Why ?

• Shortage of IP addressess both for machines and routing.
• Need for faster routing setups especially at the NAPs.
• Use what we have learned from IPv4 "Classic IP".
• Needed big switch - chance to fix all the "little things" too.

What ?

• New protocol layer based heavily on IP version 4.
• Retains the connectionless philosophies.
• Use the same upper transport layers to make development simple (TCP/UDP).
• Clean up link layer dependant issues such as ARP.

How ?

• Adding IP version 6 protocol layers to the network stacks of machines.
• Adding IP version 6 in parallel with IPv4 into the domain name service.
• Supporting dual IPv4 and IPv6 working, including tunneling arrangements.
• Defining IP version 6 for the "Socket" API to be cross platform.
• Porting of numerous applications to be protocol independant.

Conclusions

• Too much was changed - the porting has taken longer than need be.
• Not enough thought went into address writing syntax - clashes with URLs
• Backbone routing concepts were armwaved too much - only now being resolved.
• IPv6 is perhaps a bit too clever, but it will do the job intended.