In link-state routing, routers distribute descriptions of their local links so that participating routers can construct a common topology map.

Each router then computes preferred paths locally, commonly using Dijkstra's shortest-path algorithm. If a link fails, a new link-state advertisement is flooded and routers recompute routes from the updated graph.

Compared with distance-vector routing, link-state protocols exchange richer topology information and require each router to maintain more state, but they can converge without relying only on chains of neighbors' distance claims.

For example, once router A knows the weighted topology of routers and links in its routing domain, it can run Dijkstra from itself and install next hops corresponding to the resulting shortest-path tree.

The separation is important: flood topology information first; compute routes locally second.