|
||||||||
On Sun, 12 Sep 2004, Chris Buechler wrote: > I've seen it more times than I can count, but it's generally on older > equipment (I've seen a ton of 10/100 hubs that are notorious for > this), and ultra cheap switches. I generally use Cisco switches, and > have never seen them have a problem with it. I let all my gear > autonegotiate and it works fine. Do 10/100 hubs even have autonegotiation? Since the typical 10/100 hub consists of an "overlay" of a 10Mb hub and a 100Mb hub with a single speed converter between them, it seems like a "natural" for "parallel detection". Especially since deciding duplex mode is a non-issue with a hub. > I'll almost always force on a hub (on servers), because I've gone back > to clients way too many times to force half duplex when the machine > keeps setting itself to full, which significantly slows down network Because it "successfully" autonegotiated full-duplex mode with a hub, or because it has a screwed-up duplex default for the non-AN case? There have certainly been some cases where overriding AN was needed as a workaround for a coding bug, but that's different from saying that it's a necessary feature with properly working code. In this day and age, who in their right mind would even bother with a hub other than for packet capture? :-) > unnecessary. In the situations that I've done it, it couldn't be set > on the switch or hub (unmanaged equipment) so changing the port the > device is plugged into would have no effect. But replacing the hub or switch might. Or inserting a hub for capture purposes. Fred Wright |