> Or, as RFC2003 puts it:
I'd not believe rfc2003 very much. Seems, this document is pretty raw.
F.e. ttl handling proposed there contradicts to robusteness principle:
it proposes to allow to set it to preconfigured value.
Well, I did it... It was the only case in my practice when I managed
to kill Cisco :-). When tunnels loop it look like Big Bang.
> Has anyone done any work on ICMP handling for the tunnel driver?
It is really necessary, especially path mtu discovery.
I started it but found this business very boring :-) and deferred it
to the worse time :-)
Alexey Kuznetsov