Developer shares lessons from building a sync engine for a zero-dependency client-side database

A developer spent three months building an offline sync engine for ctrodb, a client-side JavaScript database they had been developing for about a year. The initial sync approach — a simple push-pull merge — failed when the same record was edited on two devices simultaneously, silently overwriting data with no recovery option. To fix this, the developer built a change-log system that records every create, update, and delete operation inside the same database transaction, preventing orphaned or lost changes. The sync engine uses cursor-based pagination for pull requests and batched pushes, with a conflict resolver that was rewritten three times before reaching a workable state. The project is open source and available on npm and GitHub, with documentation hosted on Vercel.