Are Rewrites always a Bad Idea?
summary
A C++ item response theory service is replaced with a Kotlin + Spring Boot implementation that ports an R model into plain Kotlin data classes, uses standard math (exp, ln) instead of heavy numerical libs, and validates against researcher-provided input/output cases; development splits into three deployable chunks following a strangler pattern, the team runs simulations with exposed internal APIs, and the legacy instances are decommissioned old instances—the rewrite succeeds and runs 100% in production.
AI-written summary · about
chapters (17)
- 0:00Introduction
- 1:02Speaker background
- 2:07Motivation for rewrites
- 3:11Item selection service case
- 10:30Why rewrite candidate
- 11:32Three reasons to rewrite
- 13:15No small steps available
- 16:43Missing team skills
- 20:28Small and defined scope
- 21:55Strangler pattern
- 24:46Spikes for planning
- 27:18Audit insights
- 31:11Model changes
- 33:59Kotlin and modeling
- 43:40Growth lessons
- 49:39New challenges
- 50:42Rewrites good idea?
