Traditional multicast routing is protocol-dependent. It is best suited for local area networks and less suited for sparsely distributed groups. This paper proposes a protocol independent multicast (PIM) architecture that compliments the existing multicast routing algorithms and efficiently establishes distribution trees for sparsely distributed groups.

The paper starts with a brief description of existing multicast routing algorithms and points out their limitations. The authors provide appropriate illustrations to clarify their discussion. They then discuss a more recent protocol, core-based tree (CBT). CBT addresses the broadcasting problems (sparsely distributed groups) encountered in earlier protocols, but introduces a traffic concentration problem, because CBT imposes a single shared tree for each multicast group.

The authors then discuss the two main objectives of the PIM protocol: to avoid the broadcasting and traffic concentration problems encountered in multicast and CBT, respectively, and to support sparsely distributed groups efficiently and effectively. Some of the other advantages of PIM over CBT are simplicity, versatility, and an interface check of incoming data to avoid data looping.

The authors outline the requirements of the PIM protocol--efficient sparse group support, routing protocol independence, robustness, and interoperability--and discuss how to fulfill these requirements. Then they point out several open issues (information aggregation, interaction with RSVP, and policy-based and TOS routing). Effective resolution of these issues requires more research and development.

The paper is well written. The authors must be commended for producing a much-needed concept in a timely manner.