I enabled wayland recently and there’s been a lot of annoyances to the point that I wonder why I bother considering X11 works without these issues. For example:
* Chrome needs a bunch of extra flags to launch in Wayland mode
* Firefox needs extra flags to launch in Wayland mode
* Xwayland is just broken on Nvidia with both sides refusing to compromise (implicit vs explicit synchronization - Nvidia refuses to add implicit and Xwayland refuses to take Nvidia’s patch to do explicit). What this means is that you get tearing, flickering and all sorts of terrible graphical artifacts
* Chrome only just fixed HW acceleration for Nvidia (latest m120 beta)
* enabling vulkan causes Chrome to fail to render (although this may just me needing to try reinstalling Nvidia drivers)
* I thought it was an acceleration thing, but even with HW acceleration fixed Chrome has a bug where the mouse pointer leaves behind white speckling when mousing over on a dark background. VSCode doesn’t have this issue.
None of these issues appear in X and this is from someone who thought Wayland is the right way to go (eg you didn’t see this kind of story on Mac when they switched to HW accelerated compositing).
Maybe Firefox needing flags is a per-distro thing? I’ll have to check and make sure but I think Firefox works with Waylad sessions out of the box on my Fedora laptop. That’s using an Intel iGPU instead of an Nvidia GPU though so maybe that’s also a factor.
I think that's right. To make it work on Nvidia you need to go through extra hoops in a bunch of ways (turning on config flags, installing alternative HW decoding libraries etc). Part of this story is definitely Nvidia & Wayland teams being at ends with each other for some reason so anyone on Intel & AMD GPUs probably don't see a lot of this which is kind of surprising because 10 years ago Nvidia was the only one with well behaving drivers on Linux.
I'm using sway on an Nvidia card and apart from having to use sway-nvidia (which does set some environment variables for wayland with nvidia) I don't have to set any flags for Firefox.
On my laptop with Intel graphics I didn't have to use any special flags for Firefox on wayland for several years.
One thing is that the correct environment variables should be set, but all the wayland compositors have been setting these for ages AFAIK
Seriously though, it's been a while since I've had serious issues with AMD cards on Linux. I was very annoyed with the state of my rx5700 when it first launched back in 2019. We've come a long way
Yeah the typical AMD experience on linux is about as good as should be expected. It’s a shame that there’s such a complex mix of potential problems when it comes to using AMD cards for GPU passthrough after needing to reset/reassign them.
I have three laptops each with Intel graphics, each from a different generation. They all have the same show stopping (as in screen locking gotta reset) bugs after a fresh distro install, no adjustments.
Wayland has been flawless for me, but for others not so much.
Every thread about wayland has people claiming it's broken yet they aren't reporting bugs or providing useful information at all, just vague complaints.
Many people experience no issues with wayland, and some people do experience issues, but without any information nothing can be done to even attempt to fix things. It's possible in many cases that something is misconfigured, or a known issue that has been fixed in newer releases of whatever you are using, or a genuine bug that hasn't yet been discovered. Nobody will ever know though.
Basically your complaint is “Linux users aren’t spending hours to reproduce every single issue and keep following up with additional debug info from every developer new to the conversation”. What if instead the model was done survey who has issues and pick 1 random person who’s reporting issues and build up a knowledge base of the ways it’s failing (ie formally “here’s the symptoms, here’s the root cause and here’s the fix” whether that fix is something the user has to do on their end or a new SW version. I think if you did that enough times, you’d be randomly sampling the highest effect bug rates. It’s not the cheapest way to do this but if you’re having so many foundational issues with something as it’s looking like Nvidia is with Wayland (especially given how old Wayland is), that’s what I’d be telling the Linux driver team at Nvidia to own end to end until we believed all the largest Nvidia issues were resolved (whether or not it’s a driver problem) and then go back to normal triage processes. This stuff is normal shit that happens at high performing teams when things are this bad (eg an entire component critical to Wayland working on a Linux desktop isn’t working if you’re an Nvidia customer).
Asking users to contribute to fixing their own issues is the quickest way to make sure nothing ever gets done. I’m an extremely technical user but im using Linux for my day to day engineering needs. I have other SW that needs my attention and debugging my personal desktop issues is not what I’d like to be doing. If it gets bad enough I’ll reinstall the OS from scratch or switch back to Mac and figure out how I can make Linux work in a VM so I can test the software I need. But right now the SW I’m working on is Linux-only and if I can solve and figure out a workaround for the problem by myself without having to pause and follow up about some bug is not where I want to spend my energy. I have enough trouble keeping track of conversations on GitHub. I just hope any bugs I can’t figure out myself will eventually get solved without me having to do anything. And I do follow up some times but the value of that has been hit and miss enough that I do it as rarely as possible.
I did. It was a base distro install just booted up, as I already stated.l, on three different, bog standard machines. They were not interested either.
I take my bug reports a bit more seriously, and I don't use instances of success to discount instances of failure. A bug's a bug, a good proces cares about outliers (and as you say, Wayland erroring does not seem like an outlier) too.
I understand not all combinations of hardware can work flawlessly, but Intel machines with Intel hardware seems like a good baseline to ensure always works before you mainline Wayland.
Maybe if you told us, how to replicate this show stopping bug so we could it try ourselves, someone could fix it then. Wayland being the problem or not.
Not the OP, but happens to me too. Steps to replicate: put your laptop to sleep (suspend) and try to wake it up (resume). You end up with a black screen. Often times, you can SSH into your laptop, but the screen remains unresponsive. Only hard reset helps.
There are many reports of such bug. It's common on Wayland, but happens on X11 to some folks. The only constant is it happens on devices with Nvidia cards.
I tried pretty much everything and I gave up on solving it.
I didn't even have to do anything. Just boot after a fresh install and wait a minute or two. Three generation of all Intel laptops. I uploaded some logs, but people were utterly uninterested or hostile.
It has nothing to do with Wayland though, based on the description. Wayland used the base kernel API to manage the display. That is probably broken on your given device.
X11 often gets patched by a proprietary nvidia blob (the nvidia driver), which uses a separate, non-standard, X-specific API.
I'm curious, because I have machines with Intel GPUs from different vendors and of various generations (from Ivy Bridge, through Kaby Lake up to the Raptor Lake) and I've never seen such a bug.
