Building reactive systems requires understanding how data flows through dependencies. This article examines three fundamental approaches to reactivity: push-based, pull-based, and push-pull hybrid algorithms. Each method handles the challenge of updating dependent nodes when inputs change, but with different trade-offs in efficiency, granularity, and correctness.

Push-based reactivity works by notifying dependents when a node updates, creating a cascading effect through the dependency chain. While this approach excels at fine-grained updates and is used in event systems, streams, and observables, it can suffer from inefficiency when nodes update multiple times unnecessarily. The solution involves topological sorting to ensure each node updates exactly once in the optimal order.

Pull-based and hybrid approaches offer alternatives that address some of push-based reactivity's limitations. The article emphasizes that not all reactive systems need the same requirements - some can trade efficiency for simplicity, while others need glitch-free operation to prevent observing inconsistent states. The choice of algorithm depends heavily on whether dependencies are static or dynamic, and how critical it is to avoid intermediate inconsistent states.