent" should maybe do some research on how X actually works instead of hurling insults.
In a world where internet is finally everywhere, and X (if embraced) would allow to move a game from smartphone to TV when coming home. Or a text window from notebook to tablet for discussion around a table, or.
X doesn't allow that. X has frequent round trips. Which means over a WAN it is a very high latency protocol. The fixes for that have been to semi-shatter "network transparency".
X allows that. You have to hack in at the client side, because no toolkit bothers to implement moving windows from screen to the others, but I did it for some of my applications. The same is true for roundtrips. The protocol is asynchronous, but toolkits use it in a synchronous way.
X can only work well on a mid latency setup, a LAN, the environment it evolved to handle. So no, X wouldn't allow you to do those things.
Worse it doesn't even allow you to do those things locally, Because one of the areas X also sucks at is complex graphics like TV and graphics and games.
What is wrong with OpenGL?
And that's because you can't share video buffers with applications buffers because it is client / server.
But by that standard X is not network transparent at its core. Anyone using X on a desktop (GTK, QT apps) will be using the SHM, Randr, XVideo, XInput2 etc extensions which means that after start-up a typical application is not using much of the core protocol.
That X11 had been designed to be extensible is one of its strength. That there is special support to make all overhead go away for local client does not seem to be something bad to me. As you correctly point out: this is transparent to most applications:
When you try to use one of these applications over a network the toolkits fall back to using the core X protocol just to blast lots of large pixmaps down the wire, because the core X11 protocol graphics is so outdated it just can't handle alpha blending, properly hinted fonts and so on.
You seem to be confused. X has been extended for example with XRender which supports alpha blending. But this has nothing to do with network transparency, XRender is perfectly network transparent.
How is that different from RDP, except that RDP was designed and heavily optimized for that use case (and yes, RDP can send individual windows, not just the whole desktop).
The important part is to realize that showing graphics remotely is only a minor part of a network transparent display protocol.
That's quiet apart from the fact that X11 was designed to use local, narrow-band (by modern standards) networks to render simple (by modern standards) graphics.
First of all, X was designed to be flexible. This is why it is still great, because could and has been adapted to modern requirements.
It is very 'chatty' in terms of round-trips it requires. Using it over high-latency links is not pleasant at all, even if the bandwidth of the link is orders of magnitude greater than was available in 1990. It's just not optimized well for current use.
Again, this is a misunderstanding. Many X applications do indeed suck over high-latency links. But this is not an inherent problem of the X protocol, but because nobody toolkits are doing things in a stupid way and nobody fixed it.
Wayland offers a simpler, saner design that doesn't have duplication of effort between applications, compositors, window managers and the core.
You seem to think that this monolithic design is a good idea. I disagree. Wayland was really simple in the beginning, now as it starts to become useful it is getting more complex duplicating a lot of stuff which has already been solved in X a long time ago. At the time it will be able to replace X, it will not be much simpler anymore, but a lot less powerful. Maybe some old stuff can get eliminated, but I doubt that this is worth it: I do not see how any of this 'old stuff' gets into the way of doing 'new stuff' on top of the X protocol.
It eliminates decades of subsystems that have to be supported but are duplicative or totally useless (such as the cross-platform ELF library loading, the crappy bitmap font rendering and the 'stippled line' support).
I don't see how bitmap font rendering or 'stuppled line' support can be a great maintance burden. For ELF library loading this might be different, but this does not seem to be a core problem of X.
It is built around the assumption that there may well be graphics accelerators of varying capabilities that may be used, rather than insisting that using such support is an 'extension'.
This doesn't make sense.
Most significantly, many of the core X.org developers are working on it, because they don't see that X can be taken much further.
This seems to be a myth. Some people who now work on Wayland have commited to X would be more accurate.
Decades of backwards and forwards compatibility. The real question is: What does Wayland offer?
And yes, I think it is stupid to design in 2013 a display protocol which is not network transparent at its core. In a world where internet is finally everywhere, and X (if embraced) would allow to move a game from smartphone to TV when coming home. Or a text window from notebook to tablet for discussion around a table, or.... And no, RDP does not cut it.
