Blazingly Fast or Blazingly Hyped?
summary
Compares actual Rust rewrites against battle-tested originals with numbers: ripgrep 10× faster on a cargo tree search, a PNG crate ~2× faster than the C equivalent thanks to auto-vectorization and streaming deflate, but some rewrites are 60× slower in certain piping cases; uutils multi-call binary shrinks a suite from 73MB→14MB; rewrites commonly take months to years (small months, medium 1–2 years, big 2–5 years and often 2–3× longer), introduce new bugs (Cloudflare unwrap, datefmt and async-tar regressions), and succeed only when safety, parallel performance, and platform support justify the cost.
AI-written summary · about
chapters (8)
- 0:00Introduction and agenda
- 1:26Defining the RIIR movement
- 6:01Categorizing rewrites
- 10:45Performance and analysis
- 15:06Challenges of rewrites
- 20:45Drawbacks and risks
- 26:22Conclusions and best practices
- 29:53Final remarks
