10 Billion Downloads: Insights and Trends in Open Source
summary
Dataset covers over a trillion downloads (about 2 billion/day) from registries and SDK hooks like scarf.js postinstall telemetry, with ~100 million anonymized origin IDs and gateway-level metadata. Findings: DockerHub dominates, VPNs are ~0.1% of traffic, ~75% of national governments and 96% of Fortune 500 appear in the sample, and 95% from top 1% of packages. Typical downloads-to-unique ratio is average 100:1 downloads to uniques, best packages ~15:1, and pip’s verbose user agent proves far more actionable than Kubernetes’s generic header.
AI-written summary · about
chapters (18)
- 0:00Open source insights
- 1:24Collecting trillion downloads
- 4:34Measuring open source users
- 6:39Global usage patterns
- 8:47Companies using open source
- 10:12Public cloud preferences
- 10:55Container registry traffic
- 11:59Power law in downloads
- 13:02Why download counts mislead
- 16:32Automated systems inflating
- 17:40User agent metadata best practices
- 19:03Myth of latest version upgrades
- 21:11Open source adoption funnel
- 22:12AI vs database adoption
- 23:55Data-driven decisions in open source
- 26:44Commercial open source sustainability
- 27:46Key takeaways
- 28:33Q&A
related talks
37mOpen Source Success Learnings from 1 Billion Downloads
37mThe State of Open Source Security
01h 03mWhat can we learn from 750 billion GitHub events and 42 TB of code
49mTrustworthy Open Source: The Consequences of Success
24mA Beginners Guide to Open Source Metrics and Analysis
32mExemplars, Laggards, and Hoarders: A Data-driven Look at Open Source Software Supply Chains
30mData Tales from an Open Source Research Team
28mKeynote: Incremental Revolution - What Docker Learned from the Open-Source Fire Hose