Driver and subsystem rewrites that began as experiments are now in release builds. Early data from one vendor shows memory-safety bugs in new code dropping to near zero, while the legacy C codebase continues to generate the majority of critical fixes.
The long tail
Nobody is rewriting a kernel wholesale. The strategy is to write new code in a safe language and leave old code alone until it needs touching. That means decades of mixed codebases, and tooling for the boundary is where the interesting work is.
Key takeaways
- New safe-language code shows near-zero memory bugs
- Legacy C still produces most critical fixes
- Boundary tooling is the active research area
This story is part of our ongoing coverage of cybersecurity. We will update it as more details emerge.
