rs79.vrx.palo-alto.ca.us


V6 - Epic Fail

The bombs (on Iraq - rjs) started being dropped in the middle of IETF 56.

They are still pretending that IPv6 is the future.

That's more than 50% more tenacity than Cheney, and at least as economically destructive.
- Sean Doran, March 2013.

"I think this article by Dan Bernstein is a pretty good read regarding this subject - http://cr.yp.to/djbdns/ipv6mess.html"

As usual Bernstein is right and IPV6 is ten years away from whenever you ask and will never, ever be adopted as is. Forget adoption, they haven't even finished the protocol yet.

At some point in the future there were be some devices that work a different way from V4 and V6 but can talk to them. There'll be a small one and there'll be bigger ones. They'll work with anything in the XP'ish timeframe from MS and various unicies. And then it's game over - with everytihing talking to these devices the address exhaustion issue just went away.

You won't hear about this from amy of the i* organizations (iana, ietf, icann) because they poo pooed the idea of anything "not invented here" to actually be usefull. And have rejected these ideas twice in 20 years. V6 uber alles! Let's roll!

The problem is this. Right now we have a V4 internet. Its size is "an V4 internet in size, 32 bits of address".

The i* people want us to have a bigger, a V6 internet. It's many times bigger than a V4 internet. Thing is though it's all under the control of the i* people. They're a monopoly for addresses. And they very much like it that way. It's a cash cow.

But. That's not the spirit of internet engineering. We don't regulate a scarce resource, we create new resources. Look back to the very early CATANET stuff that Postel et. al. were looking at when they designed TCP and something jumps out at you. Take the V4 internet. Let's call that "Internet 0" that is to say that two to the thirty second worth of 32 bit V4 addresses is net:0 or "The Legacy Internet" or "L0:" however you want to call it.

Now of we can only find a bit in the V4 packet header - oh wait, there's a few of unused bits...

Now, by simply using one unused bit in the pacet header and creating L1: or "internet 1" you've doubled the number of addresses. The "problem" just went away. That's why I'm not in a state of panic like Randy Bush is.

It turns out there are 2048 possible additional internets stuffed away in the V4 header and by encapsulating a V4 address into something that looks like a V6 address in length - but isn't V6 - can now shoot around the network over a V4 tunnel.

That it it goes to one V4 address, shoots about a bit then comes out the other end at some other V4 address, then goes on it's merry way to L7: or something where it's supposed to go. True end-to-end by using L0: as a core and building 2027 more internets at the edges.

All you have to do is turn off the bit that says "this is an I* brand V6 packet". Shut that nightmare off and you can now get your computer to do something usefull and you too will sleep at night.

We learned very quickly that the US Government owns, rules and regulates Internet 0 - every address and domain name is controlled by the USG. So that's fine. That war is over and now the USG - the V4 Internet utility company - simply has to "keep the lights on" and keep the V4 network working efficiently as we plan on using it a network core. The i* people as government appointed stewards of this space can help operational stability.

I know I it sounds like fantasy but I saw this work 10 years ago. I pinged a V6 looking address in Chicago that was an XP machine from my FreeBSD server in Toronto.

It worked fine. 10 years later V6 can't even do this work and frankly look to be in worse off shape and don't even appear to have any hope left. It's sad, but the net out-evolved them; routed around them as damage.

RJS February 2008