On Fri, Mar 20, 2009 at 08:52:15AM +0100, Andreas Henriksson wrote:
[SNEEP]
Okay, I do see how it's all generated from ifupdown.nw.
My thoughts are
1) I'd like to see a migration path off of vlan/vconfig sooner rather
than later. The ip-vlan script provides that now. Because
ifupdown-0.6.8+nmu1 already needs 'ip' from the 'iproute' package,
it probably belongs inside of ifupdown.
2) I like the idea of moving as much as possible away from the scripts.
But I think doing it right has the potential for being more disruptive.
2a) To avoid a lot of code repetition, there should be a bogus
address family such as "link" or "dynlink" and then have the method
specify the link type.
IE: vlan, bridge, vde2, tap....
This would also require adding some parsing of $IFACE and setting
variables from it.
2b) To avoid a lot of config repetition, there should be a bogus address
address family such as "setlink" or "linkset" that handles things such
as hwaddress and mtu.
2c) Instead of just magically turning off the old options, it might be
a good idea to put in a weaning period where ifupdown complains loudly
when mtu or hwaddress are outside of a linkset address family.
This should already be such a warning in 0.7 alpha3 if 'media' is
specified.
One last thing, there are 41 packages for unstable on i386 that
make use of /etc/network/if-* scripts.
Let's start with the #1 installed and used if-* file,
/etc/network/if-up.d/mountnfs from initscripts.
This is *NOT* a lightweight script.
Tweaking ifupdown such that mountnfs's 'exit_unless_last_interface' can
be reduced to
if [ "$IFACE" != "$LASTIFACE" ]; then
exit 0
fi
instead of it's current O(N^3) nastiness would improve the experience
for users with lots of NICs a lot more than moving vlan handling from a
script to being within ifupdown.
--
To UNSUBSCRIBE, email to [email protected]
with a subject of "unsubscribe". Trouble? Contact [email protected]
