This is not a problem with two sides. Rather, the problem is one of flexibility. Donald has very patiently explained why the current system works as it does. And he is right, if the PCI code simply scanned the bus in the other direction, there would be just as many people who suddenly find the system working against their expectations. What I think would go a long way towards helping people is a system for naming devices (be they net interfaces or scsi drives) based on the bus/slot number. Once so named, that device should have that name for as long as that device is there. This would let Dennis specify Bus 0 slot 7 as eth0 and he wouldn't need to worry about detection order. It also means that if Bus 0 Slot 2 was eth2 and the card in Bus 0 Slot 1 was removed, Bus 0 Slot 2 would *still* be eth2 and the existing config files would still work. In this way, adding a SCSI hard drive wouldn't render a system un-bootable. This seems to be the way the RedHat kudzu stuff is moving, perhaps with a few more primitives it could just automagically take care of it. Even with a system like that, there will be people who find the default behavior incorrect (ever moved a net card to a different slot in Windows?), but the point is to provide the primitives needed to alter the default behavior. -Michael kujawa@cs.ucf.edu - : send the line "unsubscribe linux-net" in the body of a message to majordomo@vger.rutgers.edu
