Application level multicasting is increasingly being used to overcome the problem of non-ubiquitous deployment of IP multicast across heterogeneous networks. To the best of our knowledge, this paper is among the first papers [A comparative study of application layer multicast protocols, Unpublished report; IEEE Network, January/February, 2003] to provide a comprehensive survey of most of the various milestone research work in application level multicast in terms of both breadth and depth. The paper classifies them into different broad categories based on their topology design, service model and architecture to facilitate better understanding of their contributions and discuss their merits and limitations. As these techniques vary widely in their goals, designs, performance evaluation metrics and evaluation strategies, it is impossible to quantify their relative performance. However, this paper is able to provide a comparative insight into their performance through the use of a set of evaluation metrics identified to be common to the techniques and are directly related to their performance and whose data can be derived and inferred from their designs. The metrics used here include scaleability measured intuitively in terms of the size of the multicast receivers it can support, the protocol efficiency in terms of the quality of data paths, control overheads, amount of state information to be maintained at each member node and failure tolerance.