Wayland 1.2.0 Released With Weston

An anonymous reader writes "Wayland 1.2 & Weston 1.2 have been released. Features of this quarterly update to the X.Org/Mir display competitor is support for color management, a new input method framework, a Raspberry Pi renderer/back-end, HiDPI output scaling, multi-seat improvements, and various other changes for this next-generation Linux desktop display protocol and compositor."

Comments

Looks good!

[CajunArson]

Wayland & Weston are coming along pretty well and we are seeing increased adoption in both GTK+/QT toolkits and in desktops with upcoming versions of KDE.

One area where the developers need to go out and evangelize is on the front of EGL for proprietary drivers. Yes it's great that Intel's open source drivers (and to a lesser extend the open-source AMD & Nvidia drivers) have EGL support, but both AMD & Nvidia need to be convinced that EGL is important to their upcoming proprietary drivers too.

The irony here is that Mir, which is is seen as a huge competitor to Wayland, could end up helping Wayland enourmously since Canonical doesn't seem to be afraid to pick up a phone and call people at AMD/Nvidia to talk about updating the drivers.

third post!

[fredan]

since I'm using X11 ;-(

Re:third post!

[crazyaxemaniac]

Client / Server is fundamental to thin clients.

There's a vnc/rdp client is about as thin as you get. Wayland compositors can do this. What else is needed?

Wrong Summary

[allo]

Wayland ist not a Xorg/Mir competitor, as mir is not affiliated in any way with xorg. Wayland is the planned successor of Xorg, while Mir is some Ubuntu project.

Re:Wrong Summary

[brodock]

Mir is another alternative/competitor to Xorg... the difference is that Mir is following the idea of a rigid protocol where Mir is following the idea of providing an API... both of then have drawbacks... but I'm confident that Mir could win the run. History has prooved that its not the "perfect engineered solution" that always win... Despite of what you think... in the end... its all about user adoption.

any decent tiling WMs?

[Trepidity]

Is Weston the only choice, or is there anything vaguely analogous to i3 or dwm in terms of how windows are laid out and managed?

Re:any decent tiling WMs?

[raxx7]

Yes and no.

Weston is only a reference implementation of a Wayland compositor.
Wayland developers don't expect it actually to be used by normal users.

Instead, they expect others to implement their own Wayland compositors, as it should not be any harder than writing a similar X Window Manager.
That is what the Gnome, KDE and Enlightmenment people plan to do, convert their current X compositors (gnome shell, kwin, e) into Wayland compositors.

So, eventually, you might get a dwm Wayland equivalent. But it doesn't exist yet.

The Real Story Behind Wayland and X.org

[Anonymous Coward]

Daniel Stone made a great presentation explaining various problems with X11 that Wayland tries to fix:
http://mirror.linux.org.au/linux.conf.au/2013/ogv/The_real_story_behind_Wayland_and_X.ogv

The same presentation is also on YouTube:
https://www.youtube.com/watch?v=RIctzAQOe44

Re:The Real Story Behind Wayland and X.org

[cheesybagel]

What I did manage to grasp from his talk is that the basic X design which he claims is terrible has remained for the most part while their fantastic new designs for things like XInput keep getting obsoleted one after the other. That he does not like the fact that X11 has a lot of extensions so his answer is to rewrite it. What will eventually happen if he ever has success is Wayland will get a lot of cruft as well.

I also noticed he gave no demos of Wayland at all. He isn't even eating his own dogfood. At least the original X designers actually created it to solve a problem they had and they actually used it.

His model of doing everything using pixmaps is also probably going to be a problem if displays keep going to higher resolutions as is happening recently. In that case you may spend a lot less bandwidth sending draw calls rather than the pixmaps.

I also disagree about the claim that VNC is good enough.

Re:The Real Story Behind Wayland and X.org

[complete loony]

What I'd like to see built as a replacement for X11's network abstractions, is a sandboxed VM running on top of the display server, passing arbitrary messages back and forth with the remote application.

For example, take the llvm based pNaCl sandbox that the google Chrome team are building. Expose wayland API's for updating and displaying window pixmaps, and receiving input.

Then you could port the widget libraries from a UI toolkit to run directly in this VM without imposing any limits to creativity and future innovation. Visual feedback for a drag operation or button push can then be animated immediately on the display, while a simple event message is sent back to the host application to trigger further processing.

Plus the application developer could implement arbitrary custom widgets and have the same low latency interaction with the user.

I'm not suggesting that the entire application should be run in this sandbox, but in some cases that may be acceptable too.

