[L2tpext] comments about draft-mkonstan-l2tpext-keyed-ipv6-tunnel

["Xialiang (Frank)" <[e-mail]>]

Hi authors,
I have reviewed this draft, and consider it a useful draft for giving a concrete use case of using mature L2TPv3 protocol as a data plane overlay technology. I also have several comments below:

1.       You use a source+des IP pair to identify a L2TP tunnel without tenant id carried on the wire. It increases the complexity when identifying tenant network's traffic using the mapping between tenant network and a large number of source+des IP pairs, especially when tenant network are highly distributed in the real network. While, using tenant id carried on the wire can reduce the complexity largely;

2.       Would you consider more general use cases, e.g., supporting IPv4 underlay network, and different types of services overlaid on it (e.g., Ethernet, PPP, IPv4, IPv6 ,etc)?

3.       I cannot find contents in the draft describing how to implement the multipoint VPN in detail. If you don't use control plane, then data plane flooding or manage plane provision should be supported. Do you follow this way?

B.R.
Frank