Bugzilla

With the non-default-output-run graph of comment 6, if an default-rate AudioContext is then connected to a MediaStreamTrack of the same graph, a CrossGraphPort will be required if output is to be delivered (from the AudioContext) to the default device.  ([Connections are not currently supported between different rate graphs.](https://bugzilla.mozilla.org/show_bug.cgi?id=1674892))

* The CrossGraphPort could either be between a separate graph for the AudioContext and the first graph or at the output of the AudioContext running on the first graph.

* Having the CrossGraphPort on the AudioContext output would be somewhat simpler as an initial step as it would be consistent with the media element output design, though we would still want the CrossGraphPort between graphs for non-default-rate `AudioContext`s at some point.

* The benefit of having an AudioContext on the same graph as the microphone track is that its output can be integrated into our existing echo cancellation mechanism, if there is only one output device.

Adding some detection of non-silent output on the AudioContext will be useful to free up the graph to choose the output device that is actually in use.  This would also have performance / energy consumption benefits.

With the non-default-output-run graph of comment 6, if a default-rate AudioContext is then connected to a MediaStreamTrack of the same graph, a CrossGraphPort will be required if output is to be delivered (from the AudioContext) to the default device.  ([Connections are not currently supported between different rate graphs.](https://bugzilla.mozilla.org/show_bug.cgi?id=1674892))

* The CrossGraphPort could either be between a separate graph for the AudioContext and the first graph or at the output of the AudioContext running on the first graph.

* Having the CrossGraphPort on the AudioContext output would be somewhat simpler as an initial step as it would be consistent with the media element output design, though we would still want the CrossGraphPort between graphs for non-default-rate `AudioContext`s at some point.

* The benefit of having an AudioContext on the same graph as the microphone track is that its output can be integrated into our existing echo cancellation mechanism, if there is only one output device.

Adding some detection of non-silent output on the AudioContext will be useful to free up the graph to choose the output device that is actually in use.  This would also have performance / energy consumption benefits.