pub const MAX_DEPTH_FOR_VOICE: usize = 64;Expand description
The depth past which tunnelled voice is refused.
Voice is bounded in latency, not in volume: at a speaker’s usual 10 ms of
framing, 64 queued messages is roughly 640 ms of backlog, and a packet that
late is worth nothing to anyone. Control messages have no such bound because
a skipped UserState leaves the client on a view that silently diverges.
One queue rather than two, because the client discards audio whose sender
session it does not know: the UserState introducing a speaker has to reach
the client before that speaker’s tunnelled audio, and two independent queues
cannot promise that.
REF: references/mumble/src/mumble/ServerHandler.cpp : handleVoicePacket
looks the sender up with ClientUser::get(senderSession) and drops the
packet when it is absent.