Re: [lwip-devel] Bridge Implementation

[Sergio R. Caprile]

Just being curious.
I always thought that the IP virtual port, when responding to an IP on a
frame from the switch, would initiate an ARP discovery, get the MAC,
pass the response to the switch and the switch would have got the
destination port also from the ARP response. Even better, it should know
the port beforehand because it already got the question frame for the IP
port first... and in the event it didn't, just forward to all ports.
So, I always thought there was no need for the management port to know
the source port, the answer would "find its way back" anyway. I know it
is probably not the most performant way but keeps OSI happy.
What is that I'm missing (besides that bypassing the switch would be
faster and easier on the CPU) ? Perhaps it is related to the comment you
made on non-unicast messages, could you point me to a doc to feed my
hunger ? Thanks!