It's not dead. It's being murdered. Microsoft, Apple, Gnome and KDE are making the experience worse with each update. Productive work becomes a chore. And the last thing we need is more experiments. We need more performance, responsiveness, consistency and less latency. Everything got worse on all 4 points for every desktop environment despite hardware getting faster by several orders of magnitude.
This also means that I heavily disagree with one of the points of the presenter. We should not use the next gen hardware to develop for the future Desktop. This is the most nonsensical thing I heard all day. We need to focus on the basics.
I agree with this. I remember when Gnome 3 came out, there were a lot of legitimate complaints that were handwaved away by the developers as "doesn't work well on a mobile interface", despite Gnome having approximately zero install cases onto anything mobile. AFAICT that probably hasn't changed, all these years later.
I don’t know. I just started distributing a gtk app and I’ve already gotten two issue reports from people using it on mobile experiencing usability problems. Not something I thought I’d have to worry about when I started but I guess they are out there.
KDE? It has great performance, it's highly configurable, and it's been improving.
Many people don't seem to like GNOME 3, but it has also been getting better, in my view.
I agree Windows and macOS have been getting worse.
I don't know how you can say this about KDE with a straight face.
They basically never remove features, and just add on more customization. You can get your desktop to behave exactly like Windows 95, if you want.
And the apps are some of the most productive around. Dolphin is the best file manager across every operating system, and it's not even close. Basic things like reading metadata is overlooked in all other file managers, but dolphin gives you a panel just for that. And then tabs, splits, thumbnails, and graph views.
KDE 3.5 -> 4 was its Gnome moment. The fall wasn't as hard, they have come up somewhat from there, but they still don't touch their KDE 3.5.10 Konqueror-primary days. Dolphin was a great example of the rot setting in during those days; Konqueror was the all-in-one browser application and Dolphin was the simplification cancer coming in. THAT particular KDE (which is TDE today) was peak KDE. Yes I am still this buttmad about Dolphin's introduction.
FWIW, this just isn't true for KDE. We hit a rough patch with the KDE 4.x series - 17 years ago - that has been difficult to live down, but have done much in the way of making amends since, including learning from and avoiding the mistakes we made back then.
For example, we intentionally optimized Plasma 5 for low-powered devices (we used to have stacks of the Pinebook at dev sprints, essentially a RaspPi-class board in a laptop shell), shedding more than half the menory and compute requirements in just that generational advance.
We also have a good half-decade of QA focus behind us, including community-elected goals like a consistency campaign, much like what you asked for.
I'm confident Plasma 5 and 6 have iteratively gotten better on all four points.
It's certainly not perfect yet, and we have many areas to still improve about the product, some of them greatly. But we're certainly not enshittifying, and the momentum remains very high. Nearly all modern, popular new distros default to KDE (e.g. Bazzite, CachyOS, Asahi, Valve SteamOS) and our donation totals from low-paying individual donors - a decent proxy for user satisfaction - have multiplied. I've been around the commnunity for about 20 to 25 years and it's never been a more vibrant project than today.
Re the fantastic talk, thanks for the little KDE shout-out in the first two minutes!
Unpopular take: Windows 95 was the peak of Desktop UX.
GUI elements were easily distinguishable from content and there was 100% consistency down to the last little detail (e.g. right click always gave you a meaningful context menu). The innovations after that are tiny in comparison and more opinionated (things like macos making the taskbar obsolete with the introduction of Exposé).
Heh, the number of points you've probably gotten for that comment, I don't think that it's that unpopular. Win 98 was my jam but it looks hella dated today, but as you said, buttons were clearly marked, but also menus were navigatible via keyboard, soms support for themes and custom coloring, UIs were designable via a GUI builder in VB or Visual Studio using MFC which was very resource friendly compared to using Electron today. Because smartphones and tablets, but even the wide variety of screen sizes also didn't exist so it was a simpler time. I can't believe how much of a step back Electron is for UI creation compared to MFC, but that wasn't cross-platform and usually elements were absolute positioned instead of the relative resizable layout that's required today.
Recently some UI ignored my action by clicking an entry in a list from drop down button. It turned out, this drop down button was additionally a normal button if you press it in the center. Awful.
> UI creation compared to MFC
Here I'd prefer Borland with (Pascal) Delphi / C++ Builder.
While it should be beneficial, the reality is awful. E.g. why is the URL input field on [1] so narrow? But if you shrinks the browser window width the text field becomes wide eventually! That's completely against expectations.
I don't think it's a stretch to call it the UI language of 95, while 2000 just adds more functionality within the bounds of that framework. Add in the Win7 search bar in the start menu, and the OS not crashing, you haven't really done anything of note with the UI beyond staying within its framework. It'll still be a Win95 UI.
Meanwhile, WinXP started to fiddle with the foundation of that framework, sometimes maybe for the better, sometimes maybe for the worse. Vista did the same. 7 mostly didn't and instead mostly fixed what Vista broke, while 8 tried to throw the whole thing out.
