Slashdot Mirror
Wayland Isn't Ready For the Fedora 24 Desktop (phoronix.com)
An anonymous reader writes: There was much hope that Fedora 24 would be the first major Linux distribution using Wayland by default in place of an X.Org Server, that didn't pan out with Fedora 24 Workstation developers deciding not to use Wayland by default but it will remain a log-in time option. Fedora Wayland has made a lot of progress but functionality like on-screen keyboard, accessibility, remote displays, USB display hot-plugging, and other functionality is incomplete for the Fedora 24 timeline. At least there are many other Fedora 24 features that made it for this next release due out in June. Wayland will turn eight years old this year.
What unsolved problem does Wayland address?
No one has any difficulty securely running GUIs across networks.
I'm sure the developers have been using things like covscan, valgrind, cppcheck, and etc., throughout the development.
So we can expect it to be fairly clean right out of the gate, right?
right?
Yes indeed
"If any question why we died, Tell them because our fathers lied."
Hm.
As an open source developer for minetest, I've been really hoping that Fedora doesn't ship wayland yet.
Our game client simply isn't usable on wayland: https://github.com/minetest/mi...
And this isn't about minetest not being able to run natively, no its a bug with the abstraction layer, xwayland. I understand the underlying issue that the devs want to forbid pointer warping, this is one of the security features of wayland, but please find a way for legacy applications like minetest to still run. All I've seen is pointer locking and pointer confinement being discussed, but no warping specifically for applications that still use and rely on X.
And other games have this precise problem too.
Its okay if the devs need some time to develop a great successor for Xorg, that's fine, Its a very large task, and I welcome the concepts of wayland. But Fedora, a fairly stable distro, really shouldn't ship immature software to their users as the default option.
// Lame joke about Hurd & DNF goes here
Confucius say, "Find worm in apple - bad. Find half a worm - worse."
XFree86 Release 4.8.0 is out NOW
4.8.0 release was released on 15 December 2008. Our next full release will be 4.9.0, and is expected to be released in the summer/winter of 2009
We make our releases on a yearly basis and each release number matches the year i.e. 2008 = 4.8.0.
Our Current Status
Currently the Project is in the development phase for 4.9.0, and we are currently accepting development code.
How's that working out for you?
Yet at the end of the day it's still... wayland.
The X protocol is not in line with how modern graphics is handled, this precludes good compression on X over ssh and also forces a lot of unnecessary complexity which is bad for security. Although I would be happier if they had just called it X12 an update that included dumping some of the old stuff in the core of the old protocol was necessary, now that they have decided to include network "transparency" equivalent functionality (single window tunnelling over ssh) and select+middle click paste the rest of it just looks like clean-up.
For Fedora, which underpins RHEL and other Enterprisey OSes, that's a major absence, even if Wayland's own developers don't consider it important.
I really hope Wayland's developers stop treating it as a minority application unworthy of serious consideration (even though it's supposedly on their long term roadmap) and actively start work on it. They have a proof of concept. They have X to show them how security can work in practice. It's time the work was done.
You are not alone. This is not normal. None of this is normal.
Also known as the Cascade of Attention Deficit Teenagers development model.
"I know I will be modded down for this": where's the option '-1, Asking for it'?
Did you miss Perl 6 hitting 1.0 last year?
I've used Wayland on Fedora on my laptop. Is it perfect? No, but absolutely usable. That's a big upgrade from just a year ago.
Replacing X is a big project. Sometimes it takes a while to generate something good.
I'm not sure anyone has a good model for handling rewrites of massive projects. The experiences of KDE 4.0 and Gnome 3.0 come to mind. Eventually, they were better, but it takes some time with a massive upgrade like that.
The other issue is that User's often have a very good idea of what they don't like. However, bulimic criticism does not help to refine a software product. It just splits the ecosystem. Ultimately the user's need to use their computer, and the new software just isn't ready. So the developer's and user's go in different directions.
Closed source isn't the solution either: with Windows 8, Microsoft split it's ecosystem. Windows 10 hasn't fixed the split (yet).
"The Fedora 24 Desktop Isn't Ready For Wayland"
My first program:
Hell Segmentation fault
Yes, Linux very badly needs a decent windowing system to make it competitive with Windows and OS X. They've both always been much better than X, but they've really kicked the living shit out of it the past half-decade.
I read a very negative review of Wayland recently.
