Dave Crane wrote:
No, whomever made apache2 "need net" instead of "use net" made the decision
for me (negating the admin). Now, I cannot use apache2 locally with my laptop
without changing rc.conf.
This makes me wonder: is there a simple method /for an initscript
author/ to make an appplication require *some* sort of network
interface, whether that's a loopback or something else, and another
application require some non-local network interface?
As this discussion shows, some applications (Apache) are useful with
only the loopback, while others (NFS, Lisa) are useless without some
external network.
The net strict checking setting serves no real-world use case that I can
see: it allows a system admin to decide whether all net applications on
his machine (and there's no fine-grain control) are useful without
external network access.
This is too general a statement to make a proper decision about. While
the final decision whether an application that has some (perhaps
limited) functionality with the loopback should launch if that's all
that's available should rest with the sysadmin, this must be a
per-application setting, and the first decision must come from the
maintainer: whether there is anything useful the application can do in
the first place.
As such, the distinction should exist between
need somenet
and
need extnet
. The names are stupid and should change, of course, but you get the
idea. The idea, in a way, would be to do away with all the net.* provide
special case, and make those net scripts provide one or both of these in
the old-fashioned way:
net.lo
-----
provide somenet
net.eth0
--------
provide somenet extnet
Definitely not an expert, but the current situation has always struck me
as strange. Can someone explain the rationale behind it?
Sebastian Redl
--
[email protected] mailing list
