2 ms·
There was a long period of chicken and egg problems: no IPv6 users, so no IPv6 services, so no IPv6 users; that's over now, there are enough users and services
by toast0 6y ago
There was a long period of chicken and egg problems: no IPv6 users, so no IPv6 services, so no IPv6 users; that's over now, there are enough users and services (even if they're mostly CDN frontends) to justify turning on IPv6 as a service or a user, although maybe not enough to require IPv6 as either.
The other reasons are all about the difficulty of the change. IPv6 is not an incremental update to IPv4. There's a ton of changes that are just totally different. Arp is replaced by ICMPv6, dhcp is mostly replaced by slaac, but dhcpv6 was added later, header processing is different. All that means, it's a good bit of work to get IPv6 to run as well as IPv4, and chicken and egg issues made it hard to justify doing the work, and hard to verify the performance.
Finally, work towards making transition easier was only started much later than the protocol design. Making it easy to use, and easy to switch to should have happened during design. Things like continuing to use ARP for IPv6 would have been simple (arp is extensible), and reduced implementation work, and gotten things moving quicker. OTOH, you would be stuck with ARP forever, instead of ICMPv6, but it doesn't seem like a big difference to me.