To Re-write Or Not To Re-write
summary
Rewrites usually fail unless the risk is low, data is small, and migration is manageable. Example outcomes: an internal MS Vacation rewrite takes ~12 months and succeeds after careful migration; a SaaS rewrite runs 18 months, doubles maintenance (legacy + new), and largely fails under a big client contract; a startup rewrite of a 10-customer product takes ~9 months and becomes the most stable release. Key tactics: use strangler fig pattern, break database coupling, and enforce separate read/write APIs when terabytes of data make lift‑and‑shift impossible.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:51The case against rewrites
- 5:48First success story
- 15:02A total failure
- 22:06Decomposition strategy
- 29:54Global expansion lessons
- 37:20Startup architecture
- 41:45Summary and conclusions
- 45:06Q&A session
related talks
36mRewrite, Rework or Refactor
36mHow to sell a big refactor or rewrite to the business?
50mThree Rewrites, One Startup - An Architect's Journey
42mSo You Want To Rewrite That...
40mHow to sell a big refactor or rewrite to the business?
47mLegacy Software: Really a Problem?
20mLegacy Migration... Differently
01h 17mRefactoring the ROI of Software