Subject: Re: NetBSD vs. RFC 1241
To: None <firstname.lastname@example.org>
From: Ignatios Souvatzis <email@example.com>
Date: 09/17/1997 11:50:12
> Matthias Scheler writes:
> > In article <199709161809.OAA18731@jekyll.piermont.com>,
> > "Perry E. Metzger" <firstname.lastname@example.org> writes:
> > > I don't think so, but it would be completely trivial to implement.
> > Trivial for your or trivial for me? :-)
> For anyone, really. A couple hours reading TCP/IP Illustrated Vol. 2
> should make the implementation dead obvious. You just add the
> necessary service routines for the packet type to the proper array,
> and all the routine really has to do is unpack the packets and stuff
> them back in again.
Perry, I think its not that simple.
The multicast tunnelling code uses IP in IP nowadays, too (source routing for
this purpose is deprecated, AFAIK), so there is already an entry for the
IP in IP in the protocol switch table (at least if you have options MROUTING).
You'd have to check for what is inside, and, if it is a multicast IP address,
forward it to the old routine, or something like that. (Currently, non-
multicast inside packets are dropped).
Code for this is (basically) in the mobile IP code I'm going to integrate soon.
I don't know if this would handle Matthias' general problem. Matthias, what's
your exact application? Building tunnels through unrelated networks?