May have a lead -- it appears that large broadcast and multicast packets do not make it to the Rio. If I restrrict the packet size
to something less than the ethernet 1500 mtu, then the broadcast and multicast packets get there.

Of course, this places way more load on both ends, since there is no window in the trio protocol -- it's a strongly handshaked exchange.

What I see is that things that are sourced elsewhere or from files don't have much proglem with 1500, but the streamsicle server I am running on the same machine as a source fights the jreceviert/trio combination enough for CPU that there are significant drop outs at the receiver output. This is sad.

I am confused why the larger packets are messed up. I do not see fragment packets as I might have expected with either the broadcast or the unicast method, but the unicast "works" where
the multi/broadcast methods do not seem to deliver a packet to
the application code on the rio, causing it to eventually come back with a "more data" message.

So, I am more informed than before, but still puzzled.

-dB