It's Time To Admit It: The X.Org Server Is Abandonware

[oiaohm]

I really believe X11 shouldn't be abandoned. And, if it really had to be replaced by something better, I don't think Wayland is the solution. The Wayland website argues that there are obsolete features in X11 which belong to the 80s and nobody uses them anymore. Well, that's fine, but such features could be deprecated if needed without starting a new thing from scratch. Moreover, then they argue that remote rendering won't be supported in Wayland because it's "orthogonal" to it and it's "too complicated". This is when they show they don't have the slightest idea of what they are doing: do they want to remove old features, or just build their dreams toy with only the things they consider cool to have? They even recommend to use Xorg if you want remote rendering!! Are they UNIX developers? I mean, don't they need to run software in a HPC server? They really look like their idea of UNIX is using Gimp in Ubuntu (because if that's your idea of UNIX, then yes, Wayland is better than X11).

Really a lot of the deprecated features in X11 have been marked as so for over 2 decades and you still have the Nvidia binary blob using them instead of their newer replacements that all the open source drivers use.

Remote rendering in X11 protocol being marked at "too complicated" is a valid answer. xpra exists because X11 protocol include remote rendering is a broken pile of crap.
https://xpra.org/trac/wiki/Usage/OpenGL The reality is you get better remote rendering X11 support using Xwayland with xpra on the application side due to standard X11 problems doing opengl.

https://gitlab.freedesktop.org/mstoeckl/waypipe

There is a recommend way of remote render wayland applications it is waypipe is not as feature rich as xpra yet.

Cesarcafe sometimes way a feature has been done is wrong. The remote render design in X11 protocol is wrong because it has all operations needing to go from the X11 application to the X11 server and back before the application can go forwards. Yes xpra/NX... sitting on the remote connection allows X11 application to proceed forwards straight away without having to worry about network lag.

Question should application really know network is there? The wayland developers answer no. This is why the wayland remote render answer is something like waypipe you will run a wayland/X11 compositor local to the application being network forwarded so the application not being stalled out by network latency issues and so you can really use a GPU to accelerate opengl and vulkan without jumping though insane number of hoops.

cesarcafe the reality is the obsolete features in the X11 protocol are obsolete even you remain using X11. X11 remote rendering in protocol is obsolete by the way xpra/nx/xvnc does remote. If you look at xpra/nx/xvnc the reality is xvnc wayland does have its ways of doing this with remote desktop protocols yes its a low level driver solution then the xpra and nx method is the waypipe method.

The wayland developer saying that remote rendering will not be support by wayland protocol is right because if you look at xpra/nx/xvnc they are not using X11 remote rendering protocol instead they are after local rendering they can send remote. If you are going to be local rendering all the application output the results is applications don't need to know about remote rendering. Remember wayland protocol is between application and compositor. If you are doing local rendering a compositor has to be local. The compositor need to know about remote rendering the applications don't need and it need to be this way so applications can perform well and that is what xpra/nx/xvnc tell us.

Cesarcafe basically as soon as you started arguement for remote rendering to be part of Wayland protocol you show lack knowledge in how remote rendering with X11 has been done over the past 15 years because the past 15 years tell you that the path Wayland protocol is on is right.

[royce]

How dare you coming here with real knowledge. Shoo.

[cesarcafe]

Really a lot of the deprecated features in X11 have been marked as so for over 2 decades and you still have the Nvidia binary blob using them instead of their newer replacements that all the open source drivers use.

Remote rendering in X11 protocol being marked at "too complicated" is a valid answer. xpra exists because X11 protocol include remote rendering is a broken pile of crap.
https://xpra.org/trac/wiki/Usage/OpenGL The reality is you get better remote rendering X11 support using Xwayland with xpra on the application side due to standard X11 problems doing opengl.

https://gitlab.freedesktop.org/mstoeckl/waypipe

There is a recommend way of remote render wayland applications it is waypipe is not as feature rich as xpra yet.

