Before describing the way in which interactive temporal random access to a video sequence can be done using the P2PSP, it is important to highlight that this possibility represent the worst working scenario of a P2P network because peers tend to retrieve data that it is only interesting for a small amount of them. In the extreme situation where only a peer requests a speciﬁc part of the sequence, the bandwidth consumption in the source side is equivalent to the bandwidth consumption in the server side of a CS system.
Keeping this fact in mind, temporal interactivity can be implemented using the P2PSP by conﬁguring a set of teams where each one of them broadcasts a diﬀerent time-shifted version of the same content. Thus, a peer which wants to move to a diﬀerent time in the stream, has to leave the current cluster and to move to the team that provides the desired part of the stream.
Obviously, this technique solves the problem of the jumping to a diﬀerent time in the same video, but does not tackle the problem of dealing with the fast forward and rewind actions. In those situations, it is not necessarily to send all the data of each GOP visited by the peer because probably only one image of the GOP is going to be played. This problem can be tackled by using scalable video media. Having a set of teams where each one of them broadcasts the lowest temporal resolution layer of a diﬀerent delayed video, fast forward and rewind can be performed when the peer visits only those teams. Finally, notice that temporal scalability does not introduce a signiﬁcant overhead in the system because its coding overhead is minimal (a video scalable in time needs approximately the same amount of memory than a non-scalable one).