All things related to multicast

[Dino Farinacci <>]

>> I don't get it; there's nobody to send prunes because there are no routers
>> involved except for S's DR and the RP.  S's DR sends Registers, the RP
>> sends Register-Stops.  (However, in this case I guess the RP can still
>> send SA's based upon the Register-Stop state..)

    The rule in the spec says that when (*,G)->rpf_nbr != (S,G)->rpf_nbr, you
    send (S,G,R) prunes toward the RP. It says nothing about the olist state
    of the (*,G). So we implemented it that way.

    Plus it allows new receivers to not have to get data from existing
    receivers on the shared-tree. Don't know how much of an advantage this is
    but it could avoid a storm of (Si,G,R) when a receiver joins for a group
    with lots of sources. Also, if the RP happen to go down during this, the
    recevier could get packets from both internal and external sources.

Dino