Cesarcafe sometimes way a feature has been done is wrong. The remote render design in X11 protocol is wrong because it has all operations needing to go from the X11 application to the X11 server and back before the application can go forwards. Yes xpra/NX... sitting on the remote connection allows X11 application to proceed forwards straight away without having to worry about network lag.

Question should application really know network is there? The wayland developers answer no. This is why the wayland remote render answer is something like waypipe you will run a wayland/X11 compositor local to the application being network forwarded so the application not being stalled out by network latency issues and so you can really use a GPU to accelerate opengl and vulkan without jumping though insane number of hoops.

cesarcafe the reality is the obsolete features in the X11 protocol are obsolete even you remain using X11. X11 remote rendering in protocol is obsolete by the way xpra/nx/xvnc does remote. If you look at xpra/nx/xvnc the reality is xvnc wayland does have its ways of doing this with remote desktop protocols yes its a low level driver solution then the xpra and nx method is the waypipe method.

The wayland developer saying that remote rendering will not be support by wayland protocol is right because if you look at xpra/nx/xvnc they are not using X11 remote rendering protocol instead they are after local rendering they can send remote. If you are going to be local rendering all the application output the results is applications don't need to know about remote rendering. Remember wayland protocol is between application and compositor. If you are doing local rendering a compositor has to be local. The compositor need to know about remote rendering the applications don't need and it need to be this way so applications can perform well and that is what xpra/nx/xvnc tell us.

Cesarcafe basically as soon as you started arguement for remote rendering to be part of Wayland protocol you show lack knowledge in how remote rendering with X11 has been done over the past 15 years because the past 15 years tell you that the path Wayland protocol is on is right.

Of course, X11 has defects and limits from the beginning, just like any technology has them. But X11 works today, and it's being rock-solid for decades. If I need to connect to a compute server from my desktop and I need to run any kind of window-based app in the server, it must be done with X11, there's no better solution. There's nothing easier nor more failsafe than setting your display properly and using X11.

Then you might say you are going to fix X11 by developing a new stuff that cannot do remote rendering. Sorry but no: start by desigining a desktop window system with rock-solid, safe, well designed remote rendering from the beginning. When you have the remote rendering properly designed, then you can start thinking in how it works on your desktop display, but not in the reverse. Wayland, unlike X11, is designed in the reverse (considering there's no network), and because of this it's broken from the beginning.

You can argue X11 has defects. Fine (and you are right). But adding proper network support to Wayland at the end of its design, will either A) End up with a design much more poor and bloated than X11, or B) End up never implementing networking and telling the user "we replaced X11 with this thing that doesn't support networking, because we thought you don't really want remote rendering, and that if you need it, you should configure your own vnc system". Honestly, neither A) nor B) convince me. X11 much better. No doubt.

[oiaohm]

Of course, X11 has defects and limits from the beginning, just like any technology has them. But X11 works today, and it's being rock-solid for decades.

Its not in fact decades. 1990s you had manual configuration it was only in 2006 you started getting auto configuration. X11 crashing was a lot more common a 2 decades back. I remember in 2000 of having X11 server lock up because it could attempt to switch to terminal now have Nvidia driver because I have attempt to switch to terminal kernel panic. Great just lost all my work and got long wait as harddrive is check and having screwed up other things.

If I need to connect to a compute server from my desktop and I need to run any kind of window-based app in the server, it must be done with X11, there's no better solution. There's nothing easier nor more failsafe than setting your display properly and using X11.

This is miss truth. xpra exists for X11 because using raw X11 over network when you have disruption in network results in X11 applications crashing/terminating.
The reality you need a "persistent remote display" solution. This is also what waypipe provides with wayland.

Then you might say you are going to fix X11 by developing a new stuff that cannot do remote rendering. Sorry but no: start by desigining a desktop window system with rock-solid, safe, well designed remote rendering from the beginning. When you have the remote rendering properly designed, then you can start thinking in how it works on your desktop display, but not in the reverse. Wayland, unlike X11, is designed in the reverse (considering there's no network), and because of this it's broken from the beginning.

