Many reliable multicast protocols have been designed that respond to a wide range of user requirements and network environments. These protocols can be differentiated by examining the set of requirements they meet, the architecture they use for interaction among the participants, the mechanisms used to effect this interaction, and the policies that determine how and when to use the mechanisms. The set of multicast protocols can be classified using the user requirements, and the architectures, mechanisms, communications patterns, and policies used to satisfy these requirements. The article provides such a classification, and illustrates it with several example protocols chosen to cover the range of features described. To help identify the components that are common across all classes of reliable multicast protocols such as point-to-multipoint and multipoint-to-multipoint, authors have developed a taxonomy based on requirements, architectures, mechanisms, communications patterns, and policies. This brings out the orthogonality between the various requirements of reliable multicast transport protocols and the ways they are used to satisfy those requirements.