State of multi-player Wayland

Original Article ↗ Hacker News Discussion ↗

Wayland’s emerging support for multiple independent mouse cursors and “seats” is opening up new possibilities such as true multi-user collaboration on a single machine, multi-pointer gaming, and separating touch or tablet input from the primary cursor. Commenters contrast this with X11’s older Multi-Pointer X extension, note that toolkits and applications often assume a single focused window per user, and explain why adding full multi-seat support is significantly more complex than just drawing extra cursors. The thread also touches on related gaps and workarounds in the modern Linux desktop stack—like screen sharing, remote desktop, and virtual display layouts—arguing that more ambitious input and seat models could unlock better everyday features.

Wayland multi-cursor / multi-seat vs X11

  • Several comments note X11 has supported multiple cursors via MPX since ~2009, but it’s fragile: breaks app focus, dragging, and can crash toolkits.
  • Wayland’s “seat” abstraction allows grouping input devices (pointer, keyboard, touch, tablet) per seat, each with its own cursor and focus.
  • Some see the article more as about “multi-player desktops” (multiple people collaborating) than just multiple pointers.

Toolkit and focus model challenges

  • Many GUI toolkits and apps assume:
    • A single focused window per application.
    • A single keyboard focus and pointer.
  • This breaks for multiple seats: you need per-seat focus and possibly per-seat active-window APIs.
  • People doubt that typical GTK/Qt applications, often struggling even with multi-touch, can handle multi-seat well without substantial library work.

Use cases and real-world setups

  • Pair programming and local multiplayer: separate mice and keyboards mapped to separate seats, sometimes already done with sway and similar compositors.
  • Touchscreens: assigning a touchscreen to its own seat lets users scroll one window while typing into another without changing pointer/focus.
  • Drawing tablets and MR/VR: common desire not to have tablet/touch move the “main” cursor; some are surprised that current multi-cursor tablet behavior feels like a bug.
  • Ergonomics and experimentation: split keyboards with trackballs, multiple mice for alternating hands, imagined dual-cursor desktops and cursor “locations” to jump between.

Screen sharing, PipeWire, and complexity

  • Disagreement on how “out-of-the-box” Wayland screen sharing is:
    • Some report it working fine on modern desktops.
    • Others highlight required infrastructure (dbus session, portals, PipeWire) and argue missing any piece breaks sharing, unlike simpler X11 setups.
  • Related discussion touches on lack of a core Wayland protocol for screen sharing vs compositor-specific/testing protocols.

Remote desktop, seats, and RDP-style mobility

  • One long subthread discusses improving Linux multi-physical-seat support and achieving Windows-style fast user switching with RDP.
  • Current logind model ties sessions immutably to seats; proposed approach involves moving sessions between logind sessions and compositors that can hot-swap backends.
  • Non-systemd environments are flagged as an open, “unclear” area.

Other tangents

  • Web PointerEvents support multiple pointers but still assume one “player” (single focus/typing).
  • VNC is debated: called “ancient,” but others point to modern GPU-accelerated capture/encoding and better clients to reduce latency.
  • Minor issues raised include mobile rendering bugs on the article site and desires like pointer inertia on trackpads.