I want to fall in love with an all-new minimalistic window manager, written from scratch for wayland. :o

I would like to have a port of i3 for Wayland, but my programming knowledge is too basic to do that myself.

I have a weird feeling that neither Wayland or Mir will ever be truly adopted over Xorg. Yes they're nice concepts but in practicality they're about as useful as a screen-door for a submarine in terms of keeping water out. I see Wayland and Mir "features" being backported into Xorg to extend it's capabilities and redraft Xorg somewhat, but as far as Wayland or Mir taking over for Xorg, not possible. Xorg/XFree86 is too widespread, too used, and too ingrained into UNIX, BSD, and Linux to be replaced on such a massive level with brutal efficiency and effectiveness equally on the same level.

XWayland being a compatibility layer kinda hints at what they want to do, but in reality, selling Wayland is going to next to difficult if not impossible to the various kernels and operating systems out their all UNIX-like and based. The problem is going to be getting everyone on board equally: hardware OEMs, software developers, and even distribution maintainers, and more. Unless everyone is on board with Wayland, more than likely it won't happen.

We will see. Most distributions have Wayland already in their repositories and the larger ones already confirmed that they will switch.
About backporting its features, mostly impossible and missing the point, since backporting those features can only be done when you break compatibility, which is, besides its widespread use, the one major point that can be counted as an advantage for X.
I personally appreciate this development, new does not have to be bad, especially when new is done in the way the Wayland developers have done it.

If done the right way, it could be completely seamless a switch over from X to Wayland, but in order for Wayland to be seamless, it has to have total support in all areas especially desktop environments, drivers, etc.

I wish the Wayland team the best, but I seriously hope they don't rush perfection and end up whipping up some half-butted replacement that only breaks your video.

This. So much of this. Mir will plow on regardless. If it doesn't work in 14.04, Canonical will continue to push it until it does work. As they did with Unity. I dare say that if Wayland isn't right in Fedora 21, many people won't give it another chance for a long time.

If you follow the Wayland development process you will see that there is no rush at all. They have used the last five years only to come to version 1.0 of the protocol.
It is already clear (and Canonical has already announced that) that Mir will not be ready for 14.04. Canonical's plans currently are:
- 13.10 will have Unity 7 running on XMir, but will fall back to Xorg if you (have to) use the proprietary AMD/Nvidia drivers
- 14.04 LTS: Unity 7 on XMir, no fallback to Xorg anymore, Canonical assumes that they will have proprietary Mir drivers at this point, let's see how that works out
- 14.10: Unity 8 (the first Mir native version) running on Mir.

Two things two keep in mind:
- Running Unity (or any other DE/WM) on XMir has no advantages at all, but is only done to force Ubuntu users to be beta testers
- XMir, at least at the current state, is not able to run X applications under Mir (no rootless X), it is solely used to provide a fullscreen window to run DEs/WMs, contrary to XWayland, which can run X applications under Wayland, but can not (and according to the developers never will) run DEs/WMs

That Canonical will continue to push Mir also will not change the technical problems that other DEs and WMs would have to port their compositors to it. Martin Gräßlin, the main developer of kwin, has some very good articles about these problems in his blog, worth a read for anyone interested.
I do neither use Ubuntu nor Unity and as it seems those will be the only options for Mir, so I don't see any point in supporting it. YMMV.

A very biased opinion here says otherwise on 13.10 Beta -> http://www.markshuttleworth.com/archives/1269

"I’m nonetheless surprised that the system feels *smoother* than it did pre-Mir. It might be coincidence, Saucy is changing pretty fast and new versions of X and Compiz have both landed while I’ve had Mir running. But watching top suggests that both Xorg and Compiz are using less memory and fewer CPU cycles under Mir than they were with X handling the hardware directly."

Less memory and fewer CPU cycles is an advantage. But, I don't know how this would translate to other DE/WM.

Would be interesting to know if the memory and CPU load you see removed from X/Compiz magically appear in Mir's top entry. And a smoother feeling is not something reliable at all. Technically it makes absolutely no sense to add a layer to the graphics system just because you can.

EDIT_ Shuttleworth's statement is also contrary to these benchmarks: http://www.phoronix.com/scan.php?pag...lle_xmir&num=1

I suspect this would be the case. A lot of weasel words were used. Like "system feels smoother".

And since when do we trust weasels?