I was too, but I'm not an expert and at the end of the day I need to get work done. It's up to Wayland devs to care about covering the 1001 edge cases (and thus have a process for user reports) or not.
If the device locks up, is that really Wayland, not the kernel driver or mesa driver?
Here on HN I see a lot of popular opinions on Intel and their drivers. Even today, proposing it as a baseline for Wayland. But could it perhaps be that Intel is just really good at marketing? AMD might be less about marketing and just delivering good drivers.
They're pushing the full-fat GSP firmware (all 60-odd megabytes of it) to linux-firmware soon, and allowing nouveau to call into it for reclocking support, so hopefully things will be less terrible for 20xx users and up.
Are you using Wayland? I'm curious since it seems like most people I've discussed with have confirmed that the 545 drivers improved things for them. I thought they might actually have fixed/improved the experience for most Wayland users but I guess things are never that simple!
I have heard these reports too but I haven’t noticed any differences in behavior by driver version. All the bugs I mention are the same ones before and after the latest beta that everyone said improved things. I imagine they fixed some particularly visible problem although I haven’t seen it. Maybe most people are on laptops and their laptop chip got some fix for an issue impacting those users? Not sure but that’s one obvious difference that might separate me. The other is that my CPU is the 13900k without an Intel GPU and I imagine most people opt for iGPU integrated (I didn’t see the point of it when running a desktop and I had enough of bad experience with Intel+Nvidia on the laptop side of things for the past 2 years with that configuration that I decided my sanity was better this way).
Yes, wayland. With driver 545 as soon as my screen goes black (due to inactivity), it never goes out of that state anymore and I have to do a hard reboot.
Another regression is that KeePassX/C AutoType doesn't work with Wayland, so now instead of a simple CTRL+V in KeePassXC, I have to separately copy and paste the user and the pass.
I'm another X holdout because of random issues like this. X just works for me, and I don't have the time and energy to care.
I try to test my system setups with qemu, but I cannot for the life of me figure out which combination of whats I need to get wayland to run there. There's apparently two? framebuffer implementations which work with none of the desktops I want to use. All the documentation points to other projects that should be documenting X or Y, but don't.
It's possible some of these things are really distro+hw related, but for the sake of argument:
> * Chrome needs a bunch of extra flags to launch in Wayland mode
I'd rather have to worry about adding a few extra flags than worry about X11. X11 is overcomplicated, over-engineered, barely maintained, is apparently broken from a security angle, ancient...
Till today I have found X11 configuration one of the toughest things in Linux. I'd rather google for some wayland specific flags than worry about X11.
> * Firefox needs extra flags to launch in Wayland mode
See argument above for Chrome.
> * Xwayland is just broken on Nvidia with both sides refusing to compromise (implicit vs explicit synchronization - Nvidia refuses to add implicit and Xwayland refuses to take Nvidia’s patch to do explicit). What this means is that you get tearing, flickering and all sorts of terrible graphical artifacts
Xwayland is a transitional thing anyways too. When X is dead we won't need to worry about Xwayland either.
> * Chrome only just fixed HW acceleration for Nvidia (latest m120 beta)
One more reason got added to use Wayland from now onwards.
> * enabling vulkan causes Chrome to fail to render (although this may just me needing to try reinstalling Nvidia drivers)
Vulkan unlike OpenGL is also quite new. Give it a couple of years for every combination to get ironed out. X11 is hardly the answer here. If you're going to gripe about Wayland, X11 is never the answer.
I don't wholly disagree with your conclusion, but it's worth considering some context.
Direct3D versions 1 – 9 were indeed released at a rapid pace, more than one per year on average. However, it took a whole five years before D3D9 was superseded by D3D10, and subsequent releases have also been relatively slow.
Direct3D 12 is eight years old, even older than Vulkan. However, it still feels to me like a "new" API, I think because even now, many games still use DirectX 11.
I'm not sure why this is. Probably some combination of the flattening of the technology curve (hardware used to change more each year than they do today), and the fact that once a technology is more complex, it takes more time to change it, and to adopt those changes. And also, legacy compatibility becomes a concern: in 1995, we didn't have multiple decades of existing code, libraries, tooling, and complete consumer software to migrate and support.
DX9 remained on the frontline for as long as it did because Windows XP was supported from 2001 all the way to 2014 due to usage share. Similarly, DX11 remains on the frontline because Windows 7. I guess it's more apt to say they are because Windows Vista and Windows 8(.1) failed spectacularly respectively, but my point remains.
Since Windows 7 and 8(.1) are both EOL now, more games should start using DX12 Ultimate.
Isn't dx12 a lower level API? Not nearly as accessible as dx/opengl. I guess not everyone in the game industry has the ability to just switch because codebases are messy, "things are already working", and there not being much to gain but performance.
Deus Ex: Mankind Divided, for example, would often crash on me under dx12, until I switched to dx11. Not everyone is id Software or Epic.
> 7 and 15 years should be more than enough time to create stable and robust software libraries
Yeah, if it were goddamn Apple forcing the ecosystem to make the change. Linux is the bazaar, no one forces anything, but it also comes with disadvantages, like not adapting new tech.
> Till today I have found X11 configuration one of the toughest things in Linux. I'd rather google for some wayland specific flags than worry about X11.
My experience here is that this used to be true more than ten years ago. At some point in between that though X11 managed to get to 0 config. At least, I installed Arch & didn't have to touch the X config at all and things just worked (on NVidia at least).
I'm not trying to say needing flags is something bad. That stuff will get ironed out for sure. But if you want to avoid Xwayland, you need flags (& it's not just Chrome & Firefox - Electron needs similar things) AND Xwayland on Nvidia is a mess right now with no end in sight.
> Xwayland is a transitional thing anyways too. When X is dead we won't need to worry about Xwayland either.
Some apps don't even support Wayland and likely won't for a long time and I think you're underestimating how long XWayland will be a thing (e.g. Tcl/TK which is powers a surprising number of developer tooling UIs doesn't have Wayland & doesn't seem to have enough interest to really push this to production though previous ports have been attempted & made progress). FWIW, this argument has been made since Wayland started 15 years ago. I won't even have the same machine in 5 years.
> X11 is overcomplicated, over-engineered, barely maintained, is apparently broken from a security angle, ancient...
