I'm considering a seven-day delay for new PyPI and npm releases to reduce exposure to short-lived malicious versions. Is a week reasonable, and how do you handle urgent fixes?
Do you wait before installing new Python package releases?
Answers
A cooldown is a useful layer, not a guarantee. Some malicious releases have been removed within hours, so a week would have kept those particular versions out. It won't catch a bad release that stays available longer or one that nobody detects.
Seven days also means waiting seven days for fixes unless you have an exception process. Make that a deliberate per-package override rather than asking everyone to disable the policy when something urgent comes up.
Reply to this answer
Three days is another option if a week feels too slow. There's a real tradeoff between giving a new release time to be scrutinized and delaying a patch you need. Pick a window your team can live with, then make sure exceptions are practical.
Reply to this answer
I'd enforce the age limit in tooling rather than rely on people remembering it. A caching proxy can apply a shared policy across package managers, but check whether every install path actually goes through it. Lockfile-based downloads and tools with their own source configuration can complicate that.
For Python-only setups, uv supports an exclude-newer setting and recent pip versions have an uploaded-prior-to option. Check your tool version and test the behavior with your existing lockfiles before treating either as an organization-wide control.
Reply to this answer
If a full week is too disruptive, pin dependencies and avoid taking releases on the day they appear unless there's a reason. It's a smaller delay, but it reduces the chance that every fresh environment immediately picks up a bad update.
Reply to this answer
Add your answer
Answers are reviewed before publication.
q.381!tavloro:
Poetry may need its own source configuration and a re-lock. Also check where locked artifact URLs point, not just where dependency resolution happens.