I don't. There is no technical reason at all for Mir's existence. So far I can see only one reason why Canonical does not use Wayland: licenses.
While Wayland, like Xorg, uses the MIT license, so that anyone can make and sell a derived work, Mir uses GPL3 together with Canonical's Contributor License Agreement (CLA). This CLA gives Canonical the right to re-license Mir under any license they want. This enables Canonical, but none of their competitors, to sell Mir under a different license to companies that don't want to use the GPL licensed version (and in general it seems that most companies in the mobile sector are not fans of the GPL3), giving Canonical an advantage over their competitors if Mir gets traction in the mobile sector.

Without driver support Mir won't get far.

It works (at the current state there seem to be problems with the radeon drivers, but Mir is still in very early development) with the FOSS drivers, but Canonical/Shuttleworth claim that they will get support from AMD/Nvidia. There were no statements from them that I know of, but eventually Mir will get driver support, at least when those companies release Wayland drivers, since those should automatically be compatible with Mir. If this will be the case for 14.04, as Canonical claims, I don't know, but my experience with AMD's driver development speed lets me think that this is unlikely.

It more than likely won't.

The FOSS drivers are one thing, but Proprietary drivers are completely another ball game.

As far as being automatically compatible between Mir and Wayland, it might look good on paper to say that, but in practice trying to support two architectures that aren't on the same page in development is a perpetual nightmare. Both projects would have to share a common code base for architecture and API, and they do not.

Even Jonathan Riddell of the Kubuntu project said that Kubuntu would be avoiding Mir in favor of X.Org until Wayland was complete:

http://blogs.kde.org/2013/06/26/kubu...ng-mir-or-xmir

This is like DirectFB and Y-Windows all over again. Something to try and replace X.Org but not doing it incorrectly, too fast, too rushed, and over-hyped breaking everything...

...and we though Canonical learned their lesson with the craptastic job they did with PulseAudio...

...but apparently, they have not.

Driver wise they actually do have most things in common, both will use the EGL driver standard (with one extension for Wayland), so if a Wayland driver occurs it should work for Mir. Sadly it isn't the other way around, but since Wayland is designed to be able to switch the driver backend (for example, with libhybris you can use Android drivers) this is not that much of a problem.

But yes, you are right, Canonical really seems to have a problem with releasing stable software, but I think that is due to its short release cycle and its heavy marketing: They always need buzzwords and features for the next release, so that they constantly are discussed in blogs and never slip out of the public sight. That this costs them their quality (by the way for me the main reason to leave Ubuntu when 10.04 was released) does not seem to matter for them.

Quality is one thing that can make or break a distribution. Sure there is always the following a distribution will have and those who will swear by it, even if everything ends up broken like ArchLinux tends to do from time to time, but even then you have to have some level of quality control on the retail version control end so that whatever you do in -Current trees, can have a free period of stability checks and testing to make sure you get it right without nuking the OS and having to start all over again rebuilding the tree branch to make sure everything is compatible and stable.

The difference between Ubuntu on one side and Arch or development branches of stable distributions, like Debian Sid or Slackware -current (though -current is still exceptionally stable), on the other side is that the users of the latter know that they will get unstable behavior or a broken system from time to time, that is simply the price to pay for living on the edge and they are prepared to handle that, while Ubuntu users get releases that pretend to be stable, while they are in reality released with bugs just because they can't move the release date due to the heavy marketing. If even their "enterprise grade" branch, the LTS versions, tell their users not to upgrade from previous LTS versions before the first point release then there can't be any other conclusion that this release model is seriously broken and incapable of delivering a stable system.
With removing the Xorg fallback in 14.04 and being totally dependent on XMir, an unnecessary additional layer, Canonical even degrades their LTS users to nothing more than beta testers.

One thing you NEVER do to end-users of a an Official release of software is to make the end-user nothing more than a Beta Tester.

I've seen companies of all kinds get a heavy handed slamming by their users and critics over that exact practice, and why very few to this day will release untested, unstable, and bleeding-edge software to the end-users without some level of warning.

Canonical is playing with matches in a room full of Gunpowder barrels. It's only a matter of time before something bad happens.

Which version of KDE will 14.1 ship with? I've noticed the acknowledgement of 4.11 but also that it might be a bit too new.

Unless we get 4.11, I suppose there isn't a lot of point in Wayland being in /extra this time. I am fine with that. I suppose that would be more and an incremental release though. For 15.0 I would hope to see 'some' Wayland support even in /extra, but I guess that is still a while off. I would be interested in seeing Wayland in action though.

Wayland is far from complete enough for proper testing in an official package. Drivers are still lacking and most features are still unfinished. It will probably not be even close to being in /testing for upwards estimated of a year or more until the project releases a stable version.