It's funny the kind of arguments that get made because you think I'm against your position. I agree with you and think Wayland should replace X11. But there are very real problems with it today & there's a reason anyone with an NVidia GPU goes "uggh. Wayland". The problem is not just on Nvidia's end and I see cultural problems from Wayland's side too (likely because they've faced so much drive by attacks). I feel like Nvidia is stepping up on their side though - I don't see the same from Wayland/XWayland camp.
> Nvidia is stepping up on their side though - I don't see the same from Wayland/XWayland camp
Genuinely curious -- what stepping up should happen from Wayland/XWayland ? Intel and AMD seem to work well on it so is it because they have some accepted some ugly hacks that Nvidia are not willing to do ? Something else ?
In terms of stepping up, they seem to be tackling through Wayland issues. The latest 545 seems to have people on forums saying “this fixed so many Wayland issues for me”. Now either they are paying professional astrosurfers, or they legitimately fixed some classes of bugs (or it’s random). My experience though is that this kind of feedback tends to be genuine. So I’m giving kudos where it’s due and it seems like the Nvidia team is starting to focus on Wayland issues. Similarly, I was observing recent bug tracker activity on Chrome and Nvidia and all that seemed to be tracking bugs on their end (which not surprising - Nvidia and Google have strong relationships between Android, Chrome, and formerly Stadia).
I don’t know the details on the Xwayland BS (which is the main visible issue I’m seeing as unhealthy). I have not done the technical research to fully understand whose position is actually justified. Heck it might be that both sides are correct on the facts (it doesn’t make sense to add the support that makes other GPUs work and it doesn’t make sense yet to add the new way for Nvidia) but the net outcome is that Xwayland on Nvidia’s propietary drivers remains broken for the foreseeable future. Both Intel and AMD both support something called implicit synchronization. But apparently explicit synchronization is “the future” because it enables more performance and Nvidia has clearly communicated they’re not going to support implicit. Xwayland only supports implicit synchronization and refuses to add explicit synchronization (I think strong arming their position is that the explicit synchronization protocol is isn’t ready or not popular enough yet in the ecosystem? I’m not sure). Nvidia has contributed some patch to fix Xwayland to use explicit synchronization when running on the propietary driver (or something like that).
> Because RedHat took over the maintenance and they want to push Wayland. They took it over to kill it.
Can we stop with these useless conteos? No one goddamn cares about linux desktop, this is the unfortunate state of things. X could be maintained, but no one goddamn cares about doing that ugly work, because they don’t get paid Cobol-bank money for maintaining an ugly legacy software that was designed at a time when GPUs didn’t exist.
RedHat is just a bad maintainer because they have a commercial interest in wayland.
They've done the same to CentOS of course - just taking it over to kill it because it's in their way of selling paid licenses.
I'm surprised how much support they have been getting from the community. They managed to frame perfectly normal open source users as "freeloaders" during the whole CentOS and closing source debacle surrounding RHEL. And the community seems to be happy parrotting this twisted narrative. The whole reason for FOSS is to share.
It's one of the main reasons I moved to FreeBSD, I don't want my OS to be a stringpuppet for big tech.
It's been kind of buried, but KDE has the first steps of tiling mode built into it now, in addition to the default screen edge tiling (on KDE 5.27.8 on up). Meta + T to enable tile editor. Shift + window drag to move a window into a tile.
Hopefully a lot more coming, including different layouts per virtual desktop & full keyboard nav for the tiles. But it's a decent start.
SwayWM at least enables some use of mouse for editing the layout: Meta+left drag to re-arrange/group/ungroup/etc. tiles, for example, and of course, Meta+right drag to resize as in i3wm. To me, even just having the Meta+left drag option seems like a significant step up.
I really wanted to try plasma with i3 but the big problem I ran into was that different monitors can't have independent workspaces, which was a deal breaker :(
The advantages of tiling window managers are so tied to muscle memory, I’m not sure many people will want to switch from, say, sway or i3 or whatever. But at least new people will be able to give it a try.
I’m not sure how much work is involved in replacing the window manager inside a desktop environment. It would be nice to just drop sway into kde. Dunno if that is possible though.
> The advantages of tiling window managers are so tied to muscle memory, I’m not sure many people will want to switch from, say, sway or i3 or whatever.
Hear, hear !
I have been stucked with awesome for years now and even though I lost some functionalities between v2 and v4 I still can't switch. I tried i3 and regolith (was really looking for all the niceties of a DE: volume, bluetooth, screen, etc.) but tiling operations are not the same.
I wasn't exactly an awesomewm power user, so I'm sure there's awesomewm functionality you couldn't mimic in kde, but my most used keyboard shortcuts I was able to import in to kde.
I haven't tried KDE tiling out yet but tiling is not just about shortcuts, it's also about tile creation patterns, tags/virtual screens (dynamic or not) and of course default settings. I am not an awesome power user either, mostly basic settings plus dynamic tag management (empty tag are deleted automatically, new tag is created when switching to next tag, number of tag start at 1). The rest is just shortcut remapping and getting a systray with audio and network applets. No application based tag, no custom script to display bitcoin trading or media player widgets.
My only problem with KDE is the lack of a feature that Awesome has: I use 2 monitors and in KDE when I switch to another workspace the workspace switches on both monitors, while in Awesome when I switch to another tag only the tag on the active monitor changes.
Thanks for this. As a new user of Ubuntu I stumbled into the meta-T shortcut but couldn't find out how to take advantage of it. Shift-T is what I was missing. I didn't have the terminology I needed to search effectively either.
> - Plasma's start time is up to a few seconds faster.
On the one hand, this is impressive. On the other, this is a vague indictment of modern software. Plasma logs you in, draws a background and the taskbar thing across the screen. There shouldn't be seconds involved in that to shave.
This must be a recurring feature, every few years someone goes through and fixes up load times.
They mentioned that the culprit for those additional load times was loading all the search backend for their application launcher which they deferred to the moment when there was actually the first search query (still shouldn't take multiple seconds)
Thanks for the clarification. Opinions will differ, of course and I cannot say whether the time required is justified or calls for optimizations, but personally, I view that moving this to the first search query is less elegant than having it happen during the initial start, when users are more accustomed to waiting a bit longer. With the new setup, the initial query will be less seamless than future ones at a time when the user is actually trying to accomplish something, leading to a degraded experience as far as I am concerned. But, as per usual, distros will make differing choices in this regard, and we shall see what ends up being shipped as the default in most cases.
