Wayland Isn't Going to Save the Linux Desktop (2022) | Hacker News Reader

Wayland Isn't Going to Save the Linux Desktop (2022)

dudemanguy.github.io

74 pointszinekeller131 comments

I updated my system recently and KDE under X11 started stuttering all the time, especially when opening new windows (which made using IntelliJ unusable, for example). Switched to Wayland and it just works. That's all it really took for me to go from a "I don't care about Wayland" guy to "I am a Wayland truther" guy.

Similar story for Pipewire; Pulseaudio had crackles and Pipewire didn't. Such is the way of life.

And as a counterexample, I've been using KDE (currently 23.08.0) + Plasma and JACKd for ages with xorg (currently 21.1.8) and have had zero troubles.

Having said that, I'm super-duper glad you found a workaround for trouble you were having.

I use kde x11 for 3/4 years and Linux more and this problem is not related to x11 but to the dependencies it has you system, the updates it makes and the combination of the version of different packages. Changing the bodywork is one way to remove the scratches you make, but it doesn't mean the problem is the body material.

So KDE worked fine, they sabotaged it to force you to switch, and now your conclusion is oh, they were right all along?

> they sabotaged it to force you to switch, and now your conclusion is oh, they were right all along?

Sure, let's go with that!

I'm using X11 here for various reasons, without any stuttering or anything. I'm not running KDE, but i3 with a nice helping of compositing, most notably blurred windows. And this isn't even some kind of absurd machine I'm talking about, but a 3 yo "ultrabook" with integrated graphics. WTH does KDE do to stutter?

I'm using an 8 year old PC with KDE/X and it doesn't stutter. shrug

I have issues on both x11 and Wayland, but the least amount of issues on Wayland.

I think it is mostly because of individual applications, not particularly KDE/Plasma. Some software works better on Wayland, some on x11, some on both.

I have no investment in Linux desktops, but this kind of sounds like "my favorite website stopped working in Firefox, now I'm a Chrome guy." Could this be characterized as it was working before but degraded as development attention shifted to Wayland?

There is nothing inherent in Wayland that makes the it work better as opposed to X11. It is all about developer focus. If KDE developers spend the majority of their resources on the one thing it obviously gets better.

I 100% hope Wayland's story mirrors PipeWire's... When PipeWire became installed by default on my distro, it was flawless.

If, when Wayland becomes default, it is as faultless as PipeWire is right now, it will join the pantheon of successful open source software, alongside the kernel, git, gcc, clang, etc.

I hope Wayland's story mirror's Pulseaudio's, in that it limps along until something actually good shows up to replace it (ie. Pipewire) and everyone, relieved, jumps on it at once.

This was my experience as well. On X11 + i3 (with no DE), latency and screen tearing was always an issue I had to fiddle with. Things came to a head when my 2021-release work Thinkpad started tearing driving an external 4K monitor over Thunderbolt. I switched over to Wayland + sway, and it was a night and day difference. Significantly better latency and no tearing, even driving two 4K displays with an Intel iGPU!

The laptop docking story is also significantly better. Waylands libinput lets me assign mouse accel and keyboard repeat rates/rebinds to specific devices, and they always get applied on hotswap, no custom udev rules needed. For audio, Pipewire is sensible and Just Works. Kanshi works sensibly in a way that I was never able to convince dockd to.

Wayland is what brought me from “I like Linux for a work/coding environment, but it’s too fiddly for a daily driver” to finally consistently using Linux in a personal and professional capacity.

A while back, a colleague updated their system and screensharing broke. I went "ah I know this", switched their session back to X11, and the issues went away.

I think its' quite common to believe that a rewrite will be wonderful and then find out that it has its own problems.

The big problems are the social ones - I think the author is correct that allowing fragmentation is a big big mistake that cannot be recovered from and wastes everyone's time.

I'm also concerned about the "justify your usecase" issues. I know why people like to say that kind of thing when they have a philosophy that they don't want to toss out but to me it's never been a way to win people over to working on your system. Usually one cannot understand every single use case or at least understand how important it is to the people who want it. Making life difficult for others is not a recipe for success.

Short term inefficienty over debating over protocols is an obvious woe. But it's corresponding long term gains over protocols being far better threshed out, debated, and revised over time is basically invisible except by conjecture and at much grander time-scales.