Looks like the "display server war" just started:

http://lists.freedesktop.org/archive...er/032881.html

Reminds me of "KDE vs. GNOME". That helped Microsoft a lot, too.

How is that ?

I don't see that as a war, it is merely a decision from Intel that they don't want to maintain patches for a one-distro-only solution upstream, those patches belong downstream. Those patches are small and it is Canonical's chosen route to go for their own solution, so of course the burden of maintenance should lie on Canonical's shoulders.

This.

As for the poll, I'm eager to see what comes of Wayland, but in the end I always stick to what Pat puts out.

Canonical is probably going to have their hands eventually forced into Wayland support (even if a separate build) if XMir doesn't get proper support by hardware vendors.

I don't blame Intel either, because Wayland is an official X.Org project, compared to XMir, which is not.

OEMs will support the official project only, not some fly-by-night one-trick-pony distribution maintainer like Canonical.

I bet, nobody will "win" this and a minimum of three incompatible display servers will be around and make stuff even more complicated for users. By then the Linux desktop will finally be on life support, and even the die-hard enthusiasts will step away from that chaos.

I can also see Valve coming up with their own solution like Google. They will get commercial-grade driver support for sure.

I kinda like the idea that one can still run X stuff in Wayland. That is for me a good reason to start and try it out a bit (in a vm at first)
Things don't have to be part of the distribution, I add software all the time, so why not the display server.

That X has flaws and a less than optimal design is well-known. However, it does the thing for me (for now). I am not averse to improvements, if they prove to be improvements. And that's something I can decide for myself best...

However, there'd need to be Nvidia drivers :-) That's a bit my issue, as I have an nvidia card...

Valve is heavily involved in the development of SDL, which for many developers is the preferred way to abstract the underlying system. SDL2 has just this week got support for Mir, so that games don't have to care if they run on Wayland or Mir. Developing another solution wouldn't make sense for Valve.

I hear Nvidia is finally providing basic support for Nouveau; maybe with Wayland/Mir approaching, they're rethinking their strategy on driver development?

Nvidia started working with Nouveau because Nouveau is going to be supporting older non-OEM driver supported chipsets, while providing a modern driver for OpenGL and X that is open source license compatible.

Nvidia has already dropped many 6x00/7x00 and earlier era chipsets down to Nouveau support.

Nvidia does nmost of the work in nouveau to have support for Tegra, the desktop chips are not a priority.
Also, unlike AMD, Nvidia supports its legacy chips well, you don't have to use nouveau if you have a Geforce 6000/7000 card.

[Sorry for the late reply, just saw this]

Thanks for the answer. Though while xterm (and maybe xpdf) might get ported, I doubt that most of the X software I use will. ctwm is old (I love old, exotic soft- and hardware), and it does not appear to me that many people are using it. Same with xv or xli/xloadimage. But if xwayland will stay, that is good news, iff Slackware some day switches to it.

They just want to improve the "out of the box experience" with nVidia hardware. Their original plan was, that you use the VESA or EFI framebuffer until you install nVidia's graphics drivers - it works this way on Windows, FreeBSD and Solaris (Nouveau is Linux-only).

But most distributions now force Nouveau even for GUI installers and Live CDs, so it is unstable or crashes before you even have a chance to get the proper drivers installed. And of course people don't blame distributors for that, they blame nVidia.

BTW: Crashes are not the only issue. If some automatically executed experimental rev-engineered code breaks your $1000 card by mis-programming clock generators, voltage converters and fan controllers, exceeding thermal/power limits, no one will pay for it.

So nVidia's efforts are targeted at getting that code stable enough, that a nVidia-powered machine is at least able to boot Linux successfully. They don't have performance or feature-completeness in mind.

They _should_ blame nVidia.

Remeber .nv? Yes, the project they cut years ago-

http://www.phoronix.com/scan.php?pag...kills_nv&num=1

As far as I know, TobiSGD is correct, almost all the efforts that nVidia have put into Nouveau is for Tegra.

Yes, they have finally released a few documents to help with Nouveau. So? Its a long time after they stopped developing .nv. Yes, what nVidia wants/'suggests' is that you use VESA until you can install the closed source drivers. That shows a limited understanding and/or empathy with the whole idea of open source.....and IMO if Linus hadn't done a bit of name calling they might not have even released the minor documentation to help nouveau.

Anyone running a Titan, Tesla or 'top end' Quadro isnt going to want nouveau. Not much point 'crippling' a $1000 card. Anyway, if its not being used for 'professional' reasons, a $1000 is card is an idiot move. Even a GTX780 is less than $750, and only numbercrunching freaks or hardcore gamers with a good job or rich family would get one of them.

nVidia should know what the situation is, they obviously knew in the past, or else they would never have made the .nv driver.