Baloo (KDE) and Tracker (Gnome) are typically the first things I disable on any Linux desktop install that requires either of them. Tracker is hot garbage and a pain to disable. Baloo at least is granular and configurable enough to turn off file indexing without breaking search on the entire DE.
I'm surprised they didn't go to a more clever solution such as "start loading it as soon as the most core components have already loaded, and CPU usage falls below 30% during 3 to 5 consecutive seconds".
That'd seem to me like a sweet spot between what was before (making the whole system startup slower), and what you mention will be now (probably making the first search to load noticeably slower).
With modern hardware and OSes the ideal answer is typically to run non-essential tasks exclusively on efficiency cores. So you leave the performance cores free to handle stuff that the user immediately cares about.
> loading all the search backend for their application launcher
Wasn't the whole initial selling point of systemd parallel startup?
Why can't this launcher thing just get set to start up as early as possible, with lowest priority, and be set to not block initialization of anything else?
I'm surprised whenever I hear Wayland works flawlessly for someone, since I've tried using it on the NixOS unstable channel for a few releases now, and there are a few major bugs that make it unusable for me.
A show stopping one is that the bottom panel is stuck in the middle of the screen, with no way to move it down that I could find, and clicks register in completely different parts of the screen. It's difficult to explain. I noticed this starts happening after I start an Xorg session for the first time, which for some reason always switches back to the default after an upgrade.
If someone knows what bug this is and how to fix it, I'd appreciate it.
Then there are other minor issues like the cursor being tiny when hovering over a Firefox window, which I guess uses Xwayland(?).
The fact each application needs to explicitly support Wayland, and I need to switch to an entirely different app ecosystem, is insane to me. I'm willing to make the jump if it offered a truly better experience than X, but I've yet to see that.
I recommend swapping your nix package to firefox-wayland -I've been using it for quite a while and it's been great. For me, it's been a huge improvement over X - thr lack of screen tearing when srolling and playing videos is huge.
If an app uses one of the normal GUI toolkits, they don't normally have to do anything crazy to support wayland, so I'd be suprised if you need to swap to a whole new ecosystem (and there's always XWayland for those that do).
Thanks, I'll try that, but the thing is that I haven't experienced screen tearing on X for many years now, on Intel, AMD or Nvidia cards, and even on high refresh rate displays (90/120hz). Sometimes when plugging in a second monitor I need to force full-screen composition with Nvidia cards to get rid of tearing, but I can always set this permanently in a config file, I just haven't gotten around to it.
I really haven't had major gripes with X for many years. It mostly just works, and is rock solid. Long gone are the days when you had to manually craft the perfect config file depending on your setup. It also works great with multiple monitors, though I think there are still some color management issues, which don't particularly bother me, and lack of HDR, obviously. Plus the client/server architecture is super useful, to this day. IME Wayland feels like a usability regression, forcing me to change many of the programs I use, and it has critical issues, which seemingly only affect me...
You don’t notice screen tearing on normal usage with most desktop environments on X due to them using compositors. Try disabling it, or alternatively, to just play a high-resolution video on a high-res screen. The latter will often have artifacts.
> The fact each application needs to explicitly support Wayland
That’s not really the case, basically every framework supports both X and Wayland, out of the box. It is just sometimes marked as experimental/non-default by some software, e.g. firefox (which is not your average GTK/Qt app).
Wayland has a few glitches, but in general it's good to use.
For example, it doesn't have an API to get current cursor position (which breaks keepassxc's browser popup, goldendict's query popup), an API (which most compositor implements) to get current window state, which makes me unable to find an alternative of autokey (kwin script can do this, though, but it lacks the ability to execute arbitary commands..).
On the other hand, I don't see a killing feature that drives people switch to wayland.. (HDR can be one, but I don't use it) I mean, yeah X11 is old and unmaintained, but it WORKS.
Lack of screentearing is the killer app for me. The only GUI programs I run are Firefox, a terminal (foot or wezterm), and video games, so everything works and works better (I dont want screentearing for any of those 3) for me under Wayland.
The lack of screen tearing is a big feature to me, but sometimes people get x11 to work without this being a problem.
The other feature for me is security. Sandboxes can access the wayland socket and run GUI stuff, without the ability to read the entire screen or run commands that effect outside of the sandbox.
The wayland socket is a unix socket, exists as a regular file. Gating access to the socket works like any other file.
X11 has some interesting issues here. It creates an "abstract socket". These exists as files on the file system, but they are not regular files.
An interesting exmaple: A sandbox that starts with a fresh root and bind mounts in only what is required. Even when you don't bind the socket into the sandbox, the sandbox can still access it!
There are ways to prevent sandboxes from accessing the x11 socket, but this is definitely not what I would have expected!
With wayland, if the socket is not bind'ed into the sandbox, it can't be accessed, it behaves as expected!
This might not seem that important, but these features bring Linux desktop out of the 1900s with regards to security. Software like flatpak would not be able to effectively sandbox GUIs at all without wayland, a feature other platforms have supported for at least a decade or more!
The Linux desktop isn't making great use of this and related technology yet, on the standard distro, by default, applications can read your GPG keys and other important data, or even delete it. Wayland is a step towards fixing this!
I really like KDE. I feel like it's what the windows DE wants to be when it grows up. Everything is configurable and it's usually fairly obvious how to do so.
Coming from a long time 'minimalist' linux users, my biggest issue with it is that there's no ~/.config/kde that you can just copy between machines and get set up the way you like it. It's spread out all over the show.
Wayland doesn't fractionally scale chrome, discord, vscode, spotify, beeper, and other appimage and non-native applications. So even if KDE enables it by default, most users will have to switch back to X.
It doesn't scale the rest, because those are electron apps that either ship with old electron version, or disable wayland. Special mention goes to vscode, which ignores *-flags.conf and provides no other way to enable wayland, except for command line argument.
So basically it is maintainers of these apps dragging their feet.
Wayland not working with NVIDIA is a massive deal-breaker.
I've used Linux desktops on and off for the past 20 years, and it's incredibly frustrating how everything is perpetually in a state of semi-brokenness as they move on to the Next Thing. I still think about the early years of KDE4, which were a horrible regression from the stable and usable KDE3.
