All things related to multicast

[Toerless Eckert <>]

>       But unfortunately, as with all of these <foo>-limit solutions,
>       it will lead to black holing of some legitimate traffic
>       if that traffic happens to be bound to an unlucky (S,G)
>       during one of these events...we need more protocol
>       mechanism to deal with this, IMHO.

Well, the idea of a per-peer SA-limit is of course that you can set a fairly
low SA-limit on all edge-peerings and a quite large ones in the core. Once you
have this set-up, the MSDP infrastructure is well protected against individual
MSDP burst originating in enterprise networks. Of course, the asumption is 
that
we can get a fairly good deployment of SA-limit on the edge, because as soon 
as
the MSDP-storm is on an inter-ISP link, it's too late, and you can not
avoid loosing third-party SAs, unless you go into much more fine grained
limiting.

I think the current approach has quite a good effort vs. result relationship.
Anything further down the road is much more likely to be more difficult to
configure right and may not prove to be much better - depending on the attack
scenario.

Cheers
        Toerless