The Five Demons of Python Packaging That Fuel Our Persistent Nightmare
summary
Python’s pain has five concrete causes: native extensions force C toolchain, ABI and build-matrix headaches (conda vs wheels matter); repository vs distribution clarifies who names, secures and verifies packages (PyPI hosting ≠ curated distro like conda‑forge); no standard app story means shipping to other users lacks a reliable installer; packages as SDKs not runtimes creates transitive bloat and impossible tree‑shaking; and diverse user tribes produce mutually incompatible defaults — plus the practical need for sustainable funding for repos and guardrails.
AI-written summary · about
chapters (9)
- 0:00Introduction to the problem
- 3:16Understanding packaging
- 5:47Native extension modules
- 11:33Repositories and distributions
- 15:58The absence of app delivery
- 17:21SDKs vs. runtimes
- 21:12Community diversity
- 26:46Sustainability and business
- 31:00Conclusion and Q&A
related talks
30mThe Evolutionary Saga of Python Packaging
40mMaking packages and packaging "just work"
25mInside the Cheeseshop: How Python Packaging Works
35mSecuring Python projects Supply Chain
32mInside the Cheeseshop: How Python Packaging Works
24mInside the Cheeseshop:How Python Packaging Works
43m5 Years of Bad Ideas
54mKey Challenges In the PyData Ecosystem and How we Can All Make a Difference