Tim wrote: > > Wouldn't they also have to be co-relating IPs to MAC addresses? Surely > they couldn't just work by the MAC, alone? > > For instance if my PC at 192.168.1.1 wants to do something with > 192.168.1.2, all that goes out on the wire is the IP addresses, hoping > that something else figures out how to connect the two together, or > hoping that they're already directly connected together. > >>From the manual (tiny bit of paper) that came with my simple switch, I > understood that it listened in on the traffic, worked out what IPs > belong to what MACs, and switched accordingly after a few initial > moments of discovering how the network was set up. If an IP or a MAC > changed for a device (just one, and not necessarily both), it'd need to > rethink things before it worked again. > > I can't say what switch I have, it's a black box, in a dark spot in the > shelf. I can't see anything to identify it, just the blinking LEDs on > the front. > If I understand things correctly, for the local network, your computer does the IP to MAC mapping. (Run arp to see it.) The packet has the MAC address as part of it. For destinations that need to use a gateway, it has the gateway MAC address. The other thing to consider is that not all traffic has an IP address. This is because TCP/IP is not the only network traffic possible on the LAN. Other protocols use different identification. What switches learn is what MAC address is on what port. This can cause problems if you change connections, depending on how fast the switch "learns" the new port. When you get beyond home-grade equipment, you may be able to tell the switch not to automaticly "learn" the new port. This prevents someone from "cloning" a MAC address and hijacking traffic. Mikkel -- Do not meddle in the affairs of dragons, for thou art crunchy and taste good with Ketchup!