Your comment here actually feels like what Wayland is good at: creating a development model where concerns & priorities can organically emerge, versus the anti-fragmentation model where there is one and only one canonical protocol or implementation:

> Usually one cannot understand every single use case or at least understand how important it is to the people who want it.

Compare versus what came before too. X dead ended itself in a lot of ways. The decentralized model seems to be working pretty well, albeit many people disagree, and many expected a faster more seamless changeover, and hang this albatross of expectations around Wayland's neck.

Most users will never care, but Wayland seems to be remarkably implementable. People kick out fun random idea implemntatjons with surprising regularity. This starting light seems like a huge win, to me, that enables a lot more ecosystem innovation that alternatives. You're just managing some OS DRM object buffers, via some protocols? Simple, nice. And then the other concerns later on, ok. This is a good gradient to me, and one that promoted a systems diversity health, that feeds forward, brilliantly. Being small & targeted & organically growing other layers keeps seeming like a huge win.

It's useless to me as I cannot share teams desktops (at least the last time I tried). Game over totally.

The idea that it's fun for some developers because they get to try things is of no use to users like me. I also question that fun if people are trying to get important features out and cannot because someone doesn't accept the use case.

Everyone has their one straw that just makes it totally unacceptable to them. And >50% of these straws seem to be that their ultra proprietary standard is a boatanchor against progress & isn't doing the good nice thing everyone in hopeful open source world has been doing for half a decade.

If you are going to doom & give up on Wayland or any idea because it's not popular enough, because it hasn't hit a critical overwhelming mass of support where the slow beleaguered giant corporations don't all support it, progress is impossible. We'll never meet your bar, change is impossible, & we shouldn't ever try doing anything but what we already did.

It has been around a long time and isn't usable for me - why should I expect it to suddenly change?

Wayland is the downfall of the Linux Desktop and the FOSS desktop ecosystem in general. It is actively developer hostile, overly complicated and the technical merits are questionable at best. It has the wrong philosophy about absolutely everything. Not only are simple and trivial things hard to do. Most things can't be done because standardized interfaces don't exist. The "process" of standardizing even the smallest features, no matter how essential through the "Wayland committee" takes literally multiple years.

What people really need is a "desktop protocol" not a "display protocol".

What is sad is that this is, unwittingly, a consolidation of unix desktop onto just 1 monoculture of Linux with Wayland should it become ubiquitous.

Since there is no network transparency, you won't have the ability to interoperate with graphical applications running on non-Linux machines either.

Meanwhile I was able to ssh -X into my FreeBSD desktop with a low end Linux laptop and run the same apps with the same settings etc. by using X11 which runs on both systems.

AIUI waypipe should actually be better for network transparency than X was. Of course, that requires that the other end is compatible and that the software runs on all the involved operating systems, which does seem to be a sticking point.

I've thought about this too but every time I've ever used remote X it's been terribly horribly slow. Do you really care? With RDP/VNC/others as alternatives, remote X has always seemed the most interesting but worst performing of them all.

I'm always cautious about the points made in blogs that have bold titles that don't accurately reflect the content, such as this one. In short, this blog title could be "My criticisms of Wayland."

That being said, I've migrated to Wayland this year because it has become more stable for what I run compared to Xorg, and from following the Wayland and Fedora projects closely, I feel it has recently reached a high level of development, and more importantly, development interest (both of which will likely cause it to supplant Xorg sooner than later). In short, Wayland definitely seems like the Way to go ;-)

The "development interest" in Wayland is still mostly carried by corporate full time paid developers. Wayland as such has had zero community mindshare in the FOSS ecosystem (until Hyprland came along but that is literally just one guy). And for a reason: The protocol is not suited for Desktops at all. It was developed for car entertainment systems and mobile phones. As a result it makes things that should be simple (like creating a window, printing a string, taking a screenshot) incredible hard for developers.

Both the Fedora and Asahi projects are stellar examples of community mindshare for Wayland in the FOSS ecosystem.

Just like GNOME!

(You know it's also GNOME devs behind it, judging purely from the feedback the author reported.)

> If you ask developers "why can't I do XYZ", you'll likely get an answer along the lines of "why do you need to do that" or "justify your usecase".

