On Mar 22, 19:47, Eric Smith wrote: a not I have seen such schemes. The small objects sold as "passive" 2-port or 3-port 10baseT hubs work like that, and two adjacent nodes can't talk directly to each other. It rarely matters. Well, I guess that depends on your network setup, actually! If you've not seen them, they look like a little black (usually) box with two RJ45 sockets on one end (into which you connect, say, two workstations or PCs) and a short RJ45 cable at the other end (which you connect to a real hub, switch, or a server). They're intended to let you connect more clients than you have hub ports. It's not a question of using some wacky scheme; rather a question of simplicity: if you want the adjacent nodes to talk to each other, you need to have the Tx from each connected to the Rx of the other. That's easy to arrange with just two 10baseT ports (that's all a crossed cable does, after all), but with three (as in the so-called passive hubs) you would end up with *everything* connected together -- or you need some electronics to isolate and filter (as in a proper hub/repeater). The passive units don't have that, so the two RJ45 sockets can each talk/listen to the RJ45 cable, but not to each other. There's a similar problem connecting three or more AUI ports; it's further complicated by the collision detect, which is normally done in the transceiver. Now if it were a multiport repeater, I'd have expected all the AUI connectors to be the same gender (since they'd be functionally equivalent). But the fanout units have one of a different gender, so I suspect that either they use a similar scheme to the passive hub I described, or they actually have more electronics than a normal repeater (or at least arranged rather differently), in order to get the signals in the right arrangement to drive a transceiver instead of a drop cable to another AUI. -- Pete Peter Turnbull Dept. of Computer Science University of York