>Wayland not working with NVIDIA is a massive deal-breaker.
To state the obvious, though: only if you use Nvidia. And honestly, the Linux community not going out of their way to support an uncooperative hardware vendor is a good thing for everyone who doesn't use that vendor; it reduces scope.
Agreed on the frustration of chronic semi-brokenness though.
I agree; this was a complete deal breaker for me. I had no choice but to ditch Nvidia as a result. It's been really great since then. Especially not having to deal with their shenanigans when I want to do something simple like update my kernel.
FTFY: GNOME Shell/Mutter and wlroots don't fractionally scale chrome, discord, vscode, spotify, beeper and other Xwayland applications.
KDE Plasma Wayland by default applies Xwayland fractional scaling for X11 apps, which while less good than Wayland-native scaling mostly works. (If it's broken for you/an app you use, you can toggle it to let the DE scale up from 100%.)
OBS works fine via xdg-desktop-portal. I think Firefox (which you can use for Meet and Zoom?) also works, but it's been ages since I tried. Not sure about Chrome and Electron apps, but for those you'll at least need a pretty new version (which Discord for example refuses to ship iirc).
I've been using Wayland KDE on my gaming machine recently, it somehow made my cursor movement stutter. Switching to KDE X11 or Gnome Wayland solved the issue.
I hope they address this kind of thing prior to making it a default.
I use Wayland KDE but have shortcuts for Steam to launch within Gamescope [].
# normal steam with experimental HDR enabled, native resolution for my laptop and 165hz refresh rate.
alias steam 'steam-run gamescope -f --hdr-enabled -e -W 2560 -H 1440 -r 165 -- steam'
# full screen gamepad ui
alias steam_gamepadui 'steam-run gamescope -f --hdr-enabled -e -W 2560 -H 1440 -r 165 -- steam -gamepadui'
Computer is Asus G513qy, a a full AMD laptop with discrete graphics. Janky thermal design, no thunderbolt... so more like a stationary desktop. Wouldn't buy something like this again but for now it does the job and runs great with any modern distro.
I've been using Wayland + KDE (currently Plasma 5.27.8) for a very long time now on my machine with an Nvidia GPU (first a 1080TI, then a 4090). There are annoying bugs, but they're not show stoppers. I'm not someone who whines about problems like these, especially because I'm at the forefront of the Linux desktop, on a known-bad combination, so there are bound to be bugs :P
For example, a persistent bug that hasn't gone away in all this time is flaky wake from sleep. Sometimes, one of my monitors breaks in such a way that it drops down to 640x480 res and can't be changed back unless I do a list of different rituals. Another bug is that KDE's "Night Color" feature doesn't work, and whenever it turns on it completely freezes the display every few minutes (although unfreezing it is just a matter of pressing the Meta key, so it's not a big deal). I keep sleep and Night Color on, even though they're broken in the hopes that I'll one day wake up to an update that fixed them. I'm 99% sure those are Nvidia driver bugs, and Nvidia has been working to improve their Wayland support recently (for example, there's an upcoming Night Light fix that I'm excited for: https://www.omgubuntu.co.uk/2023/10/gnome-night-light-nvidia...). Being able to follow progress like this is fun for me.
I'm sure there are some other issues that I'm not remembering, but overall the desktop is completely 100% usable and reliable (for both working and gaming). My only other machine is a laptop, and that too is running KDE+Wayland (though no Nvidia). I have not switched back to X11 once in the months (year?) since I switched to Wayland, and have not missed anything.
EDIT: also, I should mention that I'm on Fedora Kinoite. I feel like a lot of problems people have with KDE/Wayland/Nvidia comes down to bad configuration somewhere. If you're on a similar machine, I recommend you try switching to Kinoite since it's designed to just work out of the box, and is pretty much impossible to break.
All of that work for me on AMDGPU, so likely not a KDE bug.
Same here I have been using wayland with sway and KDE depending on the machines (some with intel gpu, but no nvidia) for a few years now and nothing to complain about (fedora, arch and ubuntu)
This might very well be a nvidia problem, their driver comes with its own suspend functions which replaces your default one during install (it usually tells you on the dnf install screen)
This is great. Running Wayland has only been getting better.
Hoping this transition to default will get the attention of developers at Zoom, and they can get screensharing to work in their native app without resorting to browser mode or some virtual camera hack. My biggest gripe for sure.
KDE doesn't have this problem, actually. At least it's not uglier than X11, since by default (since 5.27 or so) it uses X11's DPI scaling for Xwayland apps by default. You can change that under Display settings if it doesn't work for your apps and you need to revert to ugly blurry upscaling like GNOME and sway do.
Most distros have been defaulting to Wayland with GNOME for years at this point. I wouldn't hold your breath for them to wait around on Xorg for KDE despite upstream now switching.
I think I'll wait for another cycle or two before moving to Wayland. Seems like there's a long tail of small issues that need polishing. The zoom screenshare one scared me off a long time ago.
There's a new fork of barrier called input leap (not to be confused with the leap motion), https://github.com/input-leap/input-leap with work on getting Wayland support in shape. Not sure how far the support is atm, but the gnome 45 release notes mentioned "Wayland support for Input Leap" (https://release.gnome.org/45/)
last I checked nvidia on wayland was all kinds of broken, especially with multiple displays, can't rotate a display either. So I assume that has all been fixed then? (j/k)
KDE is pretty feature-rich, but it is just ugly. Look at the poor use of spacing/margins, the ugly black & white icons, the ugly font, the square edges, and just the overall look & feel. It needs a major face-lift. I dislike the attempts to dumb down GNOME but KDE being ugly makes it really hard to switch to.
Excellent spacing: no useless wide margins. Stylish icons that can blend in with various themes, which are clear and easily distinguishable. Font: most distro's actually default to Noto, not KDEs own. But you know what: this you can change, KDE didn't remove any and all options for customisation after all.
You can pry Plasma out my cold, dead hands. Gnome, Macos, and Windows: now that's amateur hour.
To be honest, there are many places in the "refreshed" Breeze look'n'feel and the Kirigami suite of apps in general that could really use some margins. It doesn't need a lot of padding, just a pixel here and a pixel there - not to make it less dense, but to make it look cleaner and consistent. To my eye, screenshots from the blog post look worse in this regard than Plasma 5.