This is the difference between application programming and infrastructure programming. When doing app development, it's best to limit scope as much as possible in order to reduce risk. If you don't need it, toss it.

On the other hand, infrastructure only exists to reduce overall risk across all apps. It does this by accepting risk and dealing with the gnarly bits so apps don't have to. Put too much opinion into infrastructure, and that pushes that risk onto all of the apps. Often, the apps don't have the leverage to deal with that, so people stop using the infrastructure.

In this case, it's the "application programmer" who is asking the "infrastructure programmer".

Also, there's a difference between limiting scope and rejecting it. People are going to walk across the street whether you paint crosswalks or not. If your infrastructure fails to accommodate a present need, people will just work around/against it.

This is an interesting post and somewhat more technical and informed than many Wayland rants, however, it's worth knowing that in 2022, some of the problems that it describes have already been solved. For example, fractional scaling is now supported.

Which of the half dozen waylands support fractional scaling? This is a rhetorical question that you don't need to go research. It is just to highlight that there is not one wayland and any issue of feature support requires knowing which one you're talking about. And that itself is a problem.

I'm talking about the official protocols, to start with (as does the post).

"now supported." to me means you can actually use it so this seemed to be talking about the implementations. I guess you were using it in the theoretically supported sense? Anyway, it's hardly an official spec (like X.org is for X11s) when the implementations pick and chose.

It's as official as, say, the HTML 5 spec, or the HTTP spec.

The implementations can be plenty different as long as they conform.

It's not that I'm not getting your point. But part of the reason I wrote this rant is better than others, is because it's not harping on the point of "omg! someone dared to write a standard that has multiple implementations! the audacity!", but instead has criticisms of the standard itself.

I understand you don't like the standard-and-multiple-impls model, but it's worth noting that the most popular Wayland implementations share plenty of auxiliary and utility code in the form of libraries (e.g. libinput, libdisplay-info, etc). It's not as fragmented as it may seem, people are pretty reasonable.

Personally I think this is a healthy way to run things, for example because changes to the standard can easily be validated in multiple implementations, which tends to promote better quality for all over time.

The problem is precisely that getting the protocol official is just the start. After that you have to go wait for every different compositor to implement it, as opposed to implementing it once in Xorg and then every window manager gets it automatically for free with zero additional effort.

This is also how the web works, or many other things where you have a standards venue and multiple implementations that contribute to and validate the standard. There may be pros and cons, but I think overall it's proven to be a healthy way of running an ecosystem.

It's also not top-down, either. The Wayland governance model consists of voting members from the compositor authors, and for protocols to enter the wp or xdg namespaces, there must be multiple open source implementations. That means that by the time the protocol is official, at least first implementations already exist. It's not moving that slowly, overall.

Also, the Wayland model doesn't preclude compositors sharing as much of their code as they want. "Implement it ones and several projects benefit" also happens in the Wayland world, for example there are multiple compositors built on libraries like wlroots.

Xorg wasn't the only X11 implementations either. Fragmentation in the X Windows space did reduce over time as the open source impl sprinted and ahead and won out over everyone else by miles, but it did exist for many years. Most people are only familiar with an X11 space that already had 25 years of maturing and consolidation on it.

> This is also how the web works, or many other things where you have a standards venue and multiple implementations that contribute to and validate the standard. There may be pros and cons, but I think overall it's proven to be a healthy way of running an ecosystem.

The web only really has 3 engines (gecko/webkit/blink), and they still vary enough that it causes problems.

> Also, the Wayland model doesn't preclude compositors sharing as much of their code as they want. "Implement it ones and several projects benefit" also happens in the Wayland world, for example there are multiple compositors built on libraries like wlroots.

Yeah, I used to have great hopes for wlroots uniting basically everything except KDE and GNOME and mostly solving the fragmentation. Then I found some cool new compositor that I wanted to try, and it ran alright so I went to configure it to my keyboard layout (I could kind of get by on QWERTY to test that it ran but I'm not gonna live like that), and hey wlroots implements keyboard configuration so it should be easy right? Yeah, so it turns out wlroots provides the ability to support kb layouts, but every compositor has to actually wire up that code and this one didn't. And that's the most recent time I gave up on wayland.

