Jump to content

Sprints/Plasma/2025/Topics/Telemetry

From KDE Community Wiki

Problems with the status quo

  • No way to change what we ask for without technical changes
  • Viewer UI is restricted so no data is publicly available, and it requires writing scary SQL to use


Actionable ideas

  • In the interim, massively increase the data collection interval for Plasma, which is super frequent right now
  • We have functionality for surveys; we could leverage this to make something like the Steam hardware survey and then just do that
  • Generate a static site with the latest data, and then aggregated data (but not raw data) can be public
  • Don't go with Privact right now since it needs more proving out in production first


Follow up testing with Apache Superset

  • Installation with docker was relatively trivial
  • It's very easy to make new graphs. I recreated all of plasma's dashboard in about an hour.
  • It supports a concept of "virtual databases" (Like a "view" in SQL) so any complex parts like filtering the latest result only have to be done once
  • We can set up pre-defined "filters" that apply to the current dashboard (i.e if we have a graph of screen resolutions) you can then filter by "only on Wayland" and "only on 6.3" that viewers can use without us needing to make more graphs ahead of time.
  • Similarly there's "cross-filtering" between graphs which is basically the same idea.
  • It supports OAuth, but the docs for setting this up are not great
  • It allows different roles for viewing uneditable dashboards vs viewing everything else.

Negatives:

  • Still had to write custom SQL for handling the separate table for screen resolutions, but at least only once.
  • On the dashboard you can access raw data (maybe it's possible to lock it down?)