Is it really too much to ask for decent open source drivers?

Nv(4) did drive the 2D units of older nVidia GPUs made for 2D GDI acceleration. Post-2007-GPUs don't have these units, because no modern desktop OS uses them anymore. So nv(4) lost its purpose.

But Nouveau might still try to drive such a card- and fail. There is not much choice anymore, because with udev now everything is plug & pray.
$1000+ cards are used on Unix/Linux primarily and the nVidia Linux driver development was mainly focused on these cards. Only recently some consumer features like "application profiles" got added due to Valve's move. 3D Vision and other stuff like that is still missing.

It was a 2D-only and had the same purpose as the VESA/EFI framebuffer.

Plus really if you look at it, nouveau will only "get it working" as well. Yes, you should actually install the Nvidia driver if you want complete support however. Relying on nouveau is not very recommended. It works, but remember well that most of this work comes through reverse engineering efforts and only some minimal contributions. It's nice and works but it does have its shortfalls here and there.

And yes, if you have a Quadro card, the Nvidia driver is what you should be using.

Annnnnd: back to the discussion; after all the nVidia-bashing ...
I think Wayland has a good idea; which will really help Linux on the desktop. I think the community as a whole can really benefit from it.
However, there should (still) be the choice: X or Wayland (or Mir); because there are places where X will be preferred and others where Wayland-likes ought to be the primary choice.

A graphical UI without X11, Mir nor Wayland: yes we can!
 

OK, that's only a proof of concept at the moment, but it's amazing!

Chen, Ping-Hsun aka penk has cooked an ISO image that is a small and live Linux distribution including a graphical browser that rely only on a framebuffer.

I've tried it on an USB stick and it works.

Just follow instructions and be aware that after having launched oxide, you'll have to use the mouse to move the cursor to input fields or hyperlinks as else the keyboard is attached to the console.

Granted, the browser doesn't include a menu, so to change websites or pages you'll have to click on hyperlinks or restart oxide, but this put aside everything works as expected (by me, at least :-), including scrolling with the wheel of the mouse.

To give it a go, just visit this page.

Caveat: you can try it in VirtualBox. In a real machine, you'll need an Intel, nVidia or AMD/ATi device with KMS support. But I think that most machines fulfill there requirements nowadays. At least my T61 does ;)

:) The so-called "free desktop" is dead, thanks to RH. I think Android helped "Linux on the desktop" more than anything else.

The only "free desktop" we may have left is BSD and Iilumos based distributions... and RH is working hard to ensure they are ended.

Quite a bold statement; can you clarify? Why would it be dead?

Survival of the fittest.

In the late 90s the Linux/BSD-based desktop was at an uprise with the advent of KDE heavily supported by SuSE, because it had a real technological advantage. But RH was unhappy with the Qt licensing, someone named Miguel de Icaza (who later became a Microsoft employee) successfully split and stalled the effort. It didn't took long and in 2001 with OS X and Windows XP viable alternatives for end-users became available.

Few years later, someone named Marc Shuttleworth tried to revive the Linux desktop and the Vista desaster helped him somewhat. But he was unsucessful, because it wasn't the 90s anymore, inferior DOS and MacOS were dead already, and their successors were good enough. Especially OS X was too much of a competition, regardless of how hard everyone tried to copy their (GUI) ideas. Following in the current decade companies like Adobe and Opera began canceling Linux support, because most of the Linux desktop crowd had moved to OS X.

Today, most "desktop environment" projects are mainly used by their own developers and now they try to copy from mobile OSes. People still using Linux on the desktop talk about how to cope with regular throw-overs and weird design decisions that shall "help the Linux desktop" (topic of this thread), while others just move on to other pastures with predictable roadmaps. Meanwhile embedded Linux invisibly did overtake the world, and the most used Linux-based end-user OS today is Android, which has its own eco-system and doesn't care about RH and their freedesktop.org. So the latter is dead.

Not just Red Hat.

Remember that Qt's owner, Trolltech, was partly owned by Canopy -- the same venture capital outfit that partly owned The SCO Group, which was at the time sueing IBM, Red Hat and others for billions of dollars, and attempting to extort a $700 per seat licence fee from each Linux end user. Ralph Yarro, the Canopy CEO who was thought to have devised this blatant scam, sat on both the SCO board and the Trolltech board. Even when QT was relicensed under the GPL, it was hard to accept this was in good faith when Trolltech's owners were simultaneously arguing in court that the GPL was unconstitutional and invalid [source].

