The Internet engineering community will be eating its own dog food tonight. For one hour, the 1,250 network experts at the IETF meeting will be able to access the Internet only through IPv6. The IETF created IPv6 in the mid-1990s, but this upgrade to the Internet's main communications protocol has not yet been widely deployed - even by the technology's biggest proponents here. Network World talked with IETF Chair Russ Housley about the group's IPv6 experiment, why the transition to IPv6 is taking so long, and whether the IETF leadership is starting to panic about IPv4 addresses running out.
PHILADELPHIA -- The Internet engineering community will be eating its own dog food tonight. For one hour, the 1,250 network experts at the Internet Engineering Task Force meeting will be able to access the Internet only through IPv6. The IETF created IPv6 in the mid-1990s, but this upgrade to the Internet's main communications protocol has not yet been widely deployed -- even by the technology's biggest proponents here. Network World National Correspondent Carolyn Duffy Marsan talked with IETF Chair Russ Housley about the group's IPv6 experiment, why the transition to IPv6 is taking so long, and whether the IETF leadership is starting to panic about IPv4 addresses running out. Here are excerpts from their conversation:
You're turning off the IPv4 network during the plenary session tonight and requiring everyone here to use IPv6. Is this significant or is this a publicity stunt?
I didn't do it as a publicity stunt. It's become more of one than I intended. It's our strong preference to preserve the end-to-end model of the Internet and deploy IPv6 and not [network address translation.] We need to have network people use the IPv6 implementations, find out where they're fragile and fix them so IPv6 deployment will be smooth. This experiment also is about finding out how hard it is to deploy IPv6. A lot of IETF people want to have their Web sites accessible during the IPv6 experiment, and they are taking the necessary steps to make their Web sites available on IPv4 and IPv6. IPv6 has been available on our network for the whole meeting, but we're turning off the IPv4 part of our network for one hour. The point of this is to get people to move out of their comfort zones with IPv4 and start using IPv6.
The IETF created IPv6 13 years ago, and it's only in a handful of production networks. Why is it taking so long?
Economics. If you are able to get your application to work on IPv4, there's no economic reason to invest in IPv6. You're just creating a second way for the same job to get done. The thing that's changing is that we're going to run out of IPv4 addresses and so we have a choice facing us. One is to have NAT upon NAT so that the Internet continues to work, or we can preserve the end-to-end model by moving away from IPv4 to IPv6. I'm hoping that IPv6 will be the answer because we want it to be easy to add new applications to the Internet. The more things that are in the middle such as NATs, the harder it will be for new applications to deploy.
Nothing. At the time MPLS and SIP were developed, there were already economic reasons to deploy them. IPv6 was developed because we knew the IPv4 address shortage was coming. But the economics for deploying IPv6 doesn't happen until the IPv4 address space becomes very scarce. All the technical pieces were there, but it's the economics that are driving when it is actually used by service providers. The IETF doesn't have economic levers.
I'll ask you the opposite question: What's good about how the IETF designed IPv6?
We saw IPv4 address depletion was coming, and we filled the need perhaps too soon. There was this idealistic view that if both IPv4 and IPv6 existed, both would be deployed. Then when IPv4 addresses ran out, there would be this smooth and easy transition to IPv6. But the economic incentives weren't there to deploy IPv6. We won't know if we designed IPv6 correctly until it's deployed and used heavily. That's what our experiment is about tonight.
The United States appears to be in a recession. Will this slow IPv6 adoption?
I doubt it because IPv4 address allocation requests have not slowed. That says that people are still deploying Internet applications. We can argue over what month IPv4 addresses are going to run out, but the prediction is that in 2010 the Internet Assigned Numbering Authority will issue the last block of IPv4 free address space. This prediction is based on the assumption that the IPv4 address space requests will continue to come in at the rate they have. A gold rush mentality will make the addresses run out quicker. That hasn't happened yet. People are still playing fair.
Is IPv6 inevitable?
No. NATs could be used to solve the problem. I just hope that they aren't. I think we will have a better Internet if we go to IPv6 rather than NAT upon NAT. IPv6 is happening in China, Japan and Korea. The U.S. Defense Department is going to IPv6. I think you'll know that IPv6 is reaching the tipping point when content providers make their stuff available on both IPv4 and IPv6. When the content providers tool up for an IPv6 population that will help with [building a critical mass.] I will be very interested to see which content providers are accessible tonight. Google has publicly accepted the challenge to support IPv6 by the time they host our IETF meeting in Minneapolis in November. We haven't heard from Yahoo or MSN. We will see tonight if their sites are accessible on IPv6.
What's the risk for a company or government agency that doesn't switch to IPv6 soon?
In the next two years, probably nothing. After that, it's going to get harder to get the IPv4 addresses they need to grow their networks and that's when the economics will kick in. Do they spend money to obtain additional IPv4 address space by buying it from others? Or do they spend money investing in IPv6? IPv6 will be a one-time cost whereas as IPv4 gets scarcer and scarcer, it's going to continue to cost more and more each time they need more addresses.
If you were giving a speech to a room full of CIOs, what advice would you give them about IPv6?
Deploy it sooner so you can have it installed and it will be robust and your people will be trained before it's a crisis. The timeframe for that is two to two-and-a-half years. There's a capital expense to get the equipment to do IPv6 versus an operational expense to buy additional IPv4 address space. You're going to pay one way or the other. The capital expense for IPv6 will move you to a place where Internet addresses are plentiful. (Compare IP address management products.)
Why is the IETF leadership so much more serious -- panicky even -- about IPv6 today than it was a few years back?
We're concerned because we're getting closer to the place where the IPv4 spigot runs dry. A very small percentage of the people here think that IPv4 trading will delay the need for IPv6. If that path comes to happen, the per-packet cost of the Internet will go up and the lack of plentiful Internet addresses will make it harder for places that don't have infrastructure such as Africa or South America to get involved in the Internet. Describing us as panicky is fair.