Beyond bus-factor: data-visuals & algorithms to better understand Apache communities
summary
Implements a small Go application plus Python notebooks that compute bus/pony factors from git and GitHub activity, and finds the standard file‑ownership truck‑factor method misses long tails and mailing‑list work. Introduces a continuous, top‑normalized contributor score, per-person contribution timelines and timezone heatmaps to reveal paid vs volunteer patterns, project lifetimes, and surprises like Ambari’s collapse after Hortonworks consolidation. Prototype runs across Ozone, Spark, Kafka, Hadoop, Flink, Dubbo and Maven, yields stable month‑over‑month signals, and is intended for migration into Kibble.
AI-written summary · about
chapters (18)
- 0:00Introduction and context
- 1:14Why measure Apache communities
- 3:38Apache Ozone project overview
- 4:46Limitations of bus factor
- 6:32Snowball factor and its flaws
- 9:29Pony factor definition and use
- 13:39Problems with pony factor metrics
- 15:51Alternative contribution analysis
- 19:01Time window impact on metrics
- 20:03Contributor activity visualization
- 23:55Pull request dynamics in communities
- 25:22Hadoop contributor lifecycle
- 27:50Paid vs voluntary contributors
- 29:14Motivation patterns in projects
- 31:21Contribution time zones analysis
- 35:53Weekend contribution patterns
- 37:19Limitations of statistical metrics
- 38:01Conclusion and future directions
related talks
40mDiscovering Emerging Open Source Communities Through...
19mOSPO Data Pyramid
48mGraphing Zephyr Development and Project Health and How Visualizing Data Help with Gr...
43mBoF: What’s Interesting about OSS Developer Communities to You? A...
01h 04mWhat can we learn from 750 billion GitHub events and 42 TB of code
27mCommunity Data is Not Community Metrics
31mNon-Code contributions (...)
32mUncovering New Open Source Communities Graphically