I apologize if this is off-topic. But I thought some of you may be interested in this question that I posed to the MPLS WG in IETF. Reply privately, I will post a summary to the list if I get any interesting replies: ---INCLUDED MESSAGE------- Hi all, I would like to get a poll on how targeted LDP sessions (and hierarchical tunnels) and MPLS data path tie together. Consider the following scenario: * denotes an MPLS tunnel between R_B and R_E R_A --> R_B *************>R_E-->R_F \ / \ / R_C --> R_D Here R_B and R_E have a targeted LDP session and an MPLS LSP between them. There is a label request traveling from R_A for an FEC_1. At R_A, MPLS does a routing table lookup and finds that next hop for this route is R_B, message is then passed on to R_B. Now at R_B, how does R_B know to associate the FEC_1 with R_E through the directed LDP session vs associating with R_C?? One way of doing this is to introduce a virtual interface between R_B and R_E with the lowest metric possible, so the routing lookup result is the virtual interface. The other way of doing this is to exchange all prefixes that the tunnel between is authorized to carry so that the LDP instance at R_B knows that FEC_1 belongs to this set. **What do you think is the best solution and are there any other solutions out there? **Why doesn't the LDP spec address this issue even though it is part of the framework document. ** Is targeted LDP important to ISPs, when you can do this much more cleanly with ER-LSPs in both CR-LDP and RSVP? Bora Akyol, Pluris http://www.pluris.com