Hi list, Nice discussion. I've been AFK over the last few days and i'm catching up here.
DMDI requires a higher bandwidth than MIDI. Cat 5 has higher throughput than bell wire. The packetization of DMIDI can be very efficient. When receiving MIDI notes from an interface DMIDI has to introduce it to the network. This high header/payload ratio for one byte is as bad as it gets, yet the network loading is still way within the limits for a 10mb network. If you send complete MIDI control sequences or sysex messages within one DMIDI packet the header/payload ratio will probably out strain the MIDI hardware interface. It would be interesting to compare this limit of throughput from an 8k DMIDI packet on different cards. RTP provides far greater control in the delivery of packets. Yes, but it's a pig to implement easily. DMIDI relies on a well set-up local network to work correctly. Turn off ARP broadcasts, don't run NFS/Samba. Treat your network like audio cable and DMIDI will work fine. MWPP is a fantastic protocol for broadcasting over public networks with reliability but it's not been designed from MIDI hardware up. It's roots lie in MPEG4 and is an extension to an AV protocol. DMIDI's implementation is far easier to program, so developing applications is far easier. The risk of packet loss will be an issue with both protocols, accept it. Regardless of what recovery methods are used there will always be a risk of packet loss. I can't think of any AV protocol, that attempts packet recovery so why do you need it on a private lan? MWPP type reliability could be achieved with DMIDI but to do so would duplicate MWPP. DMIDI is a private lan protocol and MWPP is best used for public broadcasts. They are focusing on two separate set's of users. Now, the only way to move things forward is to start developing applications using both protocols. The most important thing is to get some MIDI applications out running on Linux. I've written some code to demonstrate DMIDI, someone write the same for MWPP, and lets get coding. Phil
