Geoff Huston, an Australian researcher whose predictions about IPv4 depletion dates have proven uncannily accurate over the years, is still not certain that IPv6 will get deployed in time to avert an addressing crisis across the Internet.
Despite the success of last week's World IPv6 Launch Day - in which 60 access networks and 3,000 websites including Google, Facebook and Yahoo enabled support for IPv6 -- Huston says there hasn't been enough market momentum surrounding IPv6 to declare it a sure thing. While Huston concedes that the event caused a rise in IPv6 usage to about 1% in the United States, he says the protocol needs to be at a 20% usage rate to ensure that it will succeed.
SLIDESHOW: Why the Internet needs IPv6
I chatted with Huston from Canberra, where he is chief scientist of APNIC and an adjunct research fellow at the Centre for Advanced Internet Architectures. Here are excerpts from our conversation:
Has World IPv6 Launch Day changed your pessimistic view of IPv6?
No. Last year, we actually saw in the three to four months leading up to World IPv6 Day [a 24-hour trial of IPv6 services], an exponential increase in IPv6 usage. When I look now at the same set of statistics -- the number of networks that announce IPv6 on their Autonomous System Numbers, the number of entries in the IPv6 inter-domain routing table -- that sharp rise that happened 13 months ago didn't appear last month. It's almost like the folk that were going to play in IPv6, started to play a year ago, and the threshold for everybody else is too high. That leads me to the relatively pessimistic view about the business issues around IPv6. The customers are not going to fund this with incremental payments. The industry -- particularly the last-mile access industry -- is having a tough time coming up with the amount of money required. I'd love to be optimistic about IPv6; I just can't see it in the data.
Are you pessimistic even in the United States, which had some of its largest ISPs including Comcast, AT&T and Time Warner participate in World IPv6 Launch Day?
Yes. We measure about 600,000 random people a day to see who is using IPv6 to access YouTube. The U.S. on average had about 0.6% of customers selecting IPv6. It jumped to 0.95% over the last two weeks. To say that IPv6 usage rising from 0.6% to 0.95% is a dramatic jump is [not true.] If we were talking about increasing from 1% to 20%, I'd say, "Let's break out the champagne." Once we get to that kind of number, there's an assured outcome. Maybe I'm a tough one. Maybe the glass is half-empty. But 0.95% is not brilliant. There's still an amazing amount of work that has to happen. One in 100 is not critical mass.
You have predicted a shortfall of 800 million IPv4 addresses developing between 2012 and 2014. What about allocated-but-unused IPv4 addresses? Could they solve this shortfall?
I look at the sales projections of Apple and folk in the Android area. They're forecasting this year along the lines of 300 million to 400 million new mobile subscribers, and the same for next year, and the year after. Even if you were capable of taking the 1.84 billion unadvertised IPv4 addresses and reusing them, that would buy only three more years of life for IPv4. And we can't get them all because a lot of them are tightly tied up in networks. Yes, some IPv4 address space is being released into secondary markets. That market is inevitable, but it's a short-term palliative measure. It's not a cure. IPv4 can't cut it if what you want at the end is one network.
There's a growing market for allocated-but-unused IPv4 addresses in the U.S. primarily from bankruptcy sales. The IPv4 addresses are selling for anywhere from $8 to $13 per address. Is this price too high, too low, or just right in your opinion?
Inside a data center, an IPv4 address is worth a lot more than $10 each -- orders of magnitude more. Once you get into a market where the highest bidder gets the good, I think the prices that you are quoting me are [going to go up]. They are too low. If we go down this path of IPv4, and we're able to keep the IPv4 secondary market open enough, I suspect the pricing is going to be much higher than $10. That's going to be painful for this industry. Are rising prices for IPv4 addresses going to propel us into IPv6? I really hope so, but it's hard to tell right now.
How is the U.S. doing in terms of IPv6 adoption compared to the rest of the world?
Where the U.S. has a cable infrastructure, it has a little more control over CPE and in that case it is actually possible to turn on IPv6 without the users doing anything. Cable providers are better off than DSL providers in terms of upgrading to IPv6. The other thing that's different for the U.S. is the move by Verizon with 4G LTE. They're reinventing most of their mobile services. Verizon understood that if they wanted a coherent mobile network, it had to be IPv6. They wrote it into all of their specs and said that all of their suppliers had to be IPv6. I haven't seen many countries where a carrier has taken such a bold move.
Is last-mile infrastructure the biggest challenge to IPv6 deployment?
Yes. This is one of the hardest problems that our industry faces. Over the last 10 years, we have made the last mile the subject of intense competitive pressure. Revenue per consumer per month is measured in pennies instead of dollars. The issue for them is twofold: One, they have to upgrade their internal equipment, but they are sitting on equipment that they don't want to replace.
The second problem is that the one thing most last-mile access providers absolutely dread is when the customer rings up the call center. It costs a huge amount of money to run a call center, upwards of $60 to have the phone answered. The last thing you want is to have more phone calls. When you introduce a second protocol on the wire, the Internet can appear to go very slow. Sometimes Web pages take 30 to 40 seconds to load. Sometimes in IPv6, you don't get there. So you might ring up the help desk number. Carriers dread more phone calls from the customer. So that's why the last-mile access providers want to wait until their competitors deploy IPv6. They don't want to be at the front of the pack.
You've said it isn't clear which path the Internet will take: IPv4 with carrier-grade NATs or IPv6. Do you still think that's true after World IPv6 Launch Day?
Yes. It's the service providers that hold the fate of the Internet in their hands. The more we prevaricate, the more they are forced to put in carrier-grade NATs that constrain the Internet and give them the ability to exploit that equipment and regain a revenue stream from the content. At that point, life gets pretty bleak. All of a sudden, content is no longer open and uniformly accessible. Content has to pay a tax to the carrier. The incumbent content providers will pay because they have to. They're rich enough anyways. But any new content is completely locked out of the network. How do we prevent that? The carriage industry needs to understand that they are utilities like water or sewage. It's an ugly job, but somebody has to do it. Yes, it spins a dollar, but it doesn't spin a million. Getting them to accept that is quