Improving cursor rendering on Wayland

Original Article ↗ Hacker News Discussion ↗

Storage, Size, and “Wasting Space”

  • Some argue XCursor themes taking megabytes on disk is irrelevant in the “TB era” and compression can mitigate it further.
  • Others counter that installer images and live media still have tight size budgets, so bloat matters.
  • There’s a broader tension between “don’t waste effort on micro-optimizations” and “death by a thousand inefficiencies,” with OS and game bloat cited as warnings.

Wayland Cursor Model vs X

  • Original Wayland design has clients provide cursor surfaces, with the compositor just compositing them.
  • Critics find it “insane” that ordinary apps must handle theming, sizing, and rendering instead of just requesting a standard cursor type.
  • The newer wp_cursor_shape protocol lets clients request named cursor shapes, seen as better for most apps while still allowing custom cursors for games and drag-and-drop.

GTK/GNOME and Protocol Adoption

  • Discussion notes GTK as a hold‑out on wp_cursor_shape, though issues/MRs do exist.
  • Some see GNOME as frequently blocking or delaying Wayland extensions; others say consensus-building is slow but necessary and GNOME doesn’t have formal veto power.

Hyprcursor Format and Adoption Concerns

  • Hyprcursor (from the Hyprland ecosystem) is pitched as a superior, vector-capable, modern cursor format.
  • Concerns: dependence on a bespoke config language (Hyprlang), reduced cross-desktop appeal, and risk of yet another incompatible cursor standard.
  • Suggestions include: using a common config format (e.g., XDG-style INI), publishing a spec, and providing shims for existing XCursor APIs.

Vector vs Raster and SVG Debates

  • Many like the idea of vector cursors for high DPI, but worry full SVG is overkill: complex, heavy dependencies, even scripting support.
  • Some advocate a simpler dedicated vector format (e.g., citing Haiku’s icon format) or a constrained SVG subset.

Broader Wayland Design & Governance Issues

  • Wayland’s “minimal core, everything as extensions” is praised in theory but criticized for fragmentation when compositors can’t agree on extensions.
  • There’s recurring debate over client-side vs server-side decorations, with arguments about synchronization, VRAM use, and UI consistency.
  • Speculation about corporate (especially enterprise vendor) influence appears, but others rebut conspiracy theories and emphasize diverse contributors and real technical constraints.

Miscellaneous

  • Requests appear for features like color-inverting cursors, noting past patent concerns but likely expiration.
  • Some comment that this all feels like re-solving long-standing Xorg problems instead of building on them.