I use KDE and MacOS on different laptops. Even though I like KDE , I find that compared to MacOS, KDE is the one that comes out a bit amateurish... the attention to detail in the MacOS UI is just amazing. Am I missing something?? How is KDE better (it really feels much more raw to me, but being on Linux I don't really expect much more)?
I think it comes down to preference. I definitely don't think KDE is amateur, but it certainly lacks the fine polish Apple has the money to pay for.
That said, I find MacOS frustrating at times because basic things can feel like a pain. Switching between several windows of the same application isn't a simple Alt-Tab; the green button enters full screen mode (KDE/Windows F11 Full Screen is equivalent) instead of leaving just the menu bars and application bar on screen; tiling windows isn't as simple as ramming the application against the border you want to expand to. These are simple things I do often in other OS's that have tedious menu workflows in MacOS. I still like a lot of MacOS though.
Ramming against edges doesn't work as well in a multiple monitor arrangement. I forget which environment had this as a default and in what era but one of the smarter defaults I've seen is super(windows key) + keypad numbers. For instance 1: Bottom left quarter, 4: left half, 5: maximize
I guess MacOS is more polished than KDE. Probably more polished than anything. A few billions here and there help.
Given that KDE is an open source community project it’s quite good.
> the attention to detail in the MacOS UI is just amazing.
Please feel free to answer these questions (I think I already know the answer so this is to illustrate a point, but I will also be happy to learn a better way):
1. How do you cut and paste something on a Mac using the keyboard?
2. How do you cut a file in one folder and paste it into another in Finder using the keyboard?
3. How are you supposed to know the answer to 2. except by learning it by discussing with Mac users?
4. I haven't tested recently but it used to be that there was no single consistent way to get to the start or end of a line while selecting the text in between. CMD-shift-arrow left/right worked most places, but not everywhere. Ctrl + a/e always worked but couldn't be used to select text, only to move the cursor.
PS: I already asked for a Mac as my next machine at work. But as someone who uses and have used almost all major desktops (Gnome 2, KDE 3, 4 and 5, Windows 95 - 10, several iterations of Mac OS X) for significant time, I do feel the "attention to detail" argument for Mac is a bit overblown.
PPS: I planning to get a Mac because it has good hardware, is silent, fast and because I want to try it again after a few years. Also I have migrates my whole family to Apple phones since last I used Mac at work.
I realize that's not a very satisfying answer, but it's quite obviously a conscious UX choice, and IMO not a bad one. The problem is, where does the file go in between the time you cut and you paste? If you accidentally copy something else to your clipboard, do you loose the entire file? Does the file appear in the Trash or is it deleted perminently?
You can use cmd+c and then cmd+alt+v to copy a file to your clipboard and then move it to a different directory, but moving is distinct from cutting in that it's a synchronous operation. There is no in between point where your file is lost in the ether.
3. Assuming we're referring to "how do you move a file with the keyboard", you open the "edit" menu in Finder's global menu bar, then hold down "alt", at which point "paste" will change to "move item here".
I agree "hold down option" is not particularly discoverable. However, it's a standard across all of macOS for enabling alternative options. For example, holding down alt will also change "Minimize" to "Minimize All", and "Quit" to "Quit and Keep Windows" (or "Quit and don't keep Windows" depending on your default preference). Once you know this basic idea, it's consistent.
4. To my knowledge this works in all apps that use Carbon/Cocoa. If apps reimplement these views and disregard Apple's design guidelines, what is Apple supposed to do?
P.S. I think modern macOS is a poorly thought out mess but I love how it was a decade ago (up to OS X 10.9).
I am waiting for a Mac as my next machine and the bit about alt in the menu should help quite a bit.
I am stil a bit annoyed from last time I used Mac a decade ago and all the things I couldn't figure out without extensive web searches or asking coworkers but I have still figured out it is time to try it again :-)
The cmd+x/c/v keyboard shortcuts have been around much longer than the ctrl+x/c/v equivalents - copied by Microsoft in Windows 3.0; it was all CUA previous to that.
1. So to cut in applications (or anywhere anything is selectable other than Finder), it's cmd+x, as it has been since the early '80s; likewise, paste is cmd+v
2. In Finder, you'd use cmd+c to copy the file, cmd+v to paste a copy and cmd+opt+v to move the file. This was introduced, I think, in Lion (OS X 10.9).
3. Erm, using the menu bar to begin with? Reading the built-in help? Searching the internet? How do you learn it in Windows and/or any other $desktop or $wm?
4. *Emacs keybindings work in most places in macOS, and they have since the first beta of OS X.
Cutting stuff in word processors or content areas, and moving files, are two different operations that behave the same approximately nowhere, and it’s not obvious that they should share the same commands. Digging back into the distant past when I first encountered this on Win 95 (I don’t think I knew about it on 3.1?) it was indeed confusing, though of course now it seems OK, because I got used to it.
The example of whether to use the same command for ready-a-file-to-perhaps-move-but-leave-it-otherwise as for definitely-remove-this-thing-now-and-maybe-put-it-somewhere-else is more a matter of taste than one or the other representing better attention to detail.
The margins and paddings are objectively terrible. It is not a matter of taste. There is no coherent design here, nothing aligns, there is no visual hierarchy and often it doesn't even show more information than GNOME despite looking so much worse.
> often it doesn't even show more information than GNOME despite looking so much worse
Yeah, even as a GTK developer I'm gonna throw a yellow flag here. I've got both Qt and GTK apps that I really love and daily-drive, but none of my GTK3, GTK4 or Libadwaita apps even remotely approach the information density of a Qt app. There's nothing wrong with that; simplicity is nice for a lot of software. But programs like Cutter, Okteta or even music apps like Eliza cannot be built with the GNOME HIG. I'd even argue that certain GNOME apps like Meld, GIMP and Glade/Cambalache suffer from relying on the GNOME HIG too much.
I used to think the same as you, and I would still agree that out of the box, KDE is probably uglier than Gnome is. However, I've grown to be rather impressed with what you can do with KDE, and more importantly that KDE allows you to do it. Using Gnome was a slow descent into watching all the plugins and UI choices I liked be slowly removed or break over time.
I am not really sure how your comment is a reaction to mine? Having said that, opinions are opinions. I much prefer KDEs UI/UX choices over those made by Gnome and have no problem with the general look and feel of the interface.