1. X is stable. That its old rendering model is not used anymore by modern applications does not change this a bit. With newer extensions, you can do compositing and direct rendering just fine. I don't get your point.
2. People always used toolkits. And I don't see any technical reason why there should be a performance gain with Wayland.
How would abandoning X improve the situation for distros? If they have to maintain Xwayland they now have an even more complicated system (Wayland + X). If not, they break compatibility (which would be really really stupid).
The real reason distros are forked is because Linux userspace is an entangled mess of packages patched into shape to work with ever changing APIs. I am not speaking about kernel-level APIs here, those are rock solid. If APIs would be stable, packages would not have so many versioned dependencies and forking would not be necessary. I am already scared of the day when the idiots make their threat true to replace X with Wayland, breaking one of the remaining stable APIs with decades of backwards compatibility for no good reason.
I am not sure what you are talking about. It was a government-operated mine (after it was a commercial salt mine before) which served as a test-bed for Gorleben. The thing is: People just appoached this problem in the 60s/70s with the same naivety you can still observe here on slashdot: This nuclear waste is not a problem. Such a small amount, we can just bury it underground somewhere. It is not problem at all!
Slashdot Mirror
Easier than 'ssh -X'. I dont think so.
I am not sure that replacing X with a proprietary Microsoft extension of an ITU standard is a great idea. If you think, RDP offers a good experience, feel free to use it. But my experience with open source RDP clients was always horrible. My experience with X over network is excellent. I use it everyday over LAN and WAN.
Also there is the issue of compatibility. Even if I wanted to (and I don't - because I don't see any technical reason) switching of X would break compatibilty with old and new applications. This would be a nightmare.
Everything consists of small things described by quantum mechanics. Still we usually see classical behaviour on a macroscopic scale... It is really hard to make large systems show quantum effects, because you have to isolate it from the environment. So yes, it makes sense to be skeptical.
I wasn't really unhappy with Linux 10 years ago and a few years ago Ubuntu and other even started to polish it up to be really nice (remember project 100 paper cuts?). I don't know what happend then, but at some point it all went downhill.
They started to constantly break my user interface, by randomly changing things, removing features, or just creating new bugs. Now I am even scared to upgrade, because some programm I rely on might not work anymore (or just disappear because it was coded against some obsolete freedesktop standard), or might miss an important feature which was previously available.
I am perfectly fine with re-designing things. I am not happy about breaking stuff and especially breaking backwards compatibility. The kernel does not do it, libc does not (usually) do it, the C compiler does not do it, network protocols do not do this. Why do the freedesktop/GUI/graphics people do it all the time?
How is it a complaint when I say that it works for me? The precise technical explanation is that direct rendering does not work over the network. For 99% of all applications this is irrelevant.
X will certainly not be gone. Maybe there will be a schism in the Linux community. This might actually not be bad thing. The freedesktop crowd who break my desktop (or some applications) basically every year (I think this started in 2008 or so) can go on with their misguided attempts to redesign everything over and over again in different ways. And people who want a UNIX-like system with backwards compatibility, stability, configurability, and powerful features could have there own distribution.... This would be nice.
Sigh, this gets old. I use it everyday and it works! So yes, it is network transparent.
Yes, it is already doppler shifted because we are moving relative to it. That is why you see a large dipole in images of the background radiation (google: CMB dipole). It has been removed in many images to show the absolute values.
I am not sure what you mean by "relativistic lens" and "current". Something like a wrap drive we local areas of space carry stuff around? While there are such solutions of general relativity, that is not how our universe looks. It is described by a FLTW metric, which means that it is very simply and homogenous on a large scale. Everything expands slowly and it is possible to have global coordinate system based on the CMB (a rest frame with expanding coordinates).
Due to the Doppler effect, you see the frequency shift if you move relative to the microwave background, which would otherwise be (almost) the same blackbody radiation of temperature 2.725 K from all direction.
If one considers the rest frame of the microwave background as the rest frame of the universe, then yes, one can answer these questions.
http://en.wikipedia.org/wiki/Cosmic_microwave_background
Hm. You are right. Debian/Ubuntu seem to activate the security extension by default and Fedora/Red Hat might use XACE and selinux (I don't know). But it is a bit disappointing to learn how much this is neglected upstream.