> Xorg wasn't the only X11 implementations either. Fragmentation in the X Windows space did reduce over time as the open source impl sprinted and ahead and won out over everyone else by miles, but it did exist for many years.

Agreed; I've actually run into other X servers that were worse and I'm glad Xorg won (which is unusual; usually I prefer diverse implementations, but it turns out I prefer one working implementation).

> Most people are only familiar with an X11 space that already had 25 years of maturing and consolidation on it.

X is 39 years old[1]. Wayland is 14 years old[0]. If it's going to take another 11 years for Wayland to catch up I'm going to just wait for Arcan or whatever replaces Wayland.

[0] https://en.wikipedia.org/wiki/Wayland_(protocol)

[1] https://en.wikipedia.org/wiki/X_Window_System

X11 does not have fractional scaling at all, and it wouldn't be possible without both window manager and client involvement.

People will standardize the wayland implementation, just like they standardized the X11 one.

To tell you the truth, I'm not even aware if this is underway, or whether the computer I'm using right now uses some flavor of x11 or of wayland. But you can be certain that if every distro adopts it, every distro will have the same flavor. That problem is for early adopters only.

KDE/KWin does.

Ultimately, Wayland is where all the energy is. It has shortcomings but is good enough and all future work (HDR for example) is going to be done on Wayland.

X11 works.. when noone actually uses X11, but actually some wrapper API that was made alongside the development of X11, or suffered through effectively replaying history. Some fun things I recall encountering while trying to use it:

Want to detect a left click? Sure! How about the back/forward buttons on the mouse? nope, not in the API; well, if you try pressing them you'll see them going to mouse button constants 8 & 9, but I didn't find that in any documentation.

Want to get scrolling events? There are mouse press things for that (even left/right!). Want to also handle touchpads which may have fractional scrolls? Well, there is some confusing "valuator" stuff, and you can take a delta between two valuator event floats.. but after using it for a while, you'll notice that the app will like randomly scroll by large distances. Turns out... that valuator value is a globally accumulated value, so you have to hack in a thing to "reset" the "calibration" of that any time the mouse exits a window (try it - move mouse out & in a window, scroll one tick, repeat - you'll never scroll anywhere).

For text input, you can thankfully call a thing that'll do XCompose & most key handling for you, but to handle xmodmap I believe you have to write the mapping functionality yourself; didn't bother. Then there's some stuff about how in some context a window was not focusable through some specific method for some undetermined amount of time after being made visible, but there might be some reason for that (never bothered to figure it out).

Now, I don't know if wayland does any better on any of this, but I sure hope it does, as it does seem to at least make an attempt at being usable. Maybe wayland will go the same route, with a middle-man library that handles all the DE-specific stuff. There's already some of that for window decorations.

> Want to detect a left click? Sure! How about the back/forward buttons on the mouse? nope, not in the API; well, they are on some mouse press constants 8 & 9 if you dig around enough.

Yes? They're just more buttons; how would you expect them to work?

I'd expect them to be documented somewhere, so that I could be sure that they're not gonna be swapped or map to some other buttons (and not have to find out about it through trial/error or blogs/reading GTK's source). edit: edited my original post to be more clear

How in the world is X supposed to know what USB event your mouse sends when you click the button on the side? Did you ask the mouse manufacturer for documentation?

I don't know where that information comes from (I'd assume the kernel? wayland[0] refers to the kernel[1]), but surely if I, and every GUI app that wants to use the mouse, is supposed to know it, X definitely could too. If I recall correctly, the left/middle/right button values were documented (to 1/2/3, which doesn't match the linux header ones), so surely more could.

[0]: https://wayland.app/protocols/wayland#wl_pointer:event:butto...

[1]: https://github.com/torvalds/linux/blob/4a0fc73da97efd23a383c...

I'm not sure I understand the complaint.

Anytime I've ever had to map extra mouse buttons I've always had to do a discovery step to figure out what "button number" any specific mouse button mapped to.

Is that for back/forward, or extra buttons on top of those? Back/forward have pretty standard meanings & usages so I'd expect them to work out of the box (and they have for me). Can't comment much on other extra buttons though, quite possible those don't really have a good way of being handled.

I've never seen forward/back buttons that didn't involve the driver making interpretations.

