An interactive study companion for Martin Kleppmann's classic on reliable, scalable, and maintainable systems.
How you encode data determines compatibility, schema evolution, and cross-language interoperability.
Programs in memory represent data as objects and structs. To send data over a network or store it on disk, you must encode it as bytes. The encoding format determines size, speed, and — critically — whether schemas can evolve without breaking consumers.
Forward compatibility: old code reads new data. Backward compatibility: new code reads old data. Field tags (Protobuf) and union schemas (Avro) enable both.
Public REST APIs return JSON with OpenAPI specs. Internal microservices increasingly use gRPC with Protocol Buffers for compact, typed RPC. Kafka pipelines pair Avro schemas with Confluent Schema Registry so producers and consumers evolve independently during rolling deploys on Kubernetes.
Design state syncs over WebSockets with compact binary payloads. Internal services use Protocol Buffers so new canvas fields ship without breaking older editor builds — forward compatibility via tagged field numbers.
// Canva / gRPC — Protocol Buffers with forward-compatible field tags
// message Design { string id = 1; repeated Layer layers = 2; string title = 3; }
const payload = Design.encode({
id: designId,
layers: canvasLayers,
title: "Untitled",
}).finish();
// New field tag 4 added in v2 — old clients ignore unknown tags