A Tale of Two Feature Flags
summary
Compares two real feature‑flag lifecycles: a faceted search that becomes backend flag coupling causing hard unit tests and a long‑lived flag that leads to a Knight Capital‑style outage. Shows concrete fixes: isolate flags to one location, treat flagged code as new code, and add deadline and maintainer fields with CI that opens branches and escalates after two days. Claims flags speed releases but often become permanent technical debt; demonstrates removal is harder than implementation.
AI-written summary · about
chapters (6)
- 0:00Start
- 5:37Put a feature flag on it!
- 10:12Flags on flags on flags
- 10:45Segmented releases lead to exponential testing complexity
- 20:15Never-ending configuration
- 21:30Unattended flags are high-risk flags
related talks
56mWhen Feature flags go bad
11mFeature flag use cases you haven't heard about yet
41mFeature Flags: Separate Deployment from Feature Release
08mThe Art of Feature Flagging
41mImproving feature flag implementation to achieve flow
50mThe Story of One Feature Flag
06mFeature Flag Foul-ups
19mBeyond the Toggle: Feature Flags as Your Migration Safety Net