> > I agree. The idea of combining messages into a packet might > > be interesting > > when the source of the MIDI stream is a sequencer application instead > > of a human playing on a MIDI keyboard. Here MIDI events can be known > > ahead of time and could be sent in in bursts and the MIDI > > stream is often > > less sparse when coming from a sequencer. > > You can't still do anything with the MIDI bytes before you get the last one!
With MWPP there are always complete MIDI messages in a packet, not just MIDI stream bytes. It doesn't support an arbitrary MIDI stream. > So it makes sense to put MIDI event(s) into a packet, but it shouldn't be > restricted to one MIDI event / packet as it's not efficent. Put there either > one (human player, or if he /she is so superfast then put as many as he/she > can play) or as many as can fit (computer player) as use some reasonable and > standard way for timestamping. In this case it could be the LTB (Linear > Time Base MIDI Time Stamping method, see > > http://www.steinberg.net/infocenter/discoveries/lineartimebase/index.phtml?s > id=06023461 A delta-time (with respect to the RTP packet timestamp) can be used for multiple events in the MIDI payload, as was already said. The link to steinberg's LTB experiment doesn't say anything about how the protocol works, but even if it did, LTB is a proprietary protocol for communication between a computer and a MIDI interface over a serial connection (USB), so it hardly has anything to do with MWPP. The way LTB works as I read it in the german KEYS magazine is that the MIDI interface has it's own clock and the computer sends timestamped messages to the interface which the interface uses to adjust its clock. The interface schedules MIDI events it received from the computer using this clock. Anyway, I have an Emagic AMT8 MIDI interface :) --martijn