To their credit, Trolltech understood this, and finally got rid of Canopy involvement in 2005. From even before that happy day, Trolltech has been a model of how dual licensing should work. Today, projects are moving from GTK+ to QT, and for good reasons.

Wayland is actually a Good Thing. The thread you must be thinking of is the other popular thread.
Wayland will help the desktop by replacing a complex thing with something that "does one thing".
It won't even do input, that i see is planed for a separate library libinput. So there will be no need to run graphics through a root user.
In my opinion RH is now more of a threat to desktop linux then android.

Exactly that "replacing" is the problem. You can't call something a platform, if their core components get completely replaced every other year by incompatible concepts. Especially if there are a bunch of them competing each other like Mir, Wayland and X11.

http://imgs.xkcd.com/comics/standards.png

Did you ever try running some old Loki games?

Maintaining complex existing code is not as fun as starting the next big thing, just to leave it unmaintained behind after few years. Or do you really expect Systemd and Wayland living longer than HAL, except in the paid long-term support offered by RH? I don't.

Do you know why the C programming language (the foundation of the Linux operating system) is almost unchanged since 1989?

Android (AOSP) is is the de-facto desktop Linux, because it is a stable platform (that means interface stability). It is not a threat. It is not GNU, but that is GNU's fault.

For sure, replacing the 30 years old, not fit for modern machines, X11 is to soon, we should wait a few years more. How did you come to the "every other year"?

X11 had to be refactored and improved incrementally at a faster pace. That's what every other stable platform is doing. But that process is not fun, so it is incompatible with volunteers contributing just for fun. One of the reasons why the "free desktop" movement failed (and XFree86 is dead). Currently the best X11 implementation is the "proprietary X server" developed by nVidia (if there wasn't modular X, nVidia would have their own X server).

I would never have tried Linux in the 90s, if it had started right away with Systemd/Wayland/Mir/whatever instead of network-transparent X11 and proven Unix concepts. Because with that it becomes just another OS/2 or BeOS.
Because that is the expected lifetime of the new half-baked concepts, before they get replaced again.

wayland was cooked up by Kristian Høgsberg
he was the guy that brought you AIGLX (accelerated indirect desktop cube spinning), so yes he knows how X works
also wayland has support (and feedback and contributions) from many, years long, Xorg devs
and on top of that he is very open to input from DE makers
oh, and he is paid now by intel and intel usually does not f up things (other then instructions sets it seems)

systemd, that i personally... dislike to be political, didn't get input from anyone but GNOME (MIR anyone?)
and was not made by someone who spent years dealing with complex designs
("forget common sense and bash your head against the keyboard")
personal opinion, so ignore if your is not same



and yes i know why C is used for the kernel but a bit of history about C first

back in the decades ago Ken Thompson and Dennis Ritchie were working on Unix
they were writing assembly (the ugly AT&T assembly)
at one time they had to port that OS from a closet sized computer to another closet sized computer
so an idea was born to make a "portable assembly" language

now C is pretty low level so its fairly easy to parse into simple logic (a cpu only has simple logic)
but it's also high level enough to be.. to let humans express themselves in higher logic

like structs for example
a cpu does not know wtf structs are, but they are nice to us humans to organize data
(later cpu's got the LEA instruction to simplify getting a pointer by using a pointer to struct+offset)

"modern", "technologically advanced" languages that are "the future" are faaaaar more high level then C
(and C is already far enough from assembly, its in the golden middle)

i tried writing a small program in a newer C standard (99, 11, don't remember)
it wanted me to write a prototype for every function before the function itself
meaning i had ~50 lines for a 30 line program, 'cuz why not

but i digress

funny enough the kernel has some object orientated design in it (since it is a big state machine)
ask Linus why he doesn't want C++ in it

If those people that do exactly that job for years already decide that X11 is at a point where they can only break its design to come any further, but rather leave it alone and begin writing Wayland instead (yes, most Wayland developers are also Xorg developers) then I personally can live with that, since those people are the ones that should know. Especially since those people explicitly state what is wrong and how they plan to do it in the new protocol.
Network transparent X11 was gone when applications moved on from using Motif as their GUI frameworks. With Qt/GTK/FLTK/whatever the "network transparency" is nothing but a badly implemented VNC. By the way, network transparency can be easily implemented for Wayland, you just need to write a backend for that. Currently the developers spend their resources on getting the protocol to the point they want it to be.
Also, would you like to explain where Wayland is against the Unix philosophy?
You might want to share your insights about expected lifetimes of projects with distro developers, maybe that prevents them from switching to Wayland.