Source-implemented constraint-based routing with source routing enables traffic engineering to be performed and reservations to be made within a domain without requiring constraint information and reservation information to be disseminated to all nodes in the domain. In one embodiment, a focal node maintains a table containing metrics of links and connection reservations through the domain. When a connection is to be added, the focal node determines a path, given the constraints reflected in the table, and allocates resources on the links forming the path. The table is updated, and the path is used to generate headers for traffic associated with the connection. Traffic from the nodes toward the focal point follow the reverse path. If a node or link fails, connections carried through the failure are identified, reservations on links associated with the connections are released, and new reservations are made taking into account the new network topology.