> Valdis.Kletnieks@xxxxxx wrote: > I'm more than happy to accept any realistic projections > that point to a change in the burn rate - if you know of > something I've overlooked, please enlighten us.... The "savings" due to NAT might be underestimated, which in turn pushes the v4 exhaustion even further than anticipated. Rationale: NAT is here and here to stay. Nevertheless, allocation policies still consider the "one host - one IP" rule valid. In reality, this turns into the following situation that I have seen at many customers lately: The organization has 800 hosts, all behind NAT (they have PA space, NAT is there for renumbering ease), and there is only a small fraction of servers that have one-to-one NAT and therefore require a public IP per host. In your average 800 hosts network (if such a thing exists) it turns out that a /26 would have been enough. Where's the catch? This organization gets a frac-DS3, and the network administrator thinks "what the hell, I have 800 hosts and therefore I'll request 2 class Cs anyway even if I don't need them, it does not cost more, and who knows if I won't need them later". The organization does get 512 addresses out of which it really uses 50, but the network administrator likes seating on a cushion especially if it costs nothing. IPv4 addresses _are_ a commodity; on cheap markets (home/soho) more addresses means more money. On more expensive links (above T1) it still means money but that money is washed out in the bottom line. When home DS-3s are available for $79/mo, expect to pay more money if you need more than a handful of addresses. Pre-CIDR blocks sell on eBay, this is a gray market that I would not recommend going into but it does happen anyway. When IPv4 addresses become scarce, we will find out that lots of people that have stockpiled them would be ready to let half of their block go if there is a sound financial reason to do so. IPv4 address will never run out. They will simply be available to whoever has money to pay for them. Michel.