Hi Eric and Jeffery, Think more about using BGP mechanism to achieve the make before break at a child.
The child can do it without parent cooperation. When a child changes the parent for a Leaf A-D route, it first sends out the route with the new parent address encoded in RT; when old parent receives the route, it does not need to take any action on it; the node that RT indicates receives the route make itself as the patent for the child and start sending the mcast to the child; When the child receives the mcast packets from both old parent and new patent, it discard the packet from new parent until timer is out, now child send out withdraw Leaf A-D route with old patent ID encoded in RT. The old patent updates multicast state and new patent do nothing when receiving the withdraw leaf A-D route. Therefore, when a node receives a leaf A-D route, it first checks RT, if RT does not indicate this node, the node should not take any action. Only the child needs to set a timer. IMO: this is better approach than the one in the draft. Regarding suppress the non-parent node to re-distribute the leaf A-D route, if we take RR as an exception case and require RR to re-distribute leaf A-D; IMO, it MAY work too. This will significantly reduce Leaf A-D route messages in BGP network and reduces node process on the messages, which makes the mechanism scale better in many ways. Regarding relaxing the second rule in 6.2, IMO, it is better than the current rule because the solution allows a child to discard the packet if it comes from non-patent node. This improves network dampening upon dynamical membership changes. Regards, Lucy -----Original Message----- From: Eric C Rosen [mailto:ero...@juniper.net] Sent: Tuesday, September 01, 2015 3:15 PM To: Lucy yong; Jeffrey (Zhaohui) Zhang; draft-ietf-bess...@tools.ietf.org Cc: bess@ietf.org Subject: Re: [bess] comment on draft-ietf-bess-ir > However, a child can use one leaf A-D with the new parent with a new > label and later send withdraw leaf A-D to the old parent Lucy, Note that "both" of Leaf A-D routes you mention above have the same NLRI and the same next hop. Thus to BGP, these are really the same route. What you are suggesting is: - Send a route with a particular set of attributes to one neighbor; - Send the same route with a different set of attributes to another neighbor; - Then withdraw the route from the first neighbor without withdrawing it from the second. This sort of functionality is not really supported by the BGP distribution mechanisms. The MVPN mechanisms generally assume the typical BGP distribution mechanisms, where BGP chooses the bestpath for a particular NLRI, and then distributes it. Eric _______________________________________________ BESS mailing list BESS@ietf.org https://www.ietf.org/mailman/listinfo/bess
