flow-architectures

Key Insights

• Event streaming as the next infrastructure layer. Urquhart’s argument: just as the internet commoditized data transport and APIs commoditized request-response integration, event streaming (Kafka, cloud event buses) is becoming the standard infrastructure for real-time data flow between systems. The companies that build on this layer will have structural advantages in responsiveness and data leverage.
• The World Event Web — a vision of global event infrastructure. The book’s speculative endpoint: a universal event mesh where any system can publish and subscribe to real-time event streams from any other, with standardized schemas and discovery. This is analogous to the early vision of the World Wide Web for documents, applied to events and state changes.
• Streaming vs. batch as a fundamental architectural choice. Most business data pipelines are batch — data is collected, then processed, then reported, with latency measured in hours or days. Event streaming collapses this to near-real-time. The choice between architectures has downstream consequences for what business questions can even be asked.
• The producer/consumer decoupling as an organizational benefit. Event-driven architectures allow teams to publish data without knowing who will consume it, and to consume data without depending on the producer’s availability. This reduces coupling between teams and systems, enabling more independent development and deployment.
• Schema registries and event governance as the hard problems. The technical primitives of event streaming are relatively mature; the organizational challenges — who owns schemas, how breaking changes are managed, how events are discovered and documented — are where real implementations struggle and where the field is still developing.

— Drafted from external sources; review and edit to make your own.