At 8:23 3/21/95, yakov@watson.ibm.com wrote:

There are certain *fundamental* issues that one may choose to ignore. However, this still doesn't change the fundamental property of these issues. Relation between address space allocation/management and the ability of the Internet routing system to scale is *one of these issues*.

There are also fundamental requirements. Not all of these requirements are strictly technical (like the requirement that routing scale "nicely"). I believe that one of the non-technical requirements, but a requirement nonetheless, is for end-user organizations to be able to change providers without an exorbitant penalty. With today's technology, renumbering (of a reasonably large organization) amounts to such a penalty. I didn't say that meeting this requirement would be easy. But that doesn't mean that it can simply be ignored because it is too hard. I point out that in the U.S. the Federal Communications Commission (FCC) has mandated that 800 numbers be portable. I am sure that created a real routing headache within the telephone network. I can make all the same scaling arguments that you use about Internet routing also apply to the phone network's 800 numbers. Doesn't change the fact of the FCC action. My concern is this. If we march down the aggregation path too far we will have ISPs (both large and small) that require customers to get addresses from them where they (the ISP) maintain ownership of the address space. At some point some (perhaps large) ISP's service will degrade and customers seeking to move to another provider will suddenly realize how locked in they are. If enough customers are put into such a situation many will call for regulation of the network (if it isn't already in place by then) and demand IP address portability. We can then wind up with such portability being mandated from on high, with little thought. I would prefer to recognize this problem now and begin to see how we can address it. One solution is for IPv6 to fulfill its promise of providing autoconfiguration (complete with secure, dynamic, DNS updating). One of my fears is that autoconfiguration will not be a priority (because most users won't need its capabilities, at least immediately) so will not be fully specified and implemented until it is too late. -Jeff