I'd like to re-start the discussion of this
we can totally do this async
I think I liked this definition:
P-critical issues are potentially release blockers. They are going to be treated like a blocker but once we are close to the release we would re-consider if it worth blocking the release on a case by case basis.
something like that but better written
Yeah I think it would be too draconian to be literal about release-blockers here
P-critical for stable regressions (unless we think they aren't bugs), stable soundness holes, particularly for tier-1 seems pretty reasonable for example
stable soundness holes that we expect to arise in practical cases, at least
I'm pretty sure there are stable soundness holes that we have let lie unresolved and can continue to remain so
(Can't say I'm happy about the situation with the many soundness holes btw...)
(it undermines "Rust is a safe language")
I think for soundness holes we want to assess how likely it is that people are relying on this behavior and how long the hole has been present -- our goal is always to minimize the chance of it being super hard to fix. So if we things people are very likely to be relying on, but maybe not yet, that is worth investigating and trying to fix, etc.