One important application of SCUBA is to control a simple video conference. In this case, receivers can determine the source weights required by SCUBA via the user interface. In our implementation embedded in vic, we infer the set of sources that the user is interested in from the disposition of the user interface, and weight them accordingly with some simple heuristics.
In addition to the disposition of the vic user interface, SCUBA can utilize cues from cross-media activity. For example, video windows can be configured to automatically switch to the remote user that is currently active on the audio channel. Thus, SCUBA can utilize activity on the audio channel to configure its receiver-interest report on the video channel. Similarly, pen input to a shared whiteboard tool can be used to raise priority of media streams originating from the same source.
Some multimedia conferencing tools, e.g., vic [16] and vat [13], use a common control channel (called a ``coordination bus'') to exchange cross-media control information. For example, when vat receives the first packet of a ``talk spurt'' it announces the event over the coordination bus. Any tool listening to the bus can then react accordingly. For example, vic could use the audio channel activity notification to allocate more computational resources to decoding and displaying the corresponding participant's video stream as opposed to those from other inactive participants.
While receivers can utilize cross-media cues to control their behavior, the coordination bus model does not directly extend across the network from receivers to sources. Not only do we want the receiver to allocate more computational resources to decoding the active participant's media, but we also want those media to be transmitted at higher quality. Rather than build an ad hoc protocol for this purpose, SCUBA provides all the requisite functionality. In effect, SCUBA provides a level of indirection that captures cross-media events at the receivers and reflects the desired dependencies back to the sources via receiver interest reports.
Figure 9 illustrates an example sequence of events between a source and a receiver, each of which has an audio and video component. When the audio receiver detects data from the source, it notifies the video receiver via a coordination bus message. In response, the video receiver expresses interest in that source via SCUBA control messages. In turn, the video source increases its transmission rate precisely as desired. Thus, there is no need to build a special-purpose voiced-switched video network protocol since SCUBA completely subsumes this functionality with a clean and uniform mechanism.