X11 protocol when it started did not have a desktop window system at all. Anyone who has do bug triage for wine covering the issues with different windows mangers with X11 will tell its not stable or rock-solid.

Remote rendering was designed into X11 at the start problem it was not design to be persistent remote display to deal with network issues so expects 100 percent perfect network with no packet issues. Really the odds of a perfectly bug less network???

You can argue X11 has defects. Fine (and you are right). But adding proper network support to Wayland at the end of its design, will either A) End up with a design much more poor and bloated than X11, or B) End up never implementing networking and telling the user "we replaced X11 with this thing that doesn't support networking, because we thought you don't really want remote rendering, and that if you need it, you should configure your own vnc system". Honestly, neither A) nor B) convince me. X11 much better. No doubt.

Persistent remote display means you must have a local rendering solution first before you can go remote. X11 early history the servers did not have enough grunt to render everything so they went to a non persistent remote display out of limitations of the time to push rendering off on to the client machines. Yes the X11 design increases your points of failure.

Own VNC system guess what that is a persistent remote display system.

[cesarcafe]

Its not in fact decades. 1990s you had manual configuration it was only in 2006 you started getting auto configuration. X11 crashing was a lot more common a 2 decades back. I remember in 2000 of having X11 server lock up because it could attempt to switch to terminal now have Nvidia driver because I have attempt to switch to terminal kernel panic. Great just lost all my work and got long wait as harddrive is check and having screwed up other things.

In the 90s I didn't use Linux, but commercial Unices (mostly IRIX). X11 ran out of the box (maybe sometimes you needed to set the DISPLAY environment variable, but that was it... everything ran out of the box without any manual configuration). I remember hardware acceleration was a pain in Linux those days, but I didn't have to fight with it, because as I said, in my lab everything was commercial UNIX that ran out of the box.

This is miss truth. xpra exists for X11 because using raw X11 over network when you have disruption in network results in X11 applications crashing/terminating. The reality you need a "persistent remote display" solution. This is also what waypipe provides with wayland.

First time I heard about xpra. Never used it. Remote X11 works fine for me without that.

X11 protocol when it started did not have a desktop window system at all. Anyone who has do bug triage for wine covering the issues with different windows mangers with X11 will tell its not stable or rock-solid.

Remote rendering was designed into X11 at the start problem it was not design to be persistent remote display to deal with network issues so expects 100 percent perfect network with no packet issues. Really the odds of a perfectly bug less network???

Persistent remote display means you must have a local rendering solution first before you can go remote. X11 early history the servers did not have enough grunt to render everything so they went to a non persistent remote display out of limitations of the time to push rendering off on to the client machines. Yes the X11 design increases your points of failure.

Own VNC system guess what that is a persistent remote display system.

Any new windowing solution that doesn't consider remote rendering as a key point in its early design stage will end up either not supporting remote rendering, or not doing it in a clean way.

[oiaohm]

First time I heard about xpra. Never used it. Remote X11 works fine for me without that.

As you get into modern applications X11 cease to. Also xpra deals with using dialup and other horrible unstable connections. Sorry to say remote X11 does not work fine without it sooner or latter you will have a X11 application decide to terminate because connection between application and X11 server fails resulting in data loss. Of course rose colour glasses you most likely forgot the data loss you suffered or just accepted it as part of doing remote that you would have failures.

Any new windowing solution that doesn't consider remote rendering as a key point in its early design stage will end up either not supporting remote rendering, or not doing it in a clean way.

This is a mistake on your part. Why is wayland designed without applications being able to know where they are on screen this is not just for security. The lead developer for wayland demoed why with first demonstration of wayland working over network that leads to waypipe.

Turns out that Wayland was not design with zero consideration to remote display. This is the problem you are presuming something that is not fact.