Joi Ito's Web

Joi Ito's conversation with the living web.

Mr Blog
Practical IPv6

We finally released a project we've been working on in EarthLink R&D for some time now. I was not the lead engineer on this project but it's perhaps one of the most exciting things we've done in R&D to date, if not the most exciting thing.

Basically it's a demonstration of a practical IPv6 migration strategy. There is a sandbox that allows users to obtain their own /64 IPv6 subnet of real routable addresses (Goodbye NAT -- YEAH!)

Here's how it works: Simply get an account at to get your own personal block of 18,446,744,073,709,551,616 IPv6 addresses; install the firmware onto your standard Linksys WRT54G router, and blamo, you have IPv6. With this special code installed on your Linksys router, your IPv4 works as normal; you'll still have your NAT IPv4 LAN. But in addition to that, any IPv6 capable machine on the LAN will get a real, honest to goodness, routable IPv6 address too. It couldn't be easier. This works for Mac OS X, Linux/UNIX, as well as Windows XP. You don't have to do anything special on the machines on the LAN. They just work, as they say.

So with this code installed on the router and your IPv6 accounts setup, nothing breaks. You continue to use your LAN as before, but you suddenly also get real IPv6 addresses. Easy migration. No forklift required.

This may be a bit geeky for some people, but for anyone who's been worried about how we're going to get IPv6 everywhere, this should be good news. Congratulations Earthlink R&D! I'm going to get a WRT54G router and try this out right away...

Technorati Tags:


Thuis is good news. Dual cloud enabled s/w available for public consumption. Not to mention a block of v6 addy's !!

Bob fink, Steve deeringb et al, must be really happy that V6 is moving forward and gaininf traction.

Holy SHIT -- that's freaking cool. I assume that this isn't just for Earthlink customers, nor are they promising permanent blocks of IPs, and that it's all just a testbed, right?

Joi, that sounds great. This is how it was meant to be, no?

Obviously the next step with this is going to be to encourage other ISPs to flip the switches locally to encourage them to provide their own IPv6 service.

Is there something ICANN could do to gently encourage ISPs to do the right thing as regards offering this service? Maybe they could get some sort of preferential treatment for IP address or something?

Jason, good question. The IPv6 sandbox is indeed available for anyone. You don't need to be an EarthLink customer to participate.

We're not really promising anything with this sandbox (see disclaimers). That said, we don't expect to have to take these addresses back any time soon. If anything, the main factor that could cause us to have to shut down the testbed would be if the network load or other real costs assocuted with the IPv6 testbed hits the radar of the bean counters.

That is pretty sweet. I'll be installing that...

Ok its me again.. firstly I posted my comments when on the move w/ forgive the typo err's !!

I think the issues here is full BGP support on a v6 cloud is not possible. The v6 enable is more netted towards internal, for the wire consumption. I wonder how the other ISP will support Dual could routing ??

Anyhow, if anyone is playing in the sandbox, do send me and email. I just may get back into this --- I dropped the ball in 1999 !! :)-

There is only one problem, Joi. What incentive is there to do this if I am *not* a geek full of technocrap vanity?

... but you are Don. ;-)

I think the point is not so much that everyone should run out and do this, but that they have shown that it's not as hard as people have assumed. This will allow people to start working on applications and when those applications are interesting enough people can switch.

The other reason is to make the technology easily available to application and higher-level protocol developers. For example, a lot of effort is going into writing NAT-busting P2P applications at the moment. Obviously, IPv6 has implications for that, and this gives them a chance to experiment with their applications in an IPv6 environment.

Guilty as charged, Joi, but my point was that there is no compelling incentive for users to pay premium to switch to IPv6 service although I can see tons of incentives for developers for obvious reasons. The switchover must be driven by user demand. All that aside, I was surprised about the simplicity of what Earthlink did.

Don, You are correct it will be driven by user demand, but you are overlooking the point that the lower development costs can translate into lower cost applications or support contracts to the end user. The non-tech person might even be willing to pay more for the product if it means they will not have to waste hours on hold trying to get support just to make it work. ;) In any case the point of several years of developing transition scenarios was to make things as simple as Earthlink/LinkSys have shown.

Granting that this is a sandbox, my only real concern is that it limits the end user to a single subnet. We know that multimedia bridging is not supportable even with trained staff, so there is no hope that a non-tech consumer will get that right. We also know that vendors create new media types over time, with the home already potentially having 1394, powerline, & ethernet/wireless, so I would encourage Earthlink to build their backend processes around a multi-subnet prefix per customer.

Realize that a true IPv6 migration is not this easy. This will work for WAN interfaces, but LANs have numerous applications/devices that are not IPv6 compatible. A short list of devices/software that will not work:
most SANS
most NIDS
most Printers
many databases
many firewalls

IPv6 is not ready for primetime LAN migration.

Great News, but I have ADSL via a WAG54G. Any volunteers to modify this to include IPv6? Better still, when will Cisco support v6 in its home products. Even if not native v6 from the ISP, support for v6 tunnelling to a broker would be just as useful. The sooner the better so we can get v6 deployed into the home. More Apps and devices v6 enabled are appearing every day. They need encouragement to re-vitalise the stale Internet we have today.

Nigel: Tunneling is obsolete and 6bone adresses will be unroutable after 2006 June so no point in doing that.

Ok, you all need to stop being excited about this.
Why is it worth all the trouble to go to ipv6? What is the point? I have computers running unix, running a program that is essential for what I do, and it will ALWAYS run on ipv4 as long as I am alive, you will never see ipv6's light of day. It runs on 53k external compatible modems that costed me $500, do you think I want to spend another $500 so you jerks can say, "Omg, I am running ipv6! I get more addresses!"

Face the reality, no isp has been rejected more of the ipv4 pool when they needed it. Saying that "We'll run out of space soon!" is dumb, because we will also run out of crude oil soon for everyone, and yet you take ipv4 space as a bigger issue.

Is there an update for these instructions? I have a WRT54GS and I believe it has more recent firmware than the update you have posted.

Greg Mace