§ 4.3 Permissions Integration says: "If the descriptor does not have a deviceId, its semantic is that it queries for access to all devices of that class."
"All" here implies access to all currently connected devices AND all devices added in the future, i.e. category access.
But it's not possible to get blanket access to audio output devices as a category in this spec, because (non-microphone-associated) deviceIds are only exposed in enumerateDevices individually as a result of JS calling selectAudioOutput and users picking each device in the UA prompt.
§ 4.3 Permissions Integration says: "If the descriptor does not have a deviceId, its semantic is that it queries for access to all devices of that class."
"All" here implies access to all currently connected devices AND all devices added in the future, i.e. category access.
But it's not possible to get blanket access to audio output devices as a category in this spec, because (non-microphone-associated)
deviceIds are only exposed inenumerateDevicesindividually as a result of JS callingselectAudioOutputand users picking each device in the UA prompt.It probably makes more sense to align with https://github.com/w3c/mediacapture-main/pull/932 here.