I'm a little nervous about client-side window decorations. One of the nice things about window-system controlled decorations is that updates in function, policy and style can be done unilaterally. I'm worried that few years after the Wayland transition there will be a cluster of ugly legacy apps that will never be upgraded and don't work right- think xmms, not Google Chrome.

Maybe it is me misunderstanding things, but Wayland and X11 are not technologies in contrast with each other.

Wayland is a glorified and up-to-date framebuffer and input system, X11 is a library that draws shapes and is responsible for delivering input from the user to the correct application.

I think the current situation is confusing because current X11 implementations not only do what X11 is meant to do, but they also need a way to push pixels to the framebuffer and listen to keystrokes. Basically X servers needed their own way to talk to the graphic card and they implemented it (decades ago). That part is almost completely independent from the protocol that draws shapes and routes input to applications. What Wayland is replacing is _that_ part, the part that deals with the framebuffer (that in the modern times is no longer a simple strip of memory region with RGB pixel data).

In the end we will have Wayland as the only owner of the graphic device and many clients, one of which will be the X server. When they say "client-side decoration" they are referring to the status quo: the Wayland client called X11 will draw the decorations (asking the WM to do that); nothing different from what we have now.

Once Wayland will be stable there will be other direct clients that will want to bypass X11: games probably, but also other multi-platform toolkit like Qt. And X12, eventually ;)

To me Wayland looks as the perfect example of how to move forward: make older unsustainable technologies coexist with newer technologies.

Single page version for those who would prefer to read one large page instead of three split up pages:

http://www.h-online.com/open/features/Wayland-Beyond-X-14320...

I felt that the words like 'legacy' and 'dated' were a tad overused, and the article felt biased. A particular example was how X's protocols were described as "complex, asynchronous", contrasted with Wayland's "violently asynchronous" protocols !

It's always thrilling to have a blank slate, and you have a euphoric feeling that this time, we're going to get it right. But there are the essential complexities of any problem, and I would say that for the end user, X has delivered in spades. It will be interesting to hear developers reactions to a battle-tested Wayland code base 5 years from now.

I went to a talk about Wayland at FOSDEM. What was most interesting was that everything they are trying to solve has already been solved by X. Furthermore, the only possible benefit of Wayland (rotating windows in non 90 degree amounts) is basically a bug, not a feature.

Then there's the lack of network support.

If this is pushed on Linux users, expect forking of distributions.

One of my hats is 'system administrator'. I like X a lot more than VNC/RDP. I don't run linux on my desktop and don't really care about performance there.

I'm not looking forward to the day when X is gone. Being able to fire up system-config-* on a remote system is awesome.

It's nice that there is still interesting development work being done in this field. This approach certainly seems like a step forward. Kernel Mode Setting is noticeably faster and has better power saving abilities than previous approaches but is cannot be fully taken advantage of by X.

There is an experimental port to Arch if anyone want to give it a spin (https://wiki.archlinux.org/index.php/Wayland).

The real problem, and confusion in this discussion is because X11 is a loaded concept and when people talk about Wayland they dont seem to understand what part of X11 wayland is supposed to replace.

X11 meanings:

- X11 protocol: Allows remote applications.

- X11 server: Controlling applications and redirect input and such.

- X11 drivers: Controlling hadware.

As i understand it, Wayland is supposed to replace X11 drivers and parts of X11 server. Apparently you can still use the X11 protocol on top of Wayland if you want.

Theres a good FAQ about it: http://wayland.freedesktop.org/faq.html

While many of the changes that Wayland is making seem like improvements, I would really miss server side window decoration if it were to go away. Perhaps a nice way forward would be to have, instead of X on Wayland, another lighter weight server between Wayland and the applications? I suppose this would be very analogous to current window managers.

I get the feeling this will go the way of Plan 9 i.e. its only flaw is that it was trying to replace something very entrenched that was "good enough".

I really like Wayland and would like to see it succeed though.

Really good article. I've had so many problems with X... we can only hope.

This goes to show that it's practically impossible to do it right. At one point there are the purists who complain about bloat. At the other point, there are the purists complaining about NIH when you try to remove the bloat.

You can't have it both in this case it seems.

As a user I know one thing: X11 is huge and while it has all the nice features, it's also still a slow and laggy beast. This might be caused by bad drivers, or it might be caused by a bad architecture (or bad drivers caused by bad architecture making it impossibly hard to write good drivers).

As such, I'm curious to see whether a restart based on technology newer than the 80ies might solve the two big issues I'm having with Unix GUI at this point.

If it does, I'm happy.

If it doesn't, I'll be on the lookout for a better solution.

But I know one thing: If a long-year X11 developer (Kristian Høgsberg is working on the xorg x server) tells us that our issues are practically impossible to fix without a rewrite using a different architecture, then I believe them. Why?

Because for them it would likely be much less work to fix the existing thing (especially if it is nearly 30 years old and as such should be very mature and bug-free), so if they prefer a rewrite, there must be some truth to that.

Personally, I don't care about either bloat (unless said bloat leads to other issues than me having to buy a bit more RAM and diskspace) or NIH, but I still to this day haven't found a Linux distribution that has a GUI which works as-well as OSX or Windows (window drawing issues, multi monitor support - heck - just changing resolutions at times), so I'm certainly open to see other paths explored.

The point of Wayland isn't doing new things that weren't poissible before, it's doing things in a better and more efficient way. Less lines of code, dropping support ancient unused stuff, etc. In other words, reducing technical debt.

> I went to a talk about Wayland at FOSDEM. What was most interesting was that everything they are trying to solve has already been solved by X.

This is because Wayland is meant to be a subset of the functionality of X. It's not meant to do new things, it's meant to strip off 3 decades worth of cruft that is of little value today.

Note that Wayland is not ran by a bunch of random people who decided that X sucks and that they can do better -- a vast majority of the commits are by veteran X developers.

I'm doing occasionally some cross-platform development using Windows API and X11 and what I hope mostly for is that Wayland will reduce the complexity when programming compared to what is needed for X. Even simply things get complicated on X11 due to the client&server communication. A simple call to check the cursor-position can freeze your application for several milliseconds when the server decides to ignore your client for a few frames (and yes, depending on the application that matters - for games it's really annoying). Or adding copy&paste support will send you through a hell of no-longer-working examples and you can find lots of custom implementation where no 2 people tried the same workaround to ensure they really wait long enough for the server to respond and filter the right messages. I think the X people are also working on improving the situation when client&server are on the same system, but unfortunately you never can really ignore that communication completely and it will probably be many years before older X11 systems out there can safely be ignored.

I realize many people care a lot about the network part of X11. I suppose especially if you are working as Sysop it might be a big deal. I've not used it for over a decade, so obviously not something I care about so much. But on the programming side that client&server split makes life a lot harder and is not a solved problem.