On 20 Jan 2010, at 10:54, Paul LeoNerd Evans wrote: [...] > A major problem with IPv4 "addressing" is that global IPv4 "addresses" > have become simply names, not addresses. They do not give any information > on where to send the traffic, simply who it is. You need a BGP router > with a full route set to know where to send it. On our border routers > at $company we're currently looking at 305,000 prefixes. Supposing an > absolutely minimal implementation of, say, 5 bytes per prefix (4 address, > pack prefix length and next hop ID in a single byte), that's still 15MB. > More likely it'll take much more space than that.. possibly more than > the, say, 64MB that smaller Cisco routing boxes come with. That's every > internet-BGP-talking box in the world, has to have that table. And it > grows all the time..
Serious question: When IPv6 grows to the same size as the existing IPv4 network, what is preventing it from also not having a similar number of prefixes? Hierarchical route aggregation is a good start, but no solution.
