he prevalence of multimedia applications, such as IPTV, video conferencing, online multi-player games, over the Internet has manifested the importance of developing a scalable and efficient networking protocol to disseminate shared data to widely distributed destinations. Recently, Bloom filter based multicast has been proposed as a solution for the need, where routing information is carried in the multicast packet in the format of a Bloom filter. This family of protocols has been gaining an increasing interest in the field due to their desirable scalability over predecessor multicast protocols. This article gives a tutorial of the Bloom-filter based multicast protocols by answering some important questions. What is the fundamental issue hindering the success of the traditional multicast protocols? How has the research on multicast evolved all the way to the current state? What specific benefits can the Bloom filter based multicast bring? How could the Bloom filter based multicast benefit the multimedia applications over IP? What could be the open research issues in the field?We reveal the impetus behind the evolution path of multicast mechanisms, in order to help understanding how multicast has developed into the current state. The evolution of the multicast mechanism in fact epitomizes the evolution of Internet itself. The early IP multicast is to construct a tree structure by networking routers in a single flat topology [1]. As the domain-based hierarchical architecture adopted by Internet, multicast protocols originally designed for the flat network are extended into the inter-domain scenario, where improving IP multicast scalability has been put many efforts but still can not be thoroughly resolved. This is because the networking node involved in IP multicast has to install a forwarding state for each multicast group it is supporting, and the associated messaging and memory overhead will grow linearly.The Bloom filter based multicast takes a source routing methodology. The fundamental design is to carry routing information in the multicast packet in the format of a Bloom filter, which is a randomized data structure for representing a set and supporting membership queries [2, 3]. The routing information will be extracted by the intermediate networking node and used to compute appropriate packet copies and output interfaces. Since each node has no need maintaining the group-specific forwarding states, the Bloom filter based multicast has desirable scalability. According to the information encoded into the Bloom filter, these multicast protocols can be categorized as tree oriented [4]and destination oriented [5], where the former encodes the multicast tree branches while the latter encodes receiver domains' network prefixes. This seemingly slight difference can incur significant distinctions in protocol design and performance, which will be described later in a tutorial manner.Multicast is normally considered the most bandwidth efficient way for delivering multimedia services, but how to utilize the particular ...