How (not) to upgrade a platform
summary
Data from PyPI data mining over three months shows a clear result: the difference in absolute package counts grows (more packages still labeled Python 2 than Python 3), though Python 3 grows faster in relative terms. The author builds a future package and compares classifiers using the max heuristic, documents missing backports (sys.intern, urllib.request), and flags the real blocker: bytes vs str remains incompatible. Case studies (H Day, Apple CPU switch, MagSafe adapter) illustrate that adapter strategies or parallel interfaces work — renaming breaking APIs and backporting do too.
AI-written summary · about
chapters (18)
- 0:00Introduction
- 1:30Deprecation
- 2:02Change its name
- 3:48Unicode
- 5:13Python 3 debate
- 5:59Data mining
- 9:04Pragmatism vs idealism
- 14:46The ugly
- 18:20Traffic accidents
- 19:10MagSafe
- 25:02Is it too late
- 25:12Text data type
- 25:47Backward compatibility
- 26:19Package proliferation
- 26:34Why is it so hard to upgrade
- 27:52Extensions
- 29:52Future Dog
- 30:27Outro
related talks
43mPython 3: ten years later
26mPython 2 to 3: How to Upgrade and What Features to Start Using
28mWhyPy? Understanding Trends Driving the Usage of Python
41mYou should switch to Python 3 (two years ago)
26mEvolve or Die: How I Stopped Avoiding and Starting Loving Python Upgrades
31mPython 3 for People Who Haven't Been Paying Attention
01h 12mPython 3: the Next Generation (is here already)
32mPyPI in the face: running jokes that PyPI download stats can play on you