Subject: Re: should the default route get a new interface automatically?
To: None <itojun@iijlab.net>
From: Robert Elz <kre@munnari.OZ.AU>
List: tech-net
Date: 07/30/2001 18:49:15
    Date:        Sat, 28 Jul 2001 08:34:17 +0900
    From:        itojun@iijlab.net
    Message-ID:  <25324.996276857@itojun.org>

  | 	do you know what routing daemons are for? ;-)

They're for routers ...

[and from a later message]
  |     - listen to them on your server (routed -q)

Are you really seriously suggesting that people run RIP?

And on non-routers what's more???

I guess I could run routed -q on some of my hosts, but the only thing
they're ever going to receive is if someone somewhere forgets the '-q'
(and is running on a host with multiple interfaces where that isn't
implied or actually sets 

Real routers run OSPF, and routed -q doesn't help find them at all.

There is an IPv4 router discovery protocol (rfc1256) that NetBSD ought to
be supporting (both sending, from routers, and listening for, from hosts).
I have a vague recollection that routed might be able to listen to
such things, but it has been so long since I even thought of running
any RIP code that even if it does, it is useless there.


apb already explained the causes of the problems with the default route
being hosed after interface reconfiguration - that really should get
fixed - either by having the routing code notified when interfaces change,
and simply refreshing itself, or by adding something like "route refresh"
which would do it all (if the code to do it doesn't want to be in the
kernel, then route(8) could fetch the entire routing table, delete it
all, then install it all again - a bit of overkill, but effective).


In any case, the answer to this should not be to advise people to start
running routed ...

kre