On Tuesday 07 January 2003 06:59 pm, Tommy McNeely wrote: > I am curious about why people choose to make a certain interface > internal or external... > I notice several people pick eth0 as their outside interface, and > sorta "oh yea" the rest of the inside network is on eth1. I know the > linux kernel could really care less what they are called, its mostly a > "neatness" thing I guess... Also it seems like that leaves your box > open to attack from the time it installs (if you do a NET based > install) till the time you get around to actually putting a firewall > on it. Why would this in particular leave a box exposed? I think that the main reason for 'some one way, some the other' is random chance. However, consider this scenario: You have two NICs, eth0 and eth1. The connections on one you trust (-i eth0 -j ACCEPT), the other you don't. One of them fails, or the board works loose from it's socket, or something, so that upon booting the machine you only have one interface. No matter which board fails, the remaining board would be eth0. If eth0 is your 'trusted' internal network in normal conditions, and it fails, then suddenly the untrusted network is operating under the trusted network's rules. However, the IP assignment (if static!) would remain that of the trusted network, so as long as eth0 is configured with a static IP this shouldn't present a risk. If, however, both are dynamic, (say DHCP assigned) then this would qualify as a security hole, possibly a huge one. To be fair, this is probably a very rare intersection of situations, but if eth0 is the untrusted network, then any failure would be an annoyance, not a risk. j
