Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 28 Jun 1995 10:15:18 -0700 (PDT)
From:      "Rodney W. Grimes" <rgrimes@gndrsh.aac.dev.com>
To:        wollman@halloran-eldar.lcs.mit.edu (Garrett Wollman)
Cc:        jkh@freebsd.org, FreeBSD-current@freebsd.org
Subject:   Re: Paul Richards: sysconfig routed setting
Message-ID:  <199506281715.KAA09858@gndrsh.aac.dev.com>
In-Reply-To: <9506281556.AA00902@halloran-eldar.lcs.mit.edu> from "Garrett Wollman" at Jun 28, 95 11:56:15 am

next in thread | previous in thread | raw e-mail | index | archive | help
> 
> <<On Tue, 27 Jun 1995 22:52:17 -0700 (PDT), "Rodney W. Grimes" <rgrimes@gndrsh.aac.dev.com> said:
> 
> > ``Should'' and ``probably'' are not very convincing words, 
> 
> Let me word this more strongly:
> 
> 	A machine which is not directly involved in forwarding packets
> 	has no business running a routing process, PERIOD.  Routing
> 	processes run on routers and routers only.
> 
> > If RIP
> > packets are on the wires you should damn well be sure you listen to
> > them in one way or another.
> 
> Bull****.  If RIP packets are on the wires, then it's the business of
> the routers which are exchanging them.  If RIP packets are not on the
> wires, then it's the business of the routers which are exchanging some
> other routing protocol.  If you are not a router, you have no business
> listening to them.  HOSTS DO NOT NEED ROUTING INFORMATION.

HOSTS REQUIRE ROUTING INFORMATION!!!  A default route is ROUTING INFORMATION.

> > Give those 4 conditions the best default there is to ship the system
> > with (since you can't do the default route or gated configuration) is
> > to have routed -q started by default.
> 
> Actually, no, the best default is to ship the system with no routes at
> all, and force users if the indicate they are connected to the net to
> designate a default router.

A default router is not always the appropriate solution in a multiple
routed network environment and will fail if the default router is 
powered down and the work load takeing over by a secondary router. 
Any site who gives a hoot about network reliability has back up routers,
and runs RIP or other dynamic route listners on every single host.

> If they don't know the number to specify,
> then there should be somebody else in their shop who does.  (If their
> shop is at all sensible, it will be numbered net.sub.net.1.)  We
> actually have (broken) code in the source tree which could be used to
> listen for router advertisements and suggest a default to the user,
> but not all routers advertise.
> 
> > IMHO, and evidently CSRG's as well, routed -q should be the default
> > it covers the widest cases. 
> 
> Which is just another example of a piece of Internet technology that
> CSRG got completely wrong and we are just now recovering from.

Seems we disagree.  How many large network sites have you had to
administer, and how many times have you installed back up routers and
networks that must meet a 7 day x 24 hour up time requirement?

Have you ever been responsible for actaully deploying and makeing a
200+ node network stay alive on a day to day basis?  Do all of your
networks client class machines run with a ``default route'' or other
static route information that would fall flat on it's face if I came
in and powered down your default router and/or had to do a full
network upgrade that would replace all existing routers?

Been there, done that, default routes are fine for small static networks,
but fall to pieces in any sizeable changing structure.

You might hate routed and the RIP protocol (and I agree it is not very
eligant) but it does get the job done for the a very large mass of the
world.

-- 
Rod Grimes                                      rgrimes@gndrsh.aac.dev.com
Accurate Automation Company                 Reliable computers for FreeBSD



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199506281715.KAA09858>