Re:The Real Story Behind Wayland and X.org

[Kjella]

What I did manage to grasp from his talk is that the basic X design which he claims is terrible has remained for the most part

What has remained are the parts that you really can't replace without ceasing to be X11, he goes to great lengths to explain how toolkits, compositors and extensions all try to work around it. It's the reason they want to replace X11 with Weston, not the other way around.

while their fantastic new designs for things like XInput keep getting obsoleted one after the other.

Or as others would call it, getting new features. Do I smell a case of WORKS4ME? Didn't need it, don't want it so nobody else should either, X11 is just fine the way it is.

I also noticed he gave no demos of Wayland at all. He isn't even eating his own dogfood.

It was a presentation not a demo, don't pretend you can't find demos on YouTube... There are even LiveCDs so you can try it yourself.

His model of doing everything using pixmaps is also probably going to be a problem if displays keep going to higher resolutions as is happening recently. In that case you may spend a lot less bandwidth sending draw calls rather than the pixmaps.

Which would be relevant if anybody was using X as a drawing library, but nobody does that anymore. There is OpenGL pass-through with GLX, but the final image acts like a pixmap to the X server and I assume there will be something similar with Wayland, in fact as I understand it that's the only way Wayland will work as it has no drawing routines of its own.

Re:Benchmarks please

[jbolden]

Sure.

Speed of light 186,282 miles per second
Speed that a human can detect jitter of an icon tracing a finger 1/100th of a second
size of the earth 26k miles
circumference of the earth 24,901
fastest possible a round trip can occur from the worst 2 spots on the earth assuming 0 latency beyond the speed of light: .0683 seconds
or the earth is about 7x too big for X11 to work.

Chance of us being able to fix either the speed of light or the size of the earth 0%

Re:Can I run Wayland on top of X11?

[jbolden]

X11 runs on top of Wayland. I imagine you might be able to run Wayland on X11 by creating a fake screen but it could be brutally slow. Pretty much, yes you have to wait for driver support.

Remoteability question restated

[LaughingRadish]

Here's a very simple question with hopefully no wiggle room: Suppose I have two Linux boxes, each running Wayland. They do not run X11 in any form or fashion. I am on the console of one of them and in Wayland. Can I start a terminal emulator, ssh over to the other box, issue a command that starts some graphical program (which uses only Wayland coding, no X11), and expect that program's window to show up on the first box? Assume that ssh has already been modified to allow for this sort of thing. If this cannot be done, what prevents it from being done? I have yet received no complete answer for this.

Re:Remoteability question restated

[jbolden]

Wayland is going to be implementing some like RDP to handle this. Wayland natively does not handle this. So if your question is in terms of "Wayland as it is likely to exist" then likely you will be able to do it. If your question is "Wayland by itself with none of the supporting ecosystem" no. On the other hand normal screen sharing stuff like VNC would work.

What prevents it from being done is that Wayland applications share their graphical and application buffer. You can't pull it apart without virtualizing the entire screen like VNC.

Re:Can I run Wayland on top of X11?

[Tailhook]

Yes. Specify the x11-backend.so when starting Weston and it runs on top of X.

Re:Can I run Wayland on top of X11?

[serviscope_minor]

X11 runs on top of Wayland.

X11 can run top of Wayland. It doesn't necessarily.

I imagine you might be able to run Wayland on X11 by creating a fake screen but it could be brutally slow

If you do it badly, sure.

Wayland wants the world to be a collection of draw buffers (i.e. one per window more or less). There's no reason you couldn't simply have one X11 window for each Wayland window. Given it's all GL all the way down and X11 supports pretty much the fastest 3D graphics so far, I don't see why it would be slow.

Re:Benchmarks please

[serviscope_minor]

The problems with X11 start at things like the number of round trips the client and server have to engage in.

If we're talking about WAN network connections, then yes. For local stuff, people still complain about X being slow, but the data just doesn't support it.

150 round trips x 200 ms latency = 3 seconds till the window gets finished drawing.

X is a bit chatty. There seem to be several reasons.
1. The protocol itself is a bit too chatty. This could be improved greatly if the server could push out events for certain things so that the client can cache them reliably.

The NX people have essentially shown that this is indeed possible and the NX extended protocol seems to be one of the best remote graphical displays to use over a WAN.

2. People can't code X11 protocol for toffee. A classic example is making heavy use of XInternAtom as a synchronus call. In fact Keith Packard himself uses that as a reason for X11 having problems. If you need 100 atoms, you have to wait for 100 round trips.

Interestingly though in the protocol this is not a synchronus call. The newer XCB binding actually has it as an asynchronus call and in fact one of the first examples in the XCB documentation is how to get all your atoms in a single round trip.

Xlib (not X11) is substantially at fault. The correct thing (implementing a new C binding, namely XCB) has been done. However, even with Xlib, toolkit authors seem to love working under the assumption that they're working locally and misuse the protocol in all sorts of bad ways, making round trips where none are necessary.