[30067] in North American Network Operators' Group
Re: RFC 1918
daemon@ATHENA.MIT.EDU (Danny McPherson)
Fri Jul 14 15:14:04 2000
Message-Id: <200007141912.NAA28750@tcb.net>
To: nanog@merit.edu
From: Danny McPherson <danny@tcb.net>
Reply-To: danny@tcb.net
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Date: Fri, 14 Jul 2000 13:12:19 -0600
Errors-To: owner-nanog-outgoing@merit.edu
> We had a similar discussion a long while ago (2 years?) on whether having
> RFC1918 addressed router interface could break Path MTU discovery.
>
> The general upshot is that the RFC specifically says that no packets with a
> reserved address in the header (source or destination) should leave the
> network in question. Also, the RFC says it is not at all unreasonable (but
> not required) for a network to filter packets with RFC1918 addresses in the
> source. (To prevent attacks and things like that.)
>
> So it is nearly impossible to stay 100% compliant and address router
> interfaces with RFC1918 addresses. (Unless you NAT or something.)
Of course, if you use RFC1918 space for internal addressing,
then filter all RFC1918 SA both ingress _and egress to your
network, you'd in theory be 100% compliant (whatever that
means). You'd just be handicapping traceroute, PMTU and
the like .. but of course, if folks have a problem with it :-)
-danny