X clients tunnelled over X are untrusted X clients and do not have access to everything by default. See X security extension and the -Y option of ssh.
True, but that would apply only to the clients going through the tunnel. Well, assuming it works (on my distro ssh X11 forwarding without -Y doesn't work at all).
I don't understand. Local programs have access to everything anyway (for this user) - even without X. So why is this a problem? (Well it *is* a problem - but unrelated to X)
What do you mean that X forwarding is not working? It is usually deactivated by default and you have to turn it on with -X or in a config file. If you use '-Y' you turn off security. You have no reason to complain, if you turned security off yourself.
A bit offtopic, but what I find sad is that there is no GUI following the Unix philosophy and no GUIs which integrate well with the command line. I mean things like GUIs for combining software components which to do useful stuff (cli completion is nice, but why is there no scrollbox to select from? Why no GUI with checkboxes which I can open with right-click when I forgot an obscure option?), good graphical shells which combine a file manager with a CLI shell (there was norton commander), or terminals with graphical features. I remember seeing some interesting projects, but nothing seems to have developed to a state to be really useful. I maybe wrong, I you know such things, let me know.
This is where I would have hoped to see innovation. Instead we get dumbed down GUIs with fancy annimations.
X clients tunnelled over X are untrusted X clients and do not have access to everything by default. See X security extension and the -Y option of ssh.
Right. The X doesn't do much. It is also not slow for what it does. So why replace it?
I use CUDA for HPC. I also use X over the network everyday. Please don't educate me about bandwidth. Your point seems to be that sending OpenGL over the network is not a good idea and that it is better to send the rendering result. This is certainly true for many 3D applications such as games. Guess what. Both can be done with the X protocol. No need to break compatibility.
Yeah right, just look how abysmal it performs:
http://blogs.valvesoftware.com/linux/faster-zombies/
Left 4 Dead 2 on Windows 7 with Direct3D drivers, we get 270.6 FPS as
OpenGL implementation on Windows. Left 4 Dead 2 is now running at 303.4 FPS
Left 4 Dead 2 is running at 315 FPS on Linux.
Network transparency means that applications can render graphics to a local terminal or over a network with zero changes in code path and zero need to know anything about the underlying rendering model. Therefore, any remotely modern version of X.org is by definition not network transparent since every since modern local rendering technique such as DRI and compositing is completely incompatible with the fallback socket-based path that is used for networking
remote X programs.
No, you are confusing various things. DRI and compositing are different concepts. XRender is perfectly network transparent. OpenGL can also be used over the network via GLX. DRI allows direct access to graphics hardware and is not network transparent, but this is an optimization for local use.
Consequently, modern X is not network transparent and people who can't understand that just because it is still possible to send X pixmaps over a network socket in a kludgy manner does not mean X is "transparent" should maybe do some research on how X actually works instead of hurling insults.
glass house, stones, etc.
In a world where internet is finally everywhere, and X (if embraced) would allow to move a game from smartphone to TV when coming home. Or a text window from notebook to tablet for discussion around a table, or .
X doesn't allow that. X has frequent round trips. Which means over a WAN it is a very high latency protocol. The fixes for that have been to semi-shatter "network transparency".
X allows that. You have to hack in at the client side, because no toolkit bothers to implement moving windows from screen to the others, but I did it for some of my applications. The same is true for roundtrips. The protocol is asynchronous, but toolkits use it in a synchronous way.
X can only work well on a mid latency setup, a LAN, the environment it evolved to handle. So no, X wouldn't allow you to do those things.
Worse it doesn't even allow you to do those things locally, Because one of the areas X also sucks at is complex graphics like TV and graphics and games.
What is wrong with OpenGL?
And that's because you can't share video buffers with applications buffers because it is client / server.
You mean DRI does not exist?
But by that standard X is not network transparent at its core. Anyone using X on a desktop (GTK, QT apps) will be using the SHM, Randr, XVideo, XInput2 etc extensions which means that after start-up a typical application is not using much of the core protocol.
