[MPLS]Suggestion to discover and distribute multicast membership for P2MP automatically and dynamically

[Zheng Hewen <[e-mail]>]

Hi,

The signaling solution for the construction of P2MP MPLS-TE LSPs defined in draft-ietf-mpls-rsvp-te-p2mp-05.txt assumes that 
the ingress node for any P2MP LSP knows the identities of all of the receivers. No mechanism in that document is provided by 
which it can discover the identities of those receivers.

In multicast LDP (mLDP) defined in draft-ietf-mpls-ldp-p2mp-00.txt it is assumed that each receiver knows the identity of 
the source of the distribution tree, but no mechanism is provided whereby the receivers can discover that information.

In order to successfully operate P2MP MPLS, and to consider extending MPLS to support MP2MP it is necessary to examine
the requirements for exchanging and learning the identities of P2MP roots and leaves. One mechanism is desired to discover 
and distribute multicast membership information dynamically.

Currently some work is in progress. For L3VPN applications, it is one choice to carry and distribute multicast membership 
information through auto-discovery mechanism based on extended BGP defined in the 
draft-raggarwa-l3vpn-2547bis-mcast-bgp-01.txt. For P2MP MPLS-TE applications, one effort is to extend IGP-TE for such 
purpose using one analogous mechanism defined in draft-ietf-ccamp-te-node-cap-01.txt. For connection-oriented transport 
network basing on MPLS, one mechanism defined in RFC2022 may be referred. Extension to MSDP is another choice to satisfy 
those requirements directly if possible.

It is desired to clear the requirements for that mechanism. Is there any interest for this? I want to do something about it 
and hope to receive advices.


Best Regards,
Heaven