Software Engineering Radio - the podcast for professional software developers SE Radio 720: Martin Dilger on Understanding Eventsourcing
13 snips
May 13, 2026 
Martin Dilger, founder and CEO of Nebuilt and author on event-driven systems. He breaks down event sourcing vocabulary and how it differs from streaming, modeling, and storming. Short takes cover event modeling workshops, mapping models to code, slice-based architectures, projections, versioning, common pitfalls, when event sourcing fits, and incremental ways to start with legacy or greenfield systems.
AI Snips
Chapters
Books
Transcript
Episode notes
Persist Projections For Fast Reads
- Persist projections for runtime queries instead of replaying events on every request to avoid latency problems.
- Martin persists projections into familiar stores (e.g., relational tables) so reads behave like traditional architectures while events remain source of truth.
Evolve Events With New Types Or Versioned Upcasters
- Prefer adding new events or versioning events over changing past event shapes; use upcasters to transform older versions.
- Martin suggests CustomerRegisteredV2 for structural change and an upcaster to convert V1 to V2 during processing.
Kafka Misuse Killed An Event Sourcing Attempt
- Many companies blame event sourcing failures on wrong tooling; Martin cites Kafka being misused as an event store.
- He calls Kafka an event streaming highway and stresses it's unsuitable as an event store, explaining why prior projects 'failed' that way.