Always the same lies. People "stepped up" in the and as a result were outright banned from the gitlab (instead of e.g. just rejecting pull requests). Current maintainers refuse to do any release management and instead treat every merge into master as a new release. This kind of sabotage makes development or contributing very difficult. Also the people that used to maintain X11 (e.g. Keith Packard) had nothing to do with building Wayland.
Wayland on the other is just a insanely stupid API. Everybody advocating for Wayland should be forced to write a simple client at least once without relying on behemoths like GTK or Qt.
Yeah, lies and then you come in with shit like this. You can surely show several proofs then, right?
> Wayland on the other is just a insanely stupid API. Everybody advocating for Wayland should be forced to write a simple client at least once without relying on behemoths like GTK or Qt.
Why would you do it outside of toying around? Btw, I have and it's nothing out of ordinary.
If this was a paid job, both you and sprash would have been fired or at least PIP'ed several comments ago. This kind of behavior has ZERO place in any code project - professional or volunteer.
If by "people" you mean a fascist who doesn't know how to program, then sure. But the sensible people who don't present a security threat with their politics or with shitty code are 100% in the Wayland camp.
> Also the people that used to maintain X11 (e.g. Keith Packard) had nothing to do with building Wayland.
Those people aren't maintaining X11 today, are they? The people who are maintaining X11 today have put it in bugfix-only mode and have told you, many times, that the future is Wayland. End of discussion.
Look, you want to run a retro 90s desktop for shits and giggles, that's great. There's even an officially supported path for this use case: Ariadne Conill's Wayback. But the DEs and the toolkits are all removing X11 support within the next year or two. There is no future there. You want to keep running modern software, you will have to switch to Wayland eventually—and soon.
> Wayland on the other is just a insanely stupid API. Everybody advocating for Wayland should be forced to write a simple client at least once without relying on behemoths like GTK or Qt.
Nobody actually develops applications that way. They all use a toolkit, and the behemoths cover pretty much 90% of actual application development (modulo things like Electron). Both of those, by the way, are deprecating X11 support.
There is no X11 bitrot. Just a lack of funding. If funding for Wayland stopped today, Wayland would die much quicker than X11 because there is essentially zero community involvement whereas for X11 there are enough people that care to keep it alive for free.
The only real Wayland community effort is Hyprland. The author of that has been banned from contributing to Wayland by the corporate sponsors of Wayland.
Hence there is no community.
Also the whole "there is no Wayland" "it's just a protocol" spiel has been played so often that I believe Wayland apologists are mostly bots.
Sway and wlroots are also real community projects with no big corporate sponsors. Its main contributors are also very active in the development of wayland protocols.
Desktop Linux needs standardized and stable infrastructure. X11 delivers on that perfectly.
Wayland despite receiving huge amounts funding has actually far more moving pieces. Even for the simplest tasks you have to deal with a dbus infested portal maze, many parallel infrastructure effects and high fragmentation. The API is atrociously stupid and cumbersome.
Besides that the modesetting driver of xorg also sits "properly on top of kernel abstractions". How is this in any ways a relevant criterion. What matters is that Wayland clearly makes the wrong abstractions for Desktop applications and the vast amount of parallel infrastructure required to do even the simplest tasks shows that.
> Even for the simplest tasks you have to deal with a dbus infested portal maze
The simplest task is displaying a buffer, or changing a buffer, or handling events, and absolutely none of them have anything to do with dbus whatsoever. Also, have you seen an X11 desktop environment like KDE or Gnome? I recommend looking at all the dbus messages that are in flight there at any time.
Tell that to Havoc Pennington. Dbus was the solution he came up with based on requirements and constraints set by the DEs. A lot of people have claimed we need something better, but nobody has actually created something better. Till someone does, Dbus is the standard for client communication with Wayland compositors outside the core protocol. Sure beats piping stuff over X ClientMessage events.
If I instead don't, and let you know that the key is there in the source code, hopefully at least one deserving person might learn how to look through source code better, and none of the lazy people get it :)
This rare 22-segment display is the most interesting I've seen so far. It can display the full 7-bit ASCII range including special characters: https://www.youtube.com/watch?v=D2cyZoCat_U
> I think MGR has the greatest potential for a comeback because of its unique architecture
I think so too. But this time with modern drawing primitives. Instead of lines an circles we need shaders and textures.
In the end, even the most modern UI is nothing more than a terminal: Low bandwidth input from keyboard and mouse events and low bandwidth output (like draw checkbox at x,y). The rest is done by some drawing or blit routine which can be entirely managed on the GPU.
This also means that I heavily disagree with one of the points of the presenter. We should not use the next gen hardware to develop for the future Desktop. This is the most nonsensical thing I heard all day. We need to focus on the basics.
reply