Extra buttons starting at 8 is due to the mouse wheel taking 4-7 (both horizonntal and vertical scrolling are represented as a button)

https://www.x.org/releases/current/doc/man/man4/mousedrv.4.x...

"""

Option "ButtonMapping" "N1 N2 [...]"

Specifies how physical mouse buttons are mapped to logical buttons. Physical button 1 is mapped to logical button N1, physical button 2 to N2, and so forth. This enables the use of physical buttons that are obscured by ZAxisMapping. Default: "1 2 3 8 9 10 ...".

"""

Right, I had seen that 4-7 was scrolling, but that's still not describing which of the numbers greater than 7 will map to the back/forward buttons.

Is there an actual standard for mice at that point? I.e. have mice manufacturers ever written down what signals are backwards/forwards? The Logitech G600 has 20 extra buttons, and no backwards/forwards button.

I think it's more convention, as the applications are ultimately what are deciding to do with the mouse button. There's nothing to say that they could not put a context menu on the left click.

In a sibling comment I noted wayland referencing some kernel values[0]; and I'd hope there would be some reasonable standard. I certainly didn't have to manually configure chrome & firefox for them to know which buttons on my mouse to map to previous/next page. Maybe the more exotic mouses would need some, but I'd still expect in such case that the user is able to say "ok this button is 'back'" to their DE, and not necessarily be forced to configure that for every single application individually.

[0]: https://github.com/torvalds/linux/blob/4a0fc73da97efd23a383c...

Which is what the ButtonMapping config is for :)

That said, I do think it would be nice to have written down that 8/9 are commonly used as backwards/forwards. They are just not explicitly backward/forward buttons though.

It seems like wayland got a lot closer to generally usable in the last few years, but it's still not for everyone. Some of this is (intentionally) limited design, some of this is lack of people caring to migrate their software to work with wayland.

I was looking to switch to wayland in my VM after noticing that it does have much lower cpu utilization, and better compositing behavior, at the same time. Yet, can't do it. right or wrong, wayland's stance on security before convenience on the whole clipboard access thing means I cannot copy-paste between VM and host, because those vm tools do not support wayland, and I don't know when they will.

Also, I noticed that people who prefer obscure window managers or desktop environments are mostly in a bad spot, as the amount of WMs that support wayland isn't much.

Clipboard sharing works when using qemu & SPICE: <https://www.spice-space.org>

Guest needs to have spice-guest-tools installed and probably be using gnome - I have had no success with other environments but my testing was shallow.

Alternatively, you can share the host's wayland socket with the guest over the network via waypipe: <https://gitlab.freedesktop.org/mstoeckl/waypipe>

In this scenario the clipboard will function without any issues and with no further steps. And wouldn't require a compositor running in the guest.

You can also have a wayland connection between the guest and host through shared memory buffers via virtio-gpu. But this requires crosvm as qemu doesn't have this functionality. Guest kernel needs to be >=5.16 and be compiled with CONFIG_DRM_VIRTIO_GPU and this tool is required: <https://github.com/talex5/wayland-proxy-virtwl>

FYI: Wayland socket forwarding via the virtio-gpu device is going to get merged soon into upstream QEMU. The patch series is at qemu-devel right now. I have tested it, and works smoothly for 2D stuff. GPU rendering will be added in the future.

Can you elaborate on this? Are you saying that this obviates the need for the proxy tool above?

The network one? No, it'd just make the guest able to show windows in the host's compositor without running one inside the guest.

> Guest needs to have spice-guest-tools installed and probably be using gnome - I have had no success with other environments but my testing was shallow.

Well that's just a nonstarter.

(The waypipe thing is interesting, though)

The only thing that makes me go back to Xorg is Wayland not supporting custom notification for apps.

I use Telegram Desktop a lot and I want their custom notifications, not the ugly thing that pops up on Wayland which doesn't even take me to Telegram when I click it.

You use a seperate program for that. I use mako for notifications.

I'll look into it. Thanks.

Wayland security model fundamentally fractures automation/accessibility software. Accessibility tools need access to inspecting and injecting into other windows without root. For example, reading window properties and injecting virtual peripheral input like key/mouse. I'm hoping this has changed.

I know complaining about a site's layout and color isn't useful, but this is unreadable without reader mode for me.

The reader mode button works. Do you not use it frequently? It will make you happier.

