Challenges of Managing CoreFX repo
summary
CoreFX faces tangled dependency chains across ~30 repos that force 8–12 hour product builds and generate roughly 600 GitHub notifications per day; teams merge key pieces into dotnet/runtime, create a manual per-area notifier and split work into 60 areas so owners keep queues ~200 issues, use an ML labeler with ~70–80% accuracy, run monthly previews and accept daily/bleeding builds from customers like Bing, and see concrete wins such as cutting ~100 networking issues and ~50% closure rates on worked-up sweeps.
AI-written summary · about
chapters (21)
- 0:00Intro
- 1:39Dotnet Core
- 7:31DotNet Block
- 9:27Build turnaround
- 13:55Merge key repos
- 16:32Delete old repos
- 17:20Runtime ripples
- 18:42Traffic
- 19:24Collaborators
- 20:37Solutions
- 23:29Temporary tooling
- 24:56Repo management
- 26:07Pain points
- 27:55Optimizing labels
- 29:56What is triage
- 31:04Understanding triage
- 33:21Closing issues
- 37:09When to say no
- 43:39triage tools
- 47:31culture challenges
- 55:35contributing to community
related talks
41mAt the Heart of a Core Team
20mJuggling the dual life of an enterprise open source project
45mTools and Infrastructure for Large Teams
36mWhale Care 101 - How We Organize A 200+ Devs Project
46mBuilding Developer Community One Pull Request at a Time
33mOpen-Source Maturity in a 160-year-old organization: Technical and Legal hurdles
45mFinding problems quickly in a quickly evolving opensource project
42mBuilding DevX Teams, my story