It sounds pathetically unusable to me. I mean the tester had problems with just about every single application that was tried! There were crashes, slow performance, incompatibilities, lags, and all kinds of other just plain dumb problems like that.
How the fuck is Wayland supposed to be our savior when it can't even compete with X, for fuck's sake?!
I don't give a fuck if X does "too much crap". It's actually somewhat usable! Wayland? Fucking unusable!
Hasn't systemd deprecated Wayland? Or is that just Rawhide?
What I love about X is the flexibility one gets, which is unparalleled by any other system: I can easily start a window of an application running on another host on my machine, works fine if I am logged into that other machine using ssh. I can tunnel a whole session through sse usinv VNC and use the remote desktop directly on my local one. It supports mutliple monitors spanning one desktop or several desktop on several monitors.
Does Wayland support these things too?
For example, how do you use remote windows or something like VNC on either of those right out of the box?
Replacing X is a big project. Sometimes it takes a while to generate something good.
ISTR Gettys and Packard talking about it ten-plus years ago. X is big, but it's not that big. And before you claim I don't know, let me assure you, I have plenty of experience with X which I'm not going to detail here; trust me, I know.
Closed source isn't the solution either: with Windows 8, Microsoft split it's ecosystem. Windows 10 hasn't fixed the split (yet).
That's the UI your talking about, which many have absolutely no problem with. Wayland is trying to bring Linux up to par with the Desktop Window Manager that was introduced with Vista and is still current with Windows 10. A rock-solid, tear and flicker free, error tolerant display that provides a composited (GPU accelerated) desktop that even works with software written for Windows 95. That's the result of paying a team of dedicated developers to create something with a clear goal in mind.
I'm waiting for the Yutani release of Wayland.
That's because most people who believe Wayland is the second coming are completely unqualified to even hold an opinion. However, those who object to Wayland and it's justification tend to do so based on technical knowledge and expertise.
Wayland is doomed to epic failure until they make serious changes. One of the biggest is remote windowing. They've finally acknowledged that this is in fact a serious issue and glaring deficiency with Wayland. They might actually garner broader support to make things better and reliable.
Why not DELAY the next Fedora version so people aren't forced to upgrade as often?
But some people will use a goldfish bowl and a abacus before they use anything from microsoft.
One of the biggest is remote windowing.
This is a core feature of X11, but it hasn't worked since X11R5. In the 90s you could get an X terminal from NCD or HDS that ran nothing except an X server, and communicated over 10Mbit/s Ethernet, and this was bug-free and almost as fast as local. Since then it's bitrotted, the performance has gotten worse, and many extensions are just broken so that almost everyone uses vnc.
Instead, they should probably use NX, which performs much better than vnc, but it's proprietary. or maybe xpra, but I haven't tried it.
I agree that efficient remote display for a subset of applications is a core feature that should gate Wayland acceptance. I don't agree that X11 has this feature. People just think it does because they remember it used to and are skilled at victim-blaming doublethink when they try stuff on Unix and find it doesn't work. "You are holding it wrong." No, it's broken.
Video codecs seem like a promising way to get this feature back and keep it, like what Chrome Remote Desktop is doing.
Unfortunately Google is impossible to work with. While they do great work on VP9 and on idiot-proofing things, they arrogantly force their "cloud" on you assuming the only obstacle to using a Google Account is convenience, they do not finish the job (it keeps needing restarts when the server has a bad connection, probably because they only test with servers inside their "cloud". Xvfb servers with RandR extension are rare, probably because they just fix the distribution they use, fork-and-forget. Keyboard mapping is still all busted, yet somehow sane in both NX and traditional X client-server.), they get bored with things and kill them and always hold back some of the source code. (The last two are related because if they made proper open releases they wouldn't be able to kill things in which they lost interest.)
But if someone _else_ would invent a vnc-like remote feature based on a video codec, maybe a GPU-accelerated one, that might be a good replacement for the X protocol. We have a lot more processing power than we did in the 90s, and the bandwidth of video is not necessarily that bad: better for keyframes, worse for steady-state. For example, 90s X terminal peak bandwidth 10Mbit/s is totally acceptable for video. Currently I think it's bad compared to X11 wrt total bits sent and latency, but it's probably still the right way to do this.
One problem with the old way is that the assumption the conversation about how to compose the screen will use less bandwidth than the pixels on the screen is no longer remotely true with all this newfangled GL nonsense.
Ah, Wayland. The Hurd of windowing systems. I'm sure it will be amazing once finished.