Beauty is entirely subjective, though. I don't find KDE ugly at all. Nor do I find Gnome ugly. But I find KDE easy to use, and I don't find the same about Gnome. Even if KDE was ugly to me, usability is more important. KDE is also very configurable, so I could probably bend it into something more to my taste.
unironically if that's what you're complaining about, you might have liked kde3 a lot (yes, it's been 15 years and i still miss it).
anyway, kde is very configurable so most likely you might bend it to your liking, had you been somehow forced into kde.
> KDE being ugly makes it really hard to switch to
have you considered XFCE ? to me it's the best so far: it does all the things i need, it's fast and it isn't dumbed down like gnome.
It's not on wayland yet, but frankly I don't care much about the X11 vs Wayland debacle (though i slightly side with Xorg). Essentially I run whatever XFCE needs.
When XFCE will run on wayland, I'll run on wayland I guess.
kde3 with kdemod and the breathtaking choices of _actual_ themes (not just color schemes or "accent color" bs, but a plethora of widget themes that could make it look like anything you wanted plus rich guis to customize any and everything) plus the "emulation" of windows 9x HIG which kde apps back then used to copy and even improve upon was the pinnacle of desktop OS for me. Every major OS has been a major downgrade since... :`(
I don’t care as long as something is capable of solving issues, but for those who have the needs to have pretty DE, kde has a lot of custom themes and customisability, otherwise its just a boring old guy complaining at the sky, the linux world is better off without those kind of people
I agree on the spacing part, but it's not fair to complain about it and not submit actual solutions to them.
Have you tried getting to know their contribution process, looking at their current design guidelines and submitting a new, well-thought-out and complete design guideline proposal to KDE?
Having someone who notices the current problems with that aspect, like you, actually do the work and help fix it could really make a difference! :)
I would really like to step up here (not joking), but I am yet to understand the process and explore KDE once again. I do like it generally, but visually it’s a mess. I use swaywm on my Arch laptop and pure Gnome 45 on a couple of my Fedora 39 PCs (work and home ones).
Would appreciate if someone can help me to get quicker at understanding with their system, so I can actually contribute and not just complain about it being inconsistent here and there. Even though I not use KDE myself, I use it from time to time on other’s machines and enjoy playing with it. I would consider a switch only if the DE would highly reconsider itself towards simplifying and making other defaults. At the moment, Gnome looks like it to me.
That's great! (i mean this unironically in case that came off sarcastic)
I'm not a kde contributor myself, but they seem to have a matrix channel for that purpose at [ https://webchat.kde.org/#/room/#new-contributors:kde.org ], where the contributors themselves can probably point you to the relevant resources and get started.
About your second point: From what I've been reading on their blog, a large focus of Plasma 6 is making the defaults more sane and simplifying them a good bit.
It's obviously a huge challenge to design systems that are simple and polished to understand and use on the surface, but powerful to customize underneath, and I'm interested to see how it will be tackled.
Even as someone who’s using KDE on one machine I tend to agree. Things can be polished up a bit by changing themes and font settings but that only goes so far — it can’t do anything about widget arrangement and the general flow of UI design in KDE apps which I find somewhat awkward.
I absolutely agree. It is functional and stable, but my gosh I can’t stand the design. The oversized top decoration for windows as well as the extremely poorly rendered pop-up for hovering over elements on the task bar.
> extremely poorly rendered pop-up for hovering over elements on the task bar
The only real issue with that, for me (and it's not poorly rendered, not sure what that means) is the time it takes to activate. You have to hover over it for like a second before the pop-up shows up; this is especially frustrating when items are grouped.
I'm gonna be very controversial here, but the problem is that the vast majority of Linux users are blind to aesthetics and design. Perhaps the vast majority of people are. Mind you, being blind to something does not mean it does not exist.
Just look at the desktop wallpaper or any font choice of the average person. It's no wonder Comic Sans and Papyrus are the most common fonts in the world, even for formal business documents.
Which is perfectly fine, and design challenged people would enjoy a well-designed desktop environment as well, the issue is that they are ready to defend bad design choices to the death. So mention KDE having lower design standards than GNOME and you get a ton of people telling you "it looks fine to me, I don't know what you're moaning about."
You'd get the same answer telling the average person that, no, Comic Sans is not appropriate for an obituary.
---
That said, KDE is slowly growing on me, but the default GNOME look is much more slick, even though GTK4 does 1/20th of what QT does. All modern GTK4/Adwaita apps are a flat, slick blob of grey list widgets.
Colour theory and the use of accent colours or non-monochrome icons is unknown to the average GTK developer.
> the vast majority of Linux users are blind to aesthetics and design
Yes, my god is KDE ugly, but just because they don't prioritize aesthetics doesn't mean they don't know better.
Linux users (and developers) prioritize functionality above all else. Meanwhile the entire discipline of visual design is just not very consistent. There are far too many subjective little rules and exceptions to maintain and they change all the time. Who cares?
Deliberate oversimplification of how things are displayed from the perspective of the implementer rather than the designer is its own aesthetic. It's the choice of reason and a rejection of the imperfections of human perception. It's rebellious and edgy. It's not a mere rationalization of laziness or ignorance, but a middle finger to business types and their marketing henchmen with their annoying design guidelines and requirements that just get in the way of the actual product. KDE was not made by people getting paid to smile and nod while staying quiet in their cubicles (at least not during the hours they worked on it). Very rock and roll. I'm all for it.
This is exactly the problem. These two things are not at odds. Good design is functional. Good design makes information easy to find and digest. Visual soup is hard to organize and make sense of. Form is a component of function. A knife with another blade instead of a handle can do less cutting, not more. A knife with a contoured handle is easier to wield than one with a flat block handle. These concepts translate well into UI.
If you don't think they're at odds, then you should try your hand at implementing what you think looks good without breaking upstream features. Then try maintaining it indefinitely.
> Visual soup is hard to organize and make sense of.
Initially yes but once you know where things are, having everything be visually distinct can sidestep a lot of unnecessary GUI features (search boxes, configurable views, toolbars, etc.) to overcome the impenetrable and dishonest sameness that comes with "good design". If there's a way to avoid writing more code to maintain, they're going to find it. Why obscure the reality of the underlying logic and data when it just gets in the way of progress towards more functionality?
