Namespaces and Regions

Namespaces in Erebus are built on top of Durable Object namespaces.
Here’s the idea in simple terms:

• Per-region consistency:
  Each region has its own namespace, identified like id:{region}.
  People in the same region get both consistency and availability, since their messages stay local.

• Cross-region communication:
  When one region needs to talk to another (e.g. id:{region_1} -> id:{region_2}), the traffic goes through a Redis-backed coordination layer.
  These namespaces are grouped under "shards" to keep things scalable.

• Trade-offs:
  Within a region → you get strong guarantees (fast and consistent).
  Across regions → you get AP (availability and partition tolerance), but no strict consistency, and latency is naturally higher.

This setup balances the CAP theorem by separating guarantees regionally while still allowing global communication.