Who Authored the Kernel? Recovering Token-Level Authorship Information from Git
summary
Creates a reproducible view repository that tokenizes every C/Java/Python file using an island parser from University of Kent, rewrites history into Git and links commits to mailing-list reviews so each source token maps to the commit that introduced it; exposes token-level blame through ordinary Git/GitHub tooling. Applied to Linux up to 4.7, tokens exceed lines by ~7.6×, 5.1% of tokens predate BitKeeper, 48% of commits churn <10 tokens. Aggregate author rankings barely change, but token traces reveal many tiny surgical edits and hidden provenance.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 1:18archival framework
- 2:35git blame
- 3:50git history
- 15:49Linux history
- 23:03What code has survived
- 24:43The while loop
- 25:49Another file
- 29:28AB
- 32:09Colonel
- 36:02Other Aspects
- 37:45Churn
- 39:27Conclusion
- 39:57Proposal
related talks
35mBreaking the Chain: An Attacker's Perspective on Supply Chain Vulnerabilities and Flaws
41mThe CVE Blind Spot: Defeating "Hidden EOLs" and Repo Jacking with Engineering Triage & Code Diet
11mLightning Talk: Repurposed Purpose: Using Git's DAG for Supply Chain Artifact Resolution
46mGit internals: Why trees are a developer's best friend
41mKeynote: Wrapping Your Head Around Git
39mOpen Source Activities to Enhance Corporate Value -- From Technology Management To Infrastructure...
33mCode Archaeology: Becoming Indiana Jones
09mAspects of Dependable Linux Systems