Event Sourcing

[expede]

Summary

In a previous iteration of /design/README.md, we had planned on implementing event sourcing.

Event Sourcing

Instead of recording the createdAt and latest modifiedAt, the Fission server records immutable events. This stores more data in the database, but has multiple advantages:

• Schema migrations can be reconstituted performed directly or by playing over the log
• Post hoc analysis and debugging is possible
• Nondestructive updates makes fixing bugs in-place significantly easier

As the server is currently very focused and lightweight, we can punt on this for now. This pattern helps with several things that we expect to run into eventually, but doing so now may be premature optimisation:

• Schema changes
• Instrumentation
• Moving parts of the model to other services (i.e. a distributed event service that can materialize across multiple focused services)