I don't use a web browser that has a reader button whatsoever. Plenty of people surely is in the same situation. It's a bit rich to assume everybody has it.

Accessibility and readability, even for the sadness of the advocates of "content density", should come first.

Yeah, he complains about Wayland design choices that hurt usability and then puts grey text on a black background. The irony is interesting enough to warrant the comment.

Is copying fixed? Last I checked it seems you cannot copy-paste in many cases with wayland. There are clipboard managers for wayland that solve this but not entirely (such as in a vm, keeping formating, other edge cases IIRC). Hearing about this issue has made my never consider switching to wayland. Is this still the case?

It worked fine for me four months ago, but now I can't copy from Chrome to any other app (or vice versa) again. :-/

Guess I will check back in another 5 years ¯\_(ツ)_/¯

I have a small mini-rant about copy/paste in modern software.

copy/paste was a solved problem 20 years ago, yet nowadays it's always a crapshot as to whether or not you copied what you think you did, etc.

I've had websites insert non-printing characters into text and it will take forever to find it.

What you ran into is possibly the "security" behavior of Wayland regarding the clipboard. Basically the clipboard is attached to your focus (or mouse?) and only the active window is allowed to read the clipboard.

…which completely breaks any workflow that relies on observing the clipboard from a second application while using another…

Nope! Not in my VM. vm tools dont support wayland copy, nor screen resize. Works in X

I'm not conversant with Wayland technical intricacies. I use Plasma; Wayland works OK for me and in many respects feels like an improvement over X11, but I still have two gripes:

- Last time I checked, I could not share my screen over Zoom. This could have been fixed now. (I know, it's a proprietary application but still...)

- My favorite screenshot application (KSnip) still has reduced functionality over Wayland.

I think these issues are representative of many users' experiences, that is, things usually work as usual or improved, but there are some corner cases where there are annoyances. Let's hope they will get sorted out.

So we are still fighting over a desktop windowing system and reviving the Xorg vs Wayland battle.

Meanwhile the winners are the users who chose not to play this game with their machines or rant about how a basic app does not run on Wayland and runs on Xorg or vice versa and instead use either a Mac or Windows machine with standard defaults.

If you're app glitches on either windowing system and you continue to rant because you need to tweak the dotfiles, systemd services or dive into config files, then you might as well use WSL2 instead.

At least Windows has a much more pleasant desktop environment.

You are absolutely correct. This is why it is absolutely reasonable to assume that Wayland is actually a sabotage project.

X11 was far ahead of anything out there when we had compiz with spinning desktop cubes and wobbly windows running absolutely smooth even with the most basic GPU in 2006(!). Something had to be done about that.

Sabotage project? X was "smooth" in 2006, it hasn't moved along. Mixed framerates and resolutions are horrid to use. Performance issues appear with just one 4K screen. High-framerate, FreeSync or high bit depth displays are basically a waste with X.

The only thing X (and KDE) seems to have gotten right is mixed (and fractional) scaling, Windows can't do even decimal scaling properly within just the core UI.

Xorg has worked fine for me since I started using it in 2008.

The only problem I ever had was the poorly packaged proprietary ATI/AMD driver, and now we have AMDGPU.

I just set my refresh rate and enable TearFree (via xrandr), and I'm set.

---

Windows is the desktop that constantly interrupts my life by either breaking or running straight-up malware. And then it gives me pointless notifications I'm not allowed to disable (group policy) and bullshit ads I would never ask for.

If you really believe the Windows desktop is pleasant, you must be really easy to please, and really difficult to annoy.

> a basic app does not run on Wayland and runs on Xorg or vice versa Can you even give an example of the vice versa case?

I've never not had things work in X, so your idea about going over to 2 user-hostile OSs to solve a non-issue doesn't resonate with me.

> At least Windows has a much more pleasant desktop environment.

Boy oh boy, that's not what people are saying after using windows 11. The entire reason to be on linux for me is to recreate & extend the best of what windows UX was, but windows UX is past its glory days and into active sabotage.

I really hope Wayland won't replace Xorg anytime soon, because it still hangs my machine when I'm trying to run it. I would hate to be forced to move to macOS.

I’m sure there are enough people to support the existing setup when that happens.

