> [...] each router connected to said [Common] link:
> MUST forward traffic destined to said prefix to the respective link.
Here's the mechanism in the shncpd implementation, Steven will hopefully
tell me if that's what's intended:
- for each locally delegated prefix P, we install:
- a source-specific default unreachable route (::/0, P);
- an unreachable route to P;
- for each prefix Q assigned (and applied) to an interface E, we install:
- a route to Q through E.
All of those routes are announced over Babel, even the unreachable routes.
Babel propagates them in their usual manner. If at a given point in the
Homenet the Babel metric for Q is finite, longest-prefix routing will call
it to be routed to the nearest (according to Babel's metric) router
attached to link E. If the link quality is so bad that the metric for
Q is infinite, packets destined for Q will follow the covering route to
P and reach the router delegating P, where they will be dropped by the
unreachable route. If both routes have infinite metric, then the network
is so bad that I really don't care.
Note by the way that the non-Homenet routes are *not* announced over
Babel, unlike in the hnetd implementation, which announces all routes.
-- Juliusz
_______________________________________________
homenet mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/homenet
