I support publication. One nit / comment: Section 4.3 and 4.4 discuss MTU issues. It would also be good to note that this process could detect MTU *changes*.
i.e. someone buys a connection between A and B which has one particular MTU. Then unknown to them, some intermediary changes encapsulation methods, and MTU goes down. Most traffic will either adapt or not be affected. But it could have impact for certain corner cases. It would be better to detect the MTU change when it happens, instead of some arbitrary time later when trying to debug "Why is some traffic dropped". Alan DeKok.