I'm one of the people who does not comment on these matters because things "just work" on my machine, but I'd like to say my thoughts as well. Wayland is fine. I still am using i3wm (X11) currently, but Wayland works perfectly fine. There are a lot of things going in the right direction, and I consider Wayland clearly superior. There is absolutely nothing holding me back from switching. Yes, there are minor inconveniences like the lack of a hardware accelerated cursor. Sometimes my compositor crashes on some very old programs that still use GTK2 or some old framework for some reason, but everything outside of that is completely stable and usable. If I don't know something, I ask or browse the wiki, search up alternatives for existing X11 programs, and try to fix stuff myself. I am using the proprietary NVIDIA drivers and it works. I don't have multiple monitors, my resolution is 2560x1440. Things are fine, and they will only get better.

I've been using Wayland on Manjaro/Sway for the last 6 months. Zero issues beyond figuring out the new way of doing things.

As a former X11 application developer I will say that X11 should have been put out to pasture a long time ago. It's been moribund since the 00s and the quicker that distributions move to Wayland only, the quicker Wayland will improve.

Started using wayland a year ago with sway. A was a bit frightened at first, but in the end everything worked. From firefox, to steam and their games.

The only "drawback" was that when I launched windows (ex: wireshark) at root, it didn't work and I had to previously allow "root" to open windows on my login (with xhost). Thought it was quite neat

What? What does xhost have to do with Wayland?

Wayland seems to be slightly more performant but Firefox, Steam and games are the worst example.

Steam doesn't seem to scale properly. Firefox struggles with native Wayland, especially if you also want GPU acceleration (same with Chromium though). Games are a mixed bag. A pure X environment didn't have these issues in my experience (but it obviously has others).

Would like to read a reply to this article by one of the main Wayland devs

I wanted to implement the detachable tabs in my application, similar to that of chrome/sublime text. And I learned that this is not possible with wayland, because wayland doesn't give you the mouse's global position (for dragging and dropping, I need to know if the cursor is hovering a foreign window).

Reading this produces the same lump in the throat as the arch forums. The phallosophers want the user to disappear and finally let the computers display their beautiful art.

Why does Linux desktop need saving? I don't understand; been using Linux as daily driver for the past 20+ years. It's getting better every year.

2023 and Linux is still struggling to achieve the same level of basic functionality that windows and mac have had for decades now.

YMMV but I've had far better experiences on Linux as far back as 10 years ago then I've had on Mac or Windows even today.

Windows, at least, seems to be getting worse by the day, and it was already pretty bad 10 years ago.

Mac at least seems to not be getting worse.

Speaking on the graphical stack specifically windows is a great experience and has been since drivers moved to user space somewhere around windows 7.

(On other areas and the overall experience I might agree. WSL2 was a big step forward though)

Interesting, for me Windows keeps getting better - additions of virtual desktops and windows (windows of applications, not system itself) management became much better - things like automatic layout restore on attaching external monitor to laptop and so on are very welcome.

And vice versa I'd say. Windows rolled out the WSL for a reason, there are many workflows that Linux is best for

Programming on Windows is still a huge PITA.

I'm not sure what Linux is supposed to be struggling with.

I've been using Linux for my desktop for 25 years now and every time I try Windows or Mac, I come away disappointed; these are not professional operating systems, they are hobbyists playthings, I discover every time and time again.

So I guess; YMMV?

Can you not read the article?

I can and I did.

I don't agree with it. I don't think it makes a good argument and it reads more like an angry rant than anything else.

No idea what you are talking about, I find Windows and Macs absolutely horrible to use. To me it seems that they only make their OSs to sell me more of their stuff...

Apple decided to just never even try and do fractional scaling.

I used to use and love Apple. Then I discovered the modern Linux desktop and never looked back.

> The best statistics we have on this come from the February 2022 phoronix article on telemetry data from firefox users. It found that less than 10% of Firefox users on Linux are using Wayland in any form.

I'd guess those numbers changed when Wayland became the default for Ubuntu 22.04, which was released in April 2022. Does anyone know of newer data?

Tried Wayland, has horrible fractional scaling support (Not that X11 is much better) and went back to Windows. If the Linux guys still can't make high-dpi displays work after them being the norm for 10 years I'm not sure it will ever happen.

