KDE 6.8 drops the X11 session: what breaks isn't your apps, it's everything around them

KDE has put out the final beta of Plasma 6.8 and, barring delays, the release itself arrives on 14 October. That's the day the project turns thirty, so the birthday present comes with something taken away: 6.8 is the first release that ships no X11 session at all.

What happened

In Plasma 6.7 the X11 session could still be installed separately, as an optional package. From 6.8 onwards the desktop runs on Wayland and there's no alternative. Applications written for X11 still open, thanks to XWayland — the compatibility layer that translates between the two worlds — but the desktop itself no longer starts under X11. This was flagged more than a year ago, so it isn't a surprise.

The rest of the beta is mostly about looks. The Oxygen theme is back, with more polish for this release. The Union theming engine remains a public preview and has started styling the classic Qt widget applications — the file manager and the text editor that ship with the desktop, for instance — although KDE itself warns that support there is still rough. There's also a documentary: an Irish outfit, Cult.Repo, is making a feature-length film about the project's history, and a trailer is already out. Meanwhile, changes for 6.9 are already landing.

And alongside all that, Klassik has appeared. It isn't a fork or a port: it recreates the look and behaviour of KDE 3 on top of Plasma 6, written with Qt 6, Qt Quick and KDE Frameworks 6. It rebuilds the early-noughties panel, the K menu, the quick launcher, the task manager, the lock and logout buttons, and even the old clock. Because it sits on the modern base, it gets Wayland and fractional scaling for free. For context: KDE 3.0 shipped in April 2002 and the series ended with 3.5.10 in August 2008.

Why it matters

"Your apps still work" is not the same as "everything still works"

This is the nuance the headlines lose. XWayland translates drawing and input for an X11 application. What it doesn't translate is everything that used to be done by reaching straight into the graphics server, because X11 let any client do that:

Global keyboard shortcuts grabbed by the application rather than the desktop.
Screenshots and screen recording, including the "share your screen" button inside your own tool.
Automation: xdotool, wmctrl, scripts that move windows or inject keystrokes into another application.
Placing your own window exactly where you want it: splash screens, floating panels, kiosk setups, bars docked to an edge.
Accessibility: screen readers, magnifiers, on-screen keyboards.
Remote access to the whole session, x11vnc style.

The replacement exists and it's called portals — xdg-desktop-portal — plus the matching Wayland protocols. It's not a catastrophe, but it is work, and that work has to go into somebody's sprint.

What I'd do this week if I ship Linux desktop software

Take an honest inventory of what you touch. If your product is a web service, none of this concerns you. If you ship a desktop client — Qt, GTK, Electron, Java/AWT — go through the six items above and tick the ones you use. At least one usually comes up, normally global shortcuts or screen capture.
Test on a real Wayland session. Careful here: if your CI runs UI tests on Xvfb, it will keep going green forever, because Xvfb is X11. That green tells you nothing about what the user will see on 6.8.
Move shortcuts and capture to portals before a customer files the ticket. The cost is the same now as in January; the difference is who opens the ticket.
If you offer remote support on Linux desktops, check the tool. That's where I've seen the worst surprises: the viewer connected to the X11 session and needs a different route under Wayland.

The Klassik lesson is about maintenance, not nostalgia

The interesting part of Klassik isn't the K menu. It's the engineering call: instead of keeping Qt 3 code alive, they reimplement the behaviour on a supported base. The contract with the user — where everything is, what happens when you click — is preserved; the internals are from 2026, which is why Wayland and fractional scaling come for free.

Compare that with the alternative. Trinity (TDE) has spent years maintaining its own Qt 3 fork, called TQt, and it does it well: the latest release is 14.1.6, from April this year, and its wiki lists fourteen distributions with current packages. That's heroic and it's real, but it's a permanent cost. At the other end sits Katana, a KDE 4 fork whose repository shows no commits or releases in the past three years. A fork is a promise of indefinite maintenance, and most promises like that don't get kept.

When a client asks me for "the same thing I had, but working on the new stack", the answer is almost never to drag the old stack along: it's to respect the behaviour and rewrite on top of something that's actually supported.

What doesn't change

X11 isn't dead. What's being removed is the Plasma desktop session. The protocol is still there and your X11 applications still run under XWayland. An ssh -X to open a single remote application carries on working.
Your distribution decides, not KDE. Long-term support releases will stay on Plasma 6.7 or older for years. Nobody is pulling the X11 session out from under you on 14 October.
There are escape hatches. A recent fork, Sonic, keeps its own KWin variant — Sonic-win — with X11 support, and other desktops haven't made this move either. An escape hatch isn't a long-term plan, but it buys months.
On servers this is irrelevant. There's no desktop to remove.

If you work on backend or web, close the tab and relax. If you ship anything with windows in it, the real work is the portals, and it's better done calmly than with an angry client on the phone.

Sources: KDE's announcements and release notes and The Register's report on the beta; for Trinity, its official site.

Any questions, tell me and we'll go through it.

Best, Vicente.