Thursday, June 27, 2013

The life of IPv6

A friend today sent me a link to an old blog post from a very well respected man regarding IPv6 planning and design:

http://www.networkworld.com/community/blog/ipv6-address-design

I had seen this before, but after reading it again I sent the following to my friend. While I've had this opinion for many years, and have said as much to those I know for just as long, I had never published this opinion outside of IPv6 mailing lists and IRC channels. So, here it is here, take it for what it's worth.

Thanks for sending that!

It’s hard to argue with that man on any point in networking, but I still take issue with 64-bit interface identifiers. He admits that it’s extremely wasteful and he’s right, but the way he blows it off is troubling. Pretty much all of his design points with regards to organization using bits of the address could easily be shifted 24, 32, 40 (or more) bits to the right to make longer prefix lengths and extending the life of IPv6. Of course this would kill SLAAC, but SLAAC should never have existed in the first place (IMHO). The inclusion of a node addressing algorithm into the protocol itself is a huge fail.

The logic I use to advocate against SLAAC and 64-bit interface identifiers is based in both memory of shifting to IPv4 CIDR and also in today’s difficulty in IPv6 deployment. Yes, I fully understand that IPv6 != IPv4, but that doesn’t mean throwing away sane network design principles. Further, if you extrapolate today’s difficult and protracted IPv4->IPv6 transition to the IPv6->IPv? transition that our progeny will be stuck with then you’d start to realize that a much more conservative approach to IPv6 addressing today would mean a *much* longer life for the protocol. I’m talking about millennia, not decades, and well beyond the reach of this solar system. If something new comes along in 1200 years that deprecates IPv6 due to its superiority and simplicity to implement then that’s an easy case to make for transition. If our g-g-g-g-g-grandchildren run out of IPv6 in 400 years and have to move to IPv? then they are going to rightfully curse our names.

Or, maybe I’m wrong.
So, there you go. As I said to my friend, maybe I'm wrong, but given the facts and history I am going to stand by this opinion. What do you think?