Evaluate the necessity for MSG_CONFIRM

[chrysn]

While doing research for multicast I stumbled upon MSG_CONFIRM, which reads like it'd help reduce traffic. It's not fully clear how this would work, but given that we do have some application layer information, this might be possible.

Possibly, this could be set on messages that are let out of the buffer as a result of a successful response to a previous request, ie. an ACK or a response. For the server side, it's unclear yet as to whether to set it ("Kernel, you just got a message from there, so it's in your cache -- don't try to validate the MAC address for the response and just send it whence it came from", possibly eating an ARP spoofing but who's relying on those anyway) or not (because we don't have confirmation of a round-trip).