Investigate OS-native and shared-cache thumbnail providers #6

Closed
opened 2026-04-05 19:13:41 +00:00 by LyAhn · 0 comments
LyAhn commented 2026-04-05 19:13:41 +00:00 (Migrated from github.com)

Summary

Evaluate whether Phokus should pivot part of its thumbnail pipeline to OS-native or shared-cache providers for better performance, especially on large libraries and slower drives.

Why

The current generated-thumbnail pipeline has improved, but thumbnail and metadata throughput still appears too slow for large backlogs. This is worth a dedicated investigation rather than continued ad hoc tuning.

Research directions

  • Evaluate humb-rs for Windows/macOS native thumbnail generation
  • Evaluate Linux XDG/Freedesktop thumbnail-cache integration
  • Evaluate llmytoes or similar Linux shared-thumbnail approaches
  • Compare native/shared-cache providers vs current generated-cache pipeline
  • Identify correctness tradeoffs (real thumbnails vs icon fallback, cache freshness, metadata coverage)
  • Identify packaging/shipping implications per platform

Desired outcome

  • Decide whether to keep generated thumbnails as the primary pipeline
  • Or move to a provider model:
    • native/shared-cache first
    • generated fallback second

Notes

This is a future optimization investigation, not an immediate blocker, but it is important enough to track explicitly.

## Summary Evaluate whether Phokus should pivot part of its thumbnail pipeline to OS-native or shared-cache providers for better performance, especially on large libraries and slower drives. ## Why The current generated-thumbnail pipeline has improved, but thumbnail and metadata throughput still appears too slow for large backlogs. This is worth a dedicated investigation rather than continued ad hoc tuning. ## Research directions - [ ] Evaluate humb-rs for Windows/macOS native thumbnail generation - [ ] Evaluate Linux XDG/Freedesktop thumbnail-cache integration - [ ] Evaluate llmytoes or similar Linux shared-thumbnail approaches - [ ] Compare native/shared-cache providers vs current generated-cache pipeline - [ ] Identify correctness tradeoffs (real thumbnails vs icon fallback, cache freshness, metadata coverage) - [ ] Identify packaging/shipping implications per platform ## Desired outcome - [ ] Decide whether to keep generated thumbnails as the primary pipeline - [ ] Or move to a provider model: - [ ] native/shared-cache first - [ ] generated fallback second ## Notes This is a future optimization investigation, not an immediate blocker, but it is important enough to track explicitly.
LyAhn added the wontfix label 2026-06-12 21:54:21 +00:00
LyAhn closed this issue 2026-06-12 21:57:54 +00:00
Sign in to join this conversation.