2008 Oct 18
A bridge is like, but superior to, a repeater in that it extends the same LAN. It works digitally and only transmits packets rather than just electical signals. Thus it does not repeat noise or errors, it will only transmit complete error free packets. Each trasmitted follows the CSMA/CD rules, so collisons on one segment are isolated from all the otehr segments. One bridge can connect two or more segments. Bridges make all the segments appear to be on a single LAN but with better performance due to isolation of segments.
Bridges are called tranparent because a computer does not know how many bridges connect segments of the network. Bridges usually do more than repeat packets from one segment to another. Such bridges are called adaptive or learning bridges. The software in a learning bridge keeps an address list per interface. When a packer arrives on an interface the bridge adds its address (MAC and IP?) to the list associated with that interface.
The bridge uses these lists of addresses to know whether to retransmit a packet from one segment to another. If the address shows the destination is on the same segment then it will not retransmit the packet to any segment. If the address is a multicast or braodcast address or the address is not any list then it retransmits the packet on to all other segments.
Modern bridges are more capabale than just the above. When first powered up they can check for otehr bridges and learn the topology of the network. The distributed spanning-tree algorithm is used to decide how to optimally retransmit packets. This algorithm guarantees that a closed loop of retransmissions, especially in both directions, is not created.