Re: I've said it before and I'll say it again
Yeah, that's not really an insoluble limit, and we are already living in the world where IPv4 clients aren't directly routing to each other. There is so much address translation happening that the reality of modern IPv4 delivery is that the 90s style routing you describe is only really happening on endpoints. Even cheap layer 3 switches are in reality routing traffic to a target address that knows how to reach the destination, not necessarily the destination itself. The amount of address manipulation that happens on the cloud side of the modern internet is CRAZY. Try to operate services at cloud scale without load balancers, CDNs, etc.
This isn't really crazy, as routing through the internet works in a pretty similar way already, and routing between LANs was already addressing these issues.
You are correct in one sense, which is that with the IPv4 address space you can't build a directly addressed flat network over a certain size. We don't need or want to do that though, and IP networks never really worked that way. At no point in the public internet area did every ethernet/ip device run on the same public address space. We neither need it to or want it to. The fact that you theoretically could in IPv6 doesn't make it a problem for IPv4.
The reason IPv6 is stalling is that the workarounds to make IPv4 handle this stuff are holding the line, and IPv6 still can't handle stuff enterprise connection traffic failovers at speeds that meet users expectations.
Biting the hand that feeds IT
