Designing Data-Intensive Applications

The chapter on replication is the one that changed how this site is built. Once you accept that a browser tab is a replica and not a view, a lot of ordinary web architecture stops making sense: there is no authoritative copy to fetch, only peers with different amounts of history.

What it settled

Every renderer here reads from the local replica and writes to it too. The server is a peer with better uptime — it persists, broadcasts and enforces permissions, but it does not own the truth about what a paragraph says.

The part still being read: leaderless replication, which is the honest name for what a room full of visitors editing one document is doing.