Building for Bad Internet: The Architecture Behind Apps That Still Work

Short summary

Apps that work perfectly on a developer's laptop often fail for real users on unstable mobile networks in emerging markets. This article argues for offline-first architecture where the app treats the network as an unreliable service, saving actions locally first, updating the UI immediately, and syncing via a durable mutation queue when connectivity returns. It breaks down a six-part system covering local databases, repository layers, sync engines, and conflict resolution that impacts database design, API design, authentication, and error handling across the entire stack.

• •Treat network as unreliable: save locally first, update UI immediately, sync when connectivity returns
• •Six-part architecture: UI, local DB, repository, durable mutation queue, sync engine, remote API
• •Connectivity is not boolean — high latency, packet loss, and partial failures must all be handled

Generated with AI, which can make mistakes.