[6bone] 2001:478:: as /48

[Stephen Stuart]

> On Thu, 4 Sep 2003, Stephen Stuart wrote:
> > We lose:
> > 
> > - In combination with RPF checking, we would lose the ability to see a
> >   traceroute through an exchange point (assuming that the ICMP
> >   feedback was sourced using the IX-connected address).
> 
> You would be running "strict RPF" checking towards your upstreams?  
> Otherwise I fail to see how RPF checking would get broken here.

If I am not connected to the exchange point, and the exchange point
prefix is not in my FIB/RIB (depending on whose implementation of uRPF
we're talking about), then ICMP feedback sourced from IX-connected
routers will not make it into my network if *loose* uRPF is turned on;
from afar, a traceroute will show "* * *" rather than an IX address.

> > - The ability to ping the near and far sides of an exchange point
> >   boundary from a distance; this is sometimes useful for determining
> >   the character of asymmetric routing (when the RTTs for near and far
> >   side vary greatly).
> 
> Ping a loopback address of a remote router and compare to ping to your 
> local router?

That may not measure what I am trying to measure (the route to the
block containing the loopback may be different from the route to the
exchange point prefix).

I may not know the loopback of someone else's router.

I may be trying to diagnose an issue with my routes at an exchange
point to which I am not connected, so I may not have a "local" router.

Stephen