At 09:56 AM 20/07/00 -0400, you wrote: >None of this blabber of yours solves the problem, so why waste everyone's >time? Explaining why its broken is kind of like your mechanic telling you >why your brakes sqeak. I dont care why they sqeak, i want them to stop >sqeaking. I wasn't explaining why it was broken, I was explaing why relying on a specific interface name is broken. As Donald said, they're just arbitary labels. They have to be given out in some order - and whatever that order is, is never going to suit everyone. The best way around it is not to rely on interface names directly in any scripting. We do this on the single floppy router I work with - all references to interface name are made via an environment variable, for example $INTNAME0, which is stored in a configuration script. This completely removes dependency of the interface name and alows people to swap the function of interfaces around at will. Granted, it is easier on a small closely controlled system than a full linux distribution... >What code needs to be changed to fix it? Im sure Im not the only one who'd >like to know. Several other people have posted code which can reverse the detection order of PCI devices - but while that might "fix" things in one particular case, it doesnt actually solve the problem at all. Another suggestion was to read the hardware address of each card and swap their addresses based on that, which is an intriguing idea, but again doesnt solve every problem. Regards, Simon - : send the line "unsubscribe linux-net" in the body of a message to majordomo@vger.rutgers.edu
