On 2016/11/09 11:55, Martin Pieuchot wrote:
> On 08/11/16(Tue) 17:23, Claudio Jeker wrote:
> > On Tue, Nov 08, 2016 at 03:36:22PM +0100, Martin Pieuchot wrote:
> > > [...]
> > > To add 127.0.0.1 properly it's another story as currently netstart(8)
> > > sets it.
> > 
> > I would love to kill this part out of netstart(8). 127.0.0.1 should always
> > exists.
> 
> I'm ok with that.
> 
> > > I'm not sure to understand the benefit.  What's the use case for loop(4)? 
> > 
> > 2 name spaces, so that I don't have a conflict if I use lo1 for my
> > loopback IPs and then later on create rdomain 1.
> 
> I'm afraid this would confuse newcomers.  It seems to me that this is
> just a bandage for people already using multiple conflicting lo(4) and
> rdomains.
> 
> I'd say just put your loopback IPs on lo1000 or lo42...  But maybe this
> should be discussed by people using that ;)

I'm using that, and I think many people using an IGP will be too (you
want services - e.g. ssh, snmp, ntp, bgp - to stay working even when a
physical interface is down - and at least where the IGP is OSPF you
want those addresses hanging off an IFF_LOOPBACK interface, vether
won't do). I bet the majority of people doing this use exactly lo1.
(Changes I'd need to make on all bgp routers to handle this: ospfd,
ospf6d, pf, and rename hostname.lo1).

Changing lo0 to something like loop0 would be worse as it would bite
"normal" users fairly hard ("listen on lo0", "set skip on lo", "pass on
lo0", etc.).

Using something like lo(rdomain == 0 ? 0 : 1000+rdomain) would avoid
many conflicts with existing setups, but I don't really like that much
either, and I know there are some ports that do things with lo
interfaces in regression tests which might conflict.

Automatically creating a new interface just because a new rdomain shows
up seems like a lot of magic to me, is this something that would be sane
for users to do manually? Maybe only allow a !lo interface to be set
into an rdomain if there's already an lo interface on that domain?
Much of the diff would stand, but not the automatic interface creation.

Reply via email to