Multicasting offers tremendous savings over unicast transmission because it allows receivers to share bandwidth. The same content is transmitted only once by the sender, but is received by multiple receivers. The authors of this paper consider the practice of charging all receivers equally, based on a flat rate or usage-based pricing scheme, to be inadequate. Instead, they propose an auction mechanism with bidding.

A multicast session can be sent at multiple pre-determined rates, either using a separate multicast group for each rate (split session paradigm), or aggregating several layers (layered paradigm). A receiver places a bid per rate, indicating its willingness to pay for delivery at that rate. The marginal cost mechanism is used to determine the set of receivers that obtain the content at a particular rate. The price charged to an accepted receiver can be lower (but not more) than the original bid for that rate.

The authors first demonstrate that the problem can be solved with acceptable effort if two simplifying, but not very realistic, assumptions are added: multicasting introduces no additional costs, and there is a single source and fixed multicast tree for all groups and layers. If one of the assumptions is dropped, the problem becomes more difficult to solve (NP-hard).

There seem to be additional severe restrictions to this paradigm, however. The mathematical programming approach that is used also requires that all information (receivers, rates, bids) be known in advance, and not change afterwards, with the consequence that receivers cannot join or leave the multicast session dynamically. Also, receivers pay for a particular rate, but what if traffic metering indicates that this quality of service (QoS) was not obtained? How much then does the receiver have to pay for this lower rate? Or do we have to recompute the whole algorithm?