Tweaks to IPv4 could free up millions of addresses
theregister.com
theregister.com
Ipv6 drops so many legacy features of ipv4 and provides flow identification as part of the protocol. It's a damned shame in my opinion, as I do think it was designed "Better".
IPv4 NAT does provide some degree of privacy, but law enforcement can still identify you if the TCP source port is logged. Is this incremental privacy benefit really worth hampering the growth of the internet forever?
The goal of IP is to route packets from A to B, which fundamentally requires knowing the locations of A and B. Privacy and efficiency are contradictory goals, and it's much easier to build privacy on top of an efficient network (e.g. using proxies and VPNs) than the other way around.
Instead you will get a world that's controlled by Amazon, Google and Azure, since they are hoarding as much IPv4 space as they can get their hands on (amazon alone has over 2 billion $ worth of IPv4 space). For startups it gets more and more expensive to acquire the necessary ip space.
> providing plausible deniability as a core feature.
Not really. In the non-carrier grade NAT (CGN) case (where you just do NAT at home) you still have a 1:1 mapping to your home network and the situation is exactly as good/bad as it is with IPv6. But with CGN the ISPs now have to keep logs on all connections being made for abuse tracking.. so much better (depending on the CGN implementation of course).
> provides flow identification as part of the protocol.
That's a good thing though and not any worse than before. Routers have to load balance traffic across multiple different paths and they do that based on flows (since load balancing one flow across multiple paths can cause reordering, etc.). And since src_host and dst_host is not enough entropy for that, typically routers also look at the TCP/UDP source and destination port to identify (with funny results when you use non-tcp/udp protocols or have different hashing algorithm between different routers). With the flow label the router just has to look at the IPv6 header and not concern itself whether it's a tcp or udp packet.
Deprecate software that doesn't support IPv6.
Stop DHCP servers from providing IPv4 by default.
Then we can enjoy a world with mass 6 adoption, almost no hosts bound to 4 addresses, and billions of available 4 addresses.
We could have rolled out 6to4 with no anycast gateways, and the entire internet numbered from 2002:<ipv4>::/48. Every node could have IPv6, using only software updates and no cooperation from ISPs or RIRs.
That approach would have had some interesting positives and negatives, but the time to discuss it was >20 years ago.
It would have gained us very little and we would be in an even worse situation now (in some countries we are at > 50% IPv6 in the real world now!)