Most sync engines are an extra layer between clients and databases. But why couldn't data sync entirely replace databases and most of a typical backend? We've been building a framework to prove that for 5+ years, but realised that we're just re-inventing the parts that make up a database.

In this talk, we show you how embracing this has helped us provide new takes on must-have database features, resulting in a general-purpose tool that is more distributed (with multi-region/offline reads & writes and high fault tolerance) and more composable than traditional databases (letting you trade off properties like low-latency eventual consistency vs MVCC-style global transactionality on a per-collection basis), while providing entirely new features like granular data branching, collaborative rich-text datatypes, role-based permissions and end-to-end encryption.

You'll also see how you can still adopt it server-side-first, like a traditional database - with real-time streaming and local-first client-side use as opt-in optimisations.