Comment at end.....
> -----Original Message-----
> From: Eric Rosen [mailto:erosen@cisco.com]
> Sent: Tuesday, October 30, 2001 12:11 PM
> To: Shahram Davari
> Cc: ccamp@ops.ietf.org
> Subject: Re: CCAMP WG action items
>
>
<snip>
>
> Shahram> 4) It assumes reachability of all tunnel
> ingresses and
> Shahram> intermediate nodes by the tracing entity. So,
> for example, all
> Shahram> intermediate nodes processing GTTP have IP
> connectivity to the
> Shahram> tracing entity.
>
> Not exactly. If there is no IP connectivity to a particular
> tunnel ingress,
> you just can't trace through that tunnel. Many providers do
> not want others
> tracing through their tunnels. One could still show that
> things are okay up
> until the particular tunnel is reached.
Eric:
It's a bit more complicated than that. The tracing entity requires IP connectivity to all nodes that implement the tunnel at the particular level under test. The tunnel level under test needs to be able to carry IP. The tunnel ingress needs to silently discard all "ICMP exhausts" notifications of GTTP messages it receives (presumably by examining the forensic trail returned in the ICMP message to figure out what is GTTP). Similarly, intermediate nodes need to examine all TTL expired packets to check if they are GTTP (via finding a specific UDP port number) instead of just discarding them after generating ICMP messages.
So with a lot of provisos this thing can work, but IMHO it is stretching the fabric of the original design of ICMP and TTL handling a bit far....
Dave