That X11 had been designed to be extensible is one of its strength. That there is special support to make all overhead go away for local client does not seem to be something bad to me. As you correctly point out: this is transparent to most applications:
When you try to use one of these applications over a network the toolkits fall back to using the core X protocol just to blast lots of large pixmaps down the wire, because the core X11 protocol graphics is so outdated it just can't handle alpha blending, properly hinted fonts and so on.
You seem to be confused. X has been extended for example with XRender which supports alpha blending. But this has nothing to do with network transparency, XRender is perfectly network transparent.
How is that different from RDP, except that RDP was designed and heavily optimized for that use case (and yes, RDP can send individual windows, not just the whole desktop).
The important part is to realize that showing graphics remotely is only a minor part of a network transparent display protocol.
That's quiet apart from the fact that X11 was designed to use local, narrow-band (by modern standards) networks to render simple (by modern standards) graphics.
First of all, X was designed to be flexible. This is why it is still great, because could and has been adapted to modern requirements.
It is very 'chatty' in terms of round-trips it requires. Using it over high-latency links is not pleasant at all, even if the bandwidth of the link is orders of magnitude greater than was available in 1990. It's just not optimized well for current use.
Again, this is a misunderstanding. Many X applications do indeed suck over high-latency links. But this is not an inherent problem of the X protocol, but because nobody toolkits are doing things in a stupid way and nobody fixed it.
Wayland offers a simpler, saner design that doesn't have duplication of effort between applications, compositors, window managers and the core.
You seem to think that this monolithic design is a good idea. I disagree. Wayland was really simple in the beginning, now as it starts to become useful it is getting more complex duplicating a lot of stuff which has already been solved in X a long time ago. At the time it will be able to replace X, it will not be much simpler anymore, but a lot less powerful. Maybe some old stuff can get eliminated, but I doubt that this is worth it: I do not see how any of this 'old stuff' gets into the way of doing 'new stuff' on top of the X protocol.
It eliminates decades of subsystems that have to be supported but are duplicative or totally useless (such as the cross-platform ELF library loading, the crappy bitmap font rendering and the 'stippled line' support).
I don't see how bitmap font rendering or 'stuppled line' support can be a great maintance burden. For ELF library loading this might be different, but this does not seem to be a core problem of X.
It is built around the assumption that there may well be graphics accelerators of varying capabilities that may be used, rather than insisting that using such support is an 'extension'.
This doesn't make sense.
Most significantly, many of the core X.org developers are working on it, because they don't see that X can be taken much further.
This seems to be a myth. Some people who now work on Wayland have commited to X would be more accurate.
Decades of backwards and forwards compatibility. The real question is: What does Wayland offer?
And yes, I think it is stupid to design in 2013 a display protocol which is not network transparent at its core. In a world where internet is finally everywhere, and X (if embraced) would allow to move a game from smartphone to TV when coming home. Or a text window from notebook to tablet for discussion around a table, or .... And no, RDP does not cut it.
1. X is stable. That its old rendering model is not used anymore by modern applications does not change this a bit. With newer extensions, you can do compositing and direct rendering just fine. I don't get your point.
2. People always used toolkits. And I don't see any technical reason why there should be a performance gain with Wayland.
How would abandoning X improve the situation for distros? If they have to maintain Xwayland they now have an even more complicated system (Wayland + X). If not, they break compatibility (which would be really really stupid).
The real reason distros are forked is because Linux userspace is an entangled mess of packages patched into shape to work with ever changing APIs. I am not speaking about kernel-level APIs here, those are rock solid. If APIs would be stable, packages would not have so many versioned dependencies and forking would not be necessary. I am already scared of the day when the idiots make their threat true to replace X with Wayland, breaking one of the remaining stable APIs with decades of backwards compatibility for no good reason.
Or... it is simpy not true: http://www.reuters.com/article/2013/01/22/france-power-grid-idUSL6N0AQECN20130122
I am not sure what you are talking about. It was a government-operated mine (after it was a commercial salt mine before) which served as a test-bed for Gorleben. The thing is: People just appoached this problem in the 60s/70s with the same naivety you can still observe here on slashdot: This nuclear waste is not a problem. Such a small amount, we can just bury it underground somewhere. It is not problem at all!