KWin can now show different virtual desktops on each screen independently.

Back in 2005, a bug report was filed by Kjetil Kjernsmo, then running KDE 3.3.2 on Debian Stable. He wanted the ability to have each connected screen show a different virtual desktop independently, rather than having all displays switch as one unit.

Over the years, over 15 duplicate reports piled onto the original as more people ran into the same wall. And that’s not a surprise, because multi-monitor setups have become increasingly common.

The technical reason why this issue stayed open this long comes down to X11. Implementing it there would have required violating the EWMH specification, which has no concept of multiple virtual desktops being active at the same time.

The KWin maintainer Martin Flöser had said as much in 2013, effectively ruling it out for the entire KDE 4.x series. The only realistic path was through Wayland, and that path needed someone willing to actually walk it.

Someone finally did. The feature has now landed in KWin’s master branch and is set for a Plasma 6.7 introduction.

  • haverholm@kbin.earth
    29·
    1 day ago

    I understand the scepticism, it’s easy to be wary of assisted coding as a gut reaction these days. However, the programmer replied to his own merge request:

    I put roughly 200 hours into this (largely due to not being familiar with the codebase and technologies involved beforehand)

    …and I find it hard to believe a vibe coder would bother prompting his magic eightball for that long? Maybe that’s my personal prejudice against that particular set of (often wannabe) programmers showing.

    There’s also an “AI” disclosure in the original merge request that says:

    • None of the submitted code is AI generated. I wrote all of it myself (largely by copying existing code).

    • I used OpenAI Codex in the following ways:

      • I used it to generate a quick prototype to give me an idea of where to start. The prototype was MUCH more limited and I rewrote it from scratch (i.e. none of the code submitted here is from there).
      • I used it to help me with debugging/understanding some of the stuff new to me (e.g. how does a value get to QML, wayland protocols, understanding crashes, …).
      • I used it to review the code before submitting it (in addition to manually reviewing it myself). It caught a few minor issues.
      • I used it to write a “touchpad simulator” patch that I used for testing touchpad gestures (not submitted).

    YMMV on what side of vibe coding that usage falls.

      • haverholm@kbin.earth
        8·
        23 hours ago

        Big caveat there that “AI” still makes stuff up out of the blue, despite being ostensibly trained on “all of human knowledge”. And this guy seems to have used its generated code accordingly to better apply his PHP skills to C++.

        You can probably tell I disagree strongly with “AI” being called “just a tool” 😛

        • locuester@lemmy.zipEnglish
          5·
          16 hours ago

          Tools giving incorrect answers isn’t some new thing that AI delivered.

          AI makes stuff up and random dude on Stackoverflow posts incorrect answers. I see no difference here.

          This programmer used it the same way I do. As a tool. It’s a great tool; I find it hard to go back to being without it.

          • luciole (they/them)@beehaw.org
            3·
            9 hours ago

            The incorrect answer on Stackoverflow can be downvoted and commented by the community. With a LLM the incorrect answer is perfectly formulated and you’re alone to recognize it. I’ll favor a large community’s advice (cultural quriks and all) over a LLM’s any day.

            • locuester@lemmy.zipEnglish
              2·
              8 hours ago

              To each their own I suppose. I truly can’t imagine being without this productivity booster. I had the same opinion as you a year ago, but things have radically changed the last 6 months.

        • Powderhorn@beehaw.orgOPEnglish
          3·
          21 hours ago

          If you look further into the thread, he said he spent 200 hours coding this himself with the assistance of “AI” … look, I’m not really a coder, but give me 200 hours, and I can certainly pull off some shit. If you’re consulting an LLM like a book, I’m not really sure where the problem lies.

          • haverholm@kbin.earth
            8·
            13 hours ago

            Yeah, I also put more faith in those 200 work hours than in the original, generated code which the guy completely rewrote before submission.

            consulting an LLM like a book

            Saw a news item the other day, reporting that a significant number of university students now use “AI” bots instead of course literature. One student replied, “Nah, I opened a book like once. Anyway, the literature can be just as flawed as AI because there’s new research being made all the time”…

            There is a significant overestimation of the factuality of “AI” responses at play there. And a lack of understanding of the entire chain of fact checking, verification, and review that goes into making a book, particularly for education.

            I know that is slightly OT, but I think the comparison is fair.