Subject: Re: BIND core dumps on large DNS tables - was Re: BIND and big giant RBL PLUS!!!! table wahoo!
To: henry nelson <netb@irm.nara.kindai.ac.jp>
From: John Maier <jmaier@midamerica.net>
List: netbsd-users
Date: 06/11/2003 22:54:41
> On Wed, Jun 11, 2003 at 12:40:59AM -0600, Rick Kelly wrote:
> > >more COM domains than there are rbl domains.  It may be
> > >a matter of tuning your kernel/server or perhaps drop it
> > >into a nice openldap server and use LDAP to look it up.
> >
> > It's probably better to do MAPS RBL through BGP than through DNS.
>
> Can you recommend a good NetBSD-oriented document on how to set this up?
>

This is not a NetBSD/SMTP solution but a router solution.
See: http://mail-abuse.org/rbl/usage.html#BGP

Don't try this with a low end Cisco i.e. 25xx or 16xx, it will die a painful
death.

If you have a Bay Networks/Nortel ASN stack or BLN/BN these rock as BGP
machines, as they allocate one of the nodes of the stack as a Soloist
process, while the other nodes continue on as regular routing engines not
affected by possible BGP CPU bog.  A beefy Cisco with lots of RAM would be
good to.

Since I was exposed to RBL as DNS originally, I signed up for it and that's
what I use.

I see the advantages to a BGP4 method, as you can set your core router with
BGP and close your entire network to hosts in the list.

What I'm curious about is how much resources will it take?  I've been on the
sending end of BGP, but not receiving end..and this isn't just 13 Class C
networks using BGP3, but discrete hosts via BGP4!

jam