How Not to Destroy Data
summary
Argues for a data model that records immutable facts so state changes become inserts: only inserts no updates no deletes, identities come from natural keys or client timestamps, mutable properties move to version rows linked by a predecessor graph for mutable fields and current state is whatever version has no successor. Shows concrete SQL patterns—unique constraints, WHERE NOT EXISTS queries for “not deleted,” timestamped deletions/restorations—claims this yields easy sync, offline-first clients, no locking, and conflict detection.
AI-written summary · about
chapters (13)
- 0:00Intro
- 1:55CRUD
- 4:22Audit Log
- 10:50Event Sourcing
- 22:47Identity
- 30:33Select Where Not Deleted
- 32:07Deletion
- 39:02Select Where Not Superseded
- 43:48Mutable Properties
- 50:26Workflow
- 51:28Occasionally-Connected Apps
- 52:17Active-Active Clusters
- 54:00Microservices
related talks
18mGermany's Next Top Data Model
44mPolyglot Data
23mState in a stateless world: data serving in the cloud
17mTop 10 transactional database mistakes, and how to avoid them
45mData Modeling in an Event Centric World
46mData Integrity 2.0. The biggest advancement in data quality in 30 years!
47mRethinking Relational Database CRUD Design Patterns Since We Live in a Big Data World
34mKeynote: Things Databases Don't Do (but should).