>Your message is generally well put. However, while it is possible to send >the packets on the wire, the fundamental underlying scaling point, the >routing system, has not been properly addressed. Perhaps a solution can be >retrofitted in, but then again, who knows? This, I thought, was <largely> >the point of multi6.
Scaling of the routing system is not the objective of multi6 - multi6 is working on the principle of attempting to do no harm to the scaling properties of the routing system while allowing complex edge connectivity scenarios. Generally, scaling of the routing system requires the elimination of information from the routing domain. And, generally, the best you can do here is attempt to align deployment with some structural property of the syntax of the protocol elements you are using as routing objects, and eliminate information through aggregation of the routing objects (aggregateable address hierarchies). The alternative is to re-evaluate the routing domain and select some other attribute of the domain that does not have the same population / complexity / growth and attempt to route based on this attribute. I'm unsure what IPv6 has to do with this - it seems to make the task of scaling routing about the same as it is in IPv4 - there appears to be no intrinsic property of the protocol that alters the situation to any appreciable extent. Geoff