In a production setting desktop environments are rare anyway and that's where applications are at their most complex. That should be proof enough that visual design isn't functional. You want functional and accessible? That's a simple 1KB config file on a server somewhere that you edit from a terminal instead of an endless sea of GUI menus and a lifetime of wrist and eye issues. The ultimate "contoured handle" is a copy of vim! If you don't see the equivalence then you don't understand how much of an unnecessary and tortured metaphor the desktop really is. It's a concession, not a requirement.
Information can be dense and designed. It's actually more important for dense information to be designed well to avoid devolving into soup. KDE and its many design inconsistencies and automatic collapsing of menus into "... and X more items" leading to menus nested 5+ deep is not good design.
Away from KDE, away from GUIs, a dense text readout of key value pairs with no new lines presents a lot of information, but is functionally useless. A tiny dash of design - pairs on new lines, some boldness for keys, indentation for values that wrap to new lines - drastically increases the readability. No matter how familiar you are with this particular text dump, this remains true. You can't escape design by running off into the terminal.
In GUIs, having movable, modular panels can be good design. Blender's design is quite good, especially among its peers, and everything about it can be shuffled around in ways most apps with modular panels can only dream of.
Again, we're not talking about a 1 dimensional spectrum between design and density. Design isn't just "add padding" and "remove customizability."
Your comment is funny because Comic Sans is actually pretty good from a design and accessibility perspective.
The complaints are mostly "fun fonts are not appropriate for a business" and everyone jumped on hating it.
But those who are not that stuck-up don't hate it with a passion...
There you go proving that it’s not just that many Linux users have a bad design sensibility, but worse, they will defend bad design as good design.
Not only do you defend it, even your defense proves that Comic Sans is a bad design, because matching the needs of what it’s used for (in this case a business setting, where fun fonts are not allowed) is one of the highest factors in good design.
And you display another behavior which is common among Linux users that the parent hasn’t mentioned. Finally resorting to complaining that the Users are wrong.
Here’s the usual 3 step that prevents a lot of progress in the Linux world.
1. Have bad design ideas.
2. Defend those bad design ideas.
3. Insist users and the people who want better design are wrong.
Not all of life is business, and there's some business where fun is allowed.
Also, Comic Sans is not even something rebellious, it's literally a good-looking, quirky font, people just hate it because it looks distinctive and it stands out.
This is a bit dystopian to scream "NO FUN ALLOWED"....
Oh my, you guys came from ’90s or ’00s with your Comic Sans. I myself forgot this font ever existed, I see it nowhere these days.
It’s beyond awful, just get yourself some link to Google Fonts at least, and explore the myriads of much better looking fonts in every possible way. That’s exactly the problem with aesthetics perception of the FOSS community, having this subjective ‘oh but I like it’ stand on absolutely mediocre shit. I’m very happy that by some (unknown to me) reason modern Gnome (45) looks (and works) so good.
> 1. Have bad design ideas. 2. Defend those bad design ideas. 3. Insist users and the people who want better design are wrong.
Not sure if you're right or not about Linux, but what you've laid out there has been Apple's approach for the past 20+ years and it's been working out pretty OK for them!
Yeah, just the company making the best UI/UX and basically leading the way for everyone. So bad indeed that their just smartphone department worth more than some countries GDP.
Since we're all being frank and sharing our two cents.. The people who criticize Linux aesthetics often come across as insufferable snobs who are obsessed with small details and have a hard time acknowledging the subjectivity of their personal preferences. If they spent half as much time configuring KDE/etc as they did complaining, then they'd soon have a system that aligns with their preferences.
As a designer, the absolute warmest take I've seen towards Comic Sans is "It's fine, not good, but fine. But it's so overused and misused that it's become slightly radioactive."
You're confusing design with aesthetics. KDE design is fine. It doesn't chase the latest fads in GUI appearance which rightly makes it more usable than its peers which seem hellbent on reinventing conventions every few years for superficial reasons. This isn't just to the benefit of rigid nerds, a lot of people grew up with classic MacOS and Windows and expect some basic functionallity.
> the problem is that the vast majority of Linux users are blind to aesthetics and design.
This is just saying that people who don't share your taste are wrong (and worse, just don't know any better). I dispute that. Aesthetics are highly subjective, and things that have an aesthetic sense that I dislike are not wrong, and the people who do like it are not ignorant or idiots.
If the vast majority of people don't perceive it, it isn't bad. Some people are deathly allergic to peanuts, but that doesn't mean that food with peanuts is inherently bad, just that a small number of people react badly to it.
> the problem is that the vast majority of Linux users are blind to aesthetics and design
I'm not blind to it, but it certainly comes way down on the list compared to other things, like actually working.
As an example, my SO recently bought a new egg timer. She picked it solely based on it looking good. Turns out it's completely useless because you can barely hear it even if it's right in front of you.
Sure, if I'm buying something purely decorative, like a statue a painting, then aesthetics is all I'm going for. But for things I have to use, the thing has to do its job above anything else.
This is something a lot of designers seems to miss. Just look at Windows 11 when it launched, and the completely idiotic decision to force icons to group without text. Sure it looks less cluttered, but it's just like the egg timer, completely useless.
Honestly I don’t need my DE to look pretty, I need it to do is show me information, not crash and not search bing so all in all KDE is just fine in my opinion.
* Chrome needs a bunch of extra flags to launch in Wayland mode
* Firefox needs extra flags to launch in Wayland mode
* Xwayland is just broken on Nvidia with both sides refusing to compromise (implicit vs explicit synchronization - Nvidia refuses to add implicit and Xwayland refuses to take Nvidia’s patch to do explicit). What this means is that you get tearing, flickering and all sorts of terrible graphical artifacts
* Chrome only just fixed HW acceleration for Nvidia (latest m120 beta)
* enabling vulkan causes Chrome to fail to render (although this may just me needing to try reinstalling Nvidia drivers)
* I thought it was an acceleration thing, but even with HW acceleration fixed Chrome has a bug where the mouse pointer leaves behind white speckling when mousing over on a dark background. VSCode doesn’t have this issue.
None of these issues appear in X and this is from someone who thought Wayland is the right way to go (eg you didn’t see this kind of story on Mac when they switched to HW accelerated compositing).