All things related to multicast

[Toerless Eckert <>]

On Mon, May 12, 2003 at 10:03:58PM -0400, Marshall Eubanks wrote:
> How would you measure this ? By comparing MSDP SA source addresses with
> the MBGP tables ?

Yes, exactly. Beside the number (n% of SA available in MBGP), it
would also be nice to have an idea of the top BGP route prefixes that cover
the most non-MBGP SAs - eg: those domains that do not have MBGP and
actually are active sources.


> And what is the significance of this ?

It's related to the issue of how to do multicast RPF lookup mechanism,
distance vs. longest mask lookups and SAFI 1 vs. SAFI 3 route announcements.
Basically with BGP being used for IP multicast RPF lookup and incongruent
BGP/MBGP prefix lengths you need to interpret BGP prefixes as conditional
SAFI 3 and arbitrate purely by distance. This is extremely confusing to
users because it deviates extremely from the purely longest-mask IP unicast
routing lookup. In addition, SAFI 3 is being deprecated in IPv6 and as such
this strategy becomes even more ugly in there - and reversing the argument,
if there is a clean and easy solution for IPv6 then we should of course also
try to move towards it in IPv4.

So, basically i would like to see that we are getting to a point where there
really are no active multicast sources in the transitive core of the Internet
that are relying purely on BGP instead of actively being announced into MBGP -
at least by the edge ISP.

Cheers
        Toerless