Pretty much my experience. WSL2 has my Linux itch covered, and I can create multiple VMs easily (for different distro).

It's not like Windows can handle even decimal scaling. It's trivial to stumble upon horribly scaled text, even in Microsoft's own software (try opening the event log or group policy editor with scaling not being 100%).

Both macOS and Linux (X and Wayland both, actually) do scaling nicer on average.

Really? As soon as I turned on fractional scaling on Wayland, Chrome & Electron apps became blurry looking to the point of being unusable. Actually, a lot of apps I'm used to using were unusably blurry and horrible looking (like Jetbrains products). Maybe Windows doesn't handle decimal scaling, but at least most everything looks and works fine. You have to look really deep to find stuff that doesn't work (like event log or group policy editor, leftovers from a bygone era Windows, things I don't use day to day, and I'm sure nobody uses them every day). As it is now, as a developer, using a browser and Jetbrains IDE or VS Code, Wayland is completely unusable and makes the whole Linux space feel more like a toy than a professional tool.

So I don't know what decimal scaling you're talking about when Wayland can't even do 125% or 150% scaling without half the app ecosystem breaking.

> Maybe Windows doesn't handle decimal scaling, but at least most everything looks and works fine.

It doesn't handle even decimal I don't dare try fractional. I guess our standards differ, it's not fine.

> You have to look really deep to find stuff that doesn't work (like event log or group policy editor, leftovers from a bygone era Windows, things I don't use day to day, and I'm sure nobody uses them every day)

It really doesn't take much to stumble upon things like that when you use Windows on a daily basis. I wouldn't call system menus "leftovers from a bygone era".

If you restrict yourself to only the parts of UI that mostly scale well, of course then scaling works well. But that's silly.

> As it is now, as a developer, using a browser and Jetbrains IDE or VS Code, Wayland is completely unusable and makes the whole Linux space feel more like a toy than a professional tool.

Neither JB IDEs or VS Code are native Wayland applications. Though I can't say I've had issue with JetBrains' software under Wayland's scaling.

I’m just waiting for nvidia drivers to be supported fully, for working with blender. Iirc some work was being done there, but I don’t follow it closely after some wayland dev(s) seemed hostile to nvidia users. .. somewhat understandably, but still.

My killer app for X11 is XScreenSaver. Wayland can't run it, so forget Wayland.

People who think there can/should only be "1 linux desktop", completely misunderstood what linux is all about

It is about the experience you craft for yourself, my desktop is different than yours, it is supposed to be like that, and is the reason why linux scale from microcontrollers, to game consoles, to datacenters

These people are out of touch

it's not clear to me whether you're saying Wayland devs are out of touch, or Wayland critics are out of touch.

I don't think there should be "1 linux desktop" but I do believe the Linux desktop is far too fragmented. Yeah, it's nice that we can customize stuff but it makes it a nightmare for developers to support us.

How can anybody distribute complicated software that is supposed to support a wide range of kernels, X11 + Wayland, open + closed source drivers, several desktop environments, etc? To make matters worse there's endless combinatorics in the versions of those packages because nobody can agree on an upgrade strategy.

Maybe we could demand support if we had more users but Linux represents a small fraction of the desktop market.

No wonder I got downvoted

This was shared just recently on HN and it perfectly describe the issue:

https://notes.volution.ro/v1/2023/09/remarks/64299f31/

I have some more, perhaps it is time for me to start blog posting.. we shall see

I thought Wayland was supposed to save the Linux Phone, not the Linux Desktop?

the only think stopping me using is the unsupported nightshift / redshift / KDE night colour stuff

https://sr.ht/~kennylevinsen/wlsunset/

thank you. will try this.

Don't even get me started on this. Yes, there are 'redshift' programs for wayland, but last I tried, they were all so new that almost none of them were in the deb repos. So I install arch, thinking I'm finally going to have a nice wayland experience, but arch is a no go to me. Ironically I find the arch devs about as arrogant as the wayland devs.

Ironically I find the arch devs about as arrogant as the wayland devs

predictable.

Another thing which has only just started working for me on wayland is "flameshot" for screen grabs -- very handy.

Why was there a need to reinvent the wheel when this effort could have been used to improve X11/Xorg instead?

Reply on news.ycombinator.com