Ranji and Traian,
I have a slightly different take on the term "transparent bridge". Traian suggested that a transparent bridge did not perform any learning function. However, that would effectively degrade it to a hub, and there was no point in inventing a device that was constructed differently than a hub, yet behaved similarly to it.
In my understanding, a transparent bridge was essentially a predecessor of a switch that behaved identically to a switch - it performed MAC address learning and switched the incoming frame to another interface if and only if the recipient of the frame was connected to the other port or if it was unknown. As opposed to switches, transparent bridges were software-based devices and did not perform the switching using specialized circuitry independent of CPU like switches do. Usually, transparent bridges were simple computers with multiple network cards, running specialized software that performed bridging functions.
Switches are transparent bridges that perform the bridging operations in hardware, at rates much faster than software-based bridges, and with a significantly higher port density.
There were also translational bridges that interconnected two dissimilar data link layer technologies, such as Ethernet and Token Ring, while still creating a notion of a single broadcast domain. Translational bridges rewrote, or translated (hence the name), the frame format when data moved through the translational switch between these data link layer technologies. Autonomous wireless access points can also be considered translational bridges because the frame format used on the wireless interface is different from the wired Ethernet frame format, and access points translate between these two frame formats as data moves from the wireless interface to the wired interface, and vice versa.
Best regards,
Peter
