> > My opinion is that /etc/init.d/reboot should honor NETDOWN, and > > source it from /etc/default/halt. And then NETDOWN should be > > documented in /etc/default/halt, as requested in #703844. > > Actually, #696910 suggests that NETDOWN could also be disabled if > iSCSI is detected.
Sorry for changing my mind again, but I found yet another bug about this, #632091, which points out that /etc/init.d/networking already uses extensive code to detect iSCSI and other network mounts. So maybe we should reuse that code in halt and reboot scripts to disable -i; and also integrate and support NETDOWN at reboot, because apparently people have been using it to handle this kind of setup, in addition to its original purpose about wake-on-lan. Allowing the admin to configure /etc/default/halt is good, detecting failure cases so it works out of the box is better. And if we don't want to expand initscripts knowledge about every problematic network filesystem, considering that other packages can't edit /etc/default/halt, it might be even better to let them drop files somewhere, maybe in /run/network, to signal this. -- Pierre Ynard
