On Friday 31 August 2001 1:38 pm, Timothy Reaves wrote:
It's not a matter of being hostile, only being accurate.
There is no such thing as a switching hub; this stuff is very well defined. As a person stated, what you might have is two hubs in the same box with a segment to link the two.
Small differnece, but a difference none-the-less.
Absolutely right. When the ethernet specs were devised, no-one envisaged that there would be a 100 Mbps or 1000Mbps need at some point in the future; as a result, the layer 2 (often known as the 'data link' layer) structure was a bit sloppy (in hindsight). When it came to defining the Fast Ethernet standard (IEEE 802.3u), the major difference was in the data-link; much of the CSMA/CD is very similar. However, the two are incompatible, and cannot exist on one segment. What 'dual speed hubs' (the 3Com name for them) have is two backplanes; one 10 Mbps and one 100 Mbps; the ports are connected to both, and the backplanes are linked by a dumb 10/100 bridge (which will convert Ethernet to Fast Ethernet and vice versa). As with all hubs, effective port bandwidth on the segment is equal to the speed of the backplane (10 or 100 Mbps) divided by the number of ports in use on the backplane. Switches are different things altogether; they keep a log of MAC addresses and the corresponding ports they are associated with; and when they receive traffic for that port, they *only* repeat to that port. Each port on an early switch effectively acted as a bridge; it would forward stuff to the switch CPU which would decide where the traffic went. Nowadays, it's much more distributed; each switch port has some intelligence itself, so it doesn't need to bug the CPU as much; and each port is capable of autonegotiating Ethernet or Fast Ethernet. Just my £0.02. Gideon.