The Semantic Schema Version
summary
Argues that publishing a consistent database versioning policy — using a migrations model checked into source control + a schema_versions table that records who ran each script — lets application agents detect incompatible schema changes and pause work. Demonstrates a proof of concept where two worker agents read tasks, a migration adds a customer email column and another migration removes fax support, the history row shows version and author, and an agent safely stalls or resumes so the update deploys without downtime.
AI-written summary · about
chapters (12)
- 0:00Start
- 5:19Applications are ephemeral
- 8:47Migrations model
- 9:40State v Migrations
- 13:51What is an API?
- 14:42The database schema is an API
- 15:59Breaking promises
- 19:05Serriver: Patch version
- 19:46Serriver: Minor version
- 23:16Implementation
- 26:25Migrations as part of continuous deployment
- 28:10The goal - safe deployment
related talks
31mHelp! My Database Is The Bottleneck
37mDatabase Schema Migrations with Zero Downtime
33mDatabase Schema Migrations with Zero Downtime
21mApplying Continuous Delivery to Database Schema
38mVitess: Schema Changes at Scale
36mUp And Down Again: A Migration's Tale
57mDjangoCon 2008 Panel: Schema Evolution
56mSane schema migrations with Alembic and Postgres