... > It also depends on the application load. I can't quote any figures on it, so it'd be great to hear from someone who can, but I don't think we're talking that order of magnitude. If you route on a single interface then IMMEDIATELY a packet comes in, the router acknowledges it on one interface and forwards it to another computer on ... erm... another interface that happens to be aliased to the same phyiscal ethernet socket. Bonkl! Bonk! Packet clash! The TCP/IP stack retransmits each of the packets after a random interval, but by that time more packets have come in. Resending the packet should be no different from sending multiple packets in a row (e.g., in an FTP session). Normal Ethernet carrier sense will apply and, on a properly configured interface (e.g., no duplex mismatches), you should see no performance degradation. I wouldn't be surprised to hear that a 100Mbps connection was slowed to less than kilobits, (ie less than the speed of the DSL connection) or if it just caused hideous instability, or killer unreliability in anything trying to use the network interface. When I read about this Again, given a properly configured interface, I would be VERY surprised to see such degredation. And if the interface is not properly configured, you'll see this problem in any usage pattern. FWIW, I've been building and programming Ethernet interfaces for over 25 years. (Yes, since the days of the original thick cable.) Which is probably way too long (:-). Craig