David Conrad wrote: > Keith, > > On Sep 20, 2007, at 6:19 AM, Keith Moore wrote: >> The point is only that sooner >> or later there will be pushback associated with routing pain, and when >> that pushback happens people will look to solve their problems in other >> ways. > > Yes. The solution chosen in the past has been to deploy route > filters. I'm not sure why this wouldn't be the solution in the > future. Of course, the downside of route filters is that increasing > portions of the Internet will be unreachable. So it goes. Either that, or there's some fallback route to a shorter prefix which leads to a (replicated and possibly federated) service that knows how to tunnel the traffic to the destination. In other words, you can minimize routing complexity and routing table size by making routes less optimal and eating more bandwidth for the actual traffic. Of course that forwarding service will cost money, and be subject to failures of various kinds. And it's hard to build such a service that is federated according to the preferences of those holding PI or ULA prefixes. They can change ISPs, but their ability to have a stable address prefix that is generally reachable is still going to be dependent on some specific outside party that they can't change. This is true no matter what scheme is used for maintaining the mapping table and what scheme is used for mapping between the stable prefix and the current destination - whether it be BGP extensions, DNS or whatever. _______________________________________________ Ietf@xxxxxxxx https://www1.ietf.org/mailman/listinfo/ietf