A method and apparatus in a multicast transmission system for routing packets including a scalable architecture that supports QoS. Redundant state information is minimized in the MOSPF multicast forwarding cache of each router in the communications system, thereby rendering MOSPF considerably more scalable. In a particular embodiment, cache redundancy is minimized by using a class of graph compression algorithms to solve a graph compression problem. In another embodiment, the forwarding cache is further consolidated by separating state information pertaining to individual source routers in a multicast group from state information pertaining to the entire multicast group.