>> One of the biggest is remote windowing.
> This is a core feature of X11, but it hasn't worked since X11R5.
Don't know what you're talking about, I use it all the time. Works fine.
Usually just for random applications, but not always. As an example, I was working from home recently and needed to work on a remote VM. Sadly, my company's Linux VPN access is temporarily broken until they update some things, so I had to connect from a Windows VM. Finally, I wanted to use my home PC's display because it's much bigger than my work laptop's, but didn't want to mess around with looking for cables and rerouting stuff.
So, I ssh'ed from the PC to the laptop; started the Windows VM in VirtualBox on the laptop, displaying on the desktop; logged into the VPN in the Windows VM; connected to my work VM; and did my work. It worked fine. There was some lag, but not all that much. It was perfectly usable.
This may not be a typical use case. But really that's the entire point. No reasonable person would expect the X developers or the Wayland developers to understand every single person's needs. But on the X side, since this one basic feature - remote display - is built in, an enormous number of use cases are supported out-of-box. On the Wayland side AIUI, you can only remotely display the entire desktop, which is less flexible, more convenient, and more resource intensive.
It seems like the Wayland developers are taking a more flexible line on remote display of individual applications lately, and Wayland Over Wire is an extremely promising development. I'm hopeful that the situation will continue to improve. Now if only they'd get rid of client side window decoration...
D'oh. Meant to say that displaying the entire desktop is *less* convenient, of course.
JWZ might have been describing himself. He's a bigger man-child than ESR, and has even less credibility.
This is a core feature of X11, but it hasn't worked since X11R5
WTF is this shit? This FUD has been cropping up on every wayland thread recently. It is outright wronf. Remote windowing still works just fine. It did not stop working in X11R5. This is easily verified.
SJW n. One who posts facts.
+1
We run over 150 [Linux based] thin clients using X11R7, and before than, on X11R6. And being thin, that means remoting the ENTIRE DESKTOP SESSION- window manager, clients, everything. And those client apps come from various places on various servers, sometimes even the local machine.
Now, this is a *BUSINESS* environment.... we are not trying to push video games, music, or movies through X11. That won't work well. But Firefox, LibreOffice, Clawsmail, GIMP, Pluma, Inkscape, Pidgin, PDF viewers/writers, etc, and all our AP/GL/AR/Payroll/etc work just dandy.
Holy fuck! Can somebody please mod the parent down? He's totally wrong.
The Perl 6 1.0 he's talking about is the standard, not an implementation! There is no official Perl 6 implementation like there is for Perl 5. There are several Perl 6 implementations, and all of them are shit. Rakudo Star is the least-shitty of them all, but it's still very shitty in my experience. I have found it to be very slow, and I've had it crash on me. I don't think I've ever had the Perl 4 or Perl 5 interpreters crash on me, and I've used them for so many years now. There was also the Pugs implementation, but it died as a project when the maintainer had some sort of a gender identity crisis. There's also the damn useless Yapsi implementation, which is useless because it's implemented in Perl 6, and Perl 6 hasn't really been implemented yet!
Perl 6 1.0 doesn't actually exist in any useful form.
I ran Wayland during F23 Beta and on F23 for about ten weeks. It worked a lot better than in F22 but still has some usability issues. A couple of examples: In gedit the user should be able to drag a tab to a new window. In Wayland this causes gedit to crash. In XWayland apps such as Firefox or LibreOffice, the cursor will randomly disappear. Once that happens you have to restart Gnome. I am a bit frustrated with the lack of Wayland fixes in F23 promised here https://fedoramagazine.org/hel... That being said I encourage anyone running F23 to try Wayland. Most of the time you won't notice a difference. I'll run it on a spare system when F24 beta lands.
> From the XFree86 web page:
>> XFree86 Release 4.8.0 is out NOW
>> 4.8.0 release was released on 15 December 2008. Our next full release will
>> be 4.9.0, and is expected to be released in the summer/winter of 2009
[...deletia...]
> How's that working out for you?
In case you missed it, there was an internal revolt inside the XFree86 group, and XFree86 code was forked as Xorg, which is the current implementation. The last person to leave the XFree86 project forgot to turn off the lights.
XFree86 is passed on! This project is no more! It has ceased to be! It's expired and gone to meet its maker! It's a stiff! Bereft of life, it rests in peace! If you hadn't nailed it to the perch it'd be pushing up the daisies! Its metabolic processes are now history! It's off the twig! It's kicked the bucket, It's shuffled off its mortal coil, run down the curtain and joined the bleedin' choir invisibile!! THIS IS AN EX-PROJECT!!
