Network Routing

Network routing is the distributed decision-making process by which data packets find paths through a network topology without a central map. Each node makes local forwarding decisions based on partial information — routing tables, link-state advertisements, distance vectors — and the global behavior (end-to-end connectivity) emerges from the accumulation of these local choices. The discipline is one of the purest engineering instantiations of stigmergy: nodes leave traces (route updates, link metrics) in a shared environment, and other nodes respond to those traces without ever coordinating a global plan.

The design problem is ancient in form and modern in scale: how do you coordinate behavior across agents who cannot see the whole? The answer, as in markets and termite nests, is to make local information globally legible through a compact signal. In routing, the signal is the cost metric — a scalar that compresses bandwidth, latency, congestion, and reliability into a single number that guides packet flows. The result is a self-organizing system that adapts to failure, topology change, and traffic shifts without human intervention.

But the compression is lossy. Routing protocols know nothing of application requirements, user priorities, or security context. A video stream and a software update receive the same treatment if their packets carry the same destination. The metric is legible to the network, not to the user. This is the price signal problem in a different register: the system coordinates efficiently on dimensions it can measure, and silently fails on dimensions it cannot.