See the current Xorg location http://www.x.org/wiki/ It actually has stuff from late last month, rather than late last decade.
I'm not repeating myself
I'm an X window user; I'm an ex-Windows user
People doing payroll on a LAN practically do not matter. People sharing windows in teleconferences outnumber them by several orders of magnitude, and people use Windows and Macs for that type of use case. You may not know that MS and Apple got into a brief escalation/competition around 2000 over desktop conferencing, and in the process leapfrogged X network transparency considerably.
X cannot share a window with 10 or 20 people efficiently. Linux users reach for VNC for that use case, and it is an inefficient throwback... nothing more than a bitmap-tosser.
That is why we should have been putting effort into something 100% backwards compatible with X11... X12. All kinds of things COULD have been rolled in- compression, local cursor, broadcast, etc.
Dude, he left the Mozilla project because they opted to rewrite from scratch rather than continue to maintain the Netscape code he worked on back in the day. He wrote, and continues to maintain, Xscreensaver in straight xlib. No GTK, no Qt, straight Xlib.
And the goal post shifting keeps happening.
If someone demonstrates that X can be used for something the Waylanders claim it can't they bring out another example, and the dance continues.
Elsewhere this kind of debate pattern has been likened to moat and bailey fort building.
http://blog.practicalethics.ox.ac.uk/2014/09/motte-and-bailey-doctrines/
No wonder Wayland is taking so long. The core Wayland team shares a lot of the same people as the core Xorg team. You do realize that people who made Xorg are mostly the people making Wayland because Xorg is such crap from the multi-decades of backwards comparability and old architecture that is no longer reflective of modern systems. Bailing wire and duct tape.
Replacing X is a big project. Sometimes it takes a while to generate something good.
I'm not sure anyone has a good model for handling rewrites of massive projects. The experiences of KDE 4.0 and Gnome 3.0 come to mind. Eventually, they were better, but it takes some time with a massive upgrade like that.
The other issue is that User's often have a very good idea of what they don't like. However, bulimic criticism does not help to refine a software product. It just splits the ecosystem. Ultimately the user's need to use their computer, and the new software just isn't ready. So the developer's and user's go in different directions.
Closed source isn't the solution either: with Windows 8, Microsoft split it's ecosystem. Windows 10 hasn't fixed the split (yet).
Perhaps we just defer Fedora24 for six months, and allow the F24 enhancements to be rolled into Fedora23. From a workspace user, there is little difference between F22 and F23. So, we could say, Fedora23 is a rolling release. And Fedora23 becomes Fedora23.1
Leslie Satenstein Montreal Quebec Canada
Anyone remember Fresco?
No?
Me neither.
Seeing a lot of comparisons between X and other remote-access protocols such as VNC. From a personal use perspective I've found:
One nice thing about X is that the active window is independent of a particular desktop, or parent etc. When I use VNC, one frustration is that everything is bound inside the parent window (which is generally also restricted to a particular monitor). Larger desktops tend to suck, performance-wise, as you end up with a lot of pricey redraws.
One *nice* thing about VNC is that you can push stuff that's been accelerated on the remote GPU.
There's also stuff like Citrix, which I've that in concept is nice enough to draw things as separate windows without an MDI-style parent or virtual desktop, but in practice weird crap starts happening as soon as I move said windows between monitors etc. One good way to mess with Citrix is to have the overall desktop size change (say by hotplugging a monitor). It's also sometimes ugly with overlapping windows etc.
At the moment my current setup is a combination of a remote X window and VNC (X11VNC+tightvnc). I've been working on OpenGL-based applications, which won't render across machines with X, however the beefy hardware is not on my main box. To that end, I've got the remote machine running X11VNC (which is able to grab the framebuffer) and tightvnc locally for rendering. That at least gets me the hardware acceleration for rendering. In addition, running the actual code editor+compiler via X allows me to have all the dialogs, debuggers, etc on my local machine and move stufff between monitors with ease.
I believe there are tricks to make this work all in X-land, e.g. with xvfb, but I haven't had much luck with that yet. What would be nice is to see those capabilities all pulled together in something like Wayland where the network stack, framebuffer, etc and all the userland stuff play nicely with each other out-of-the box with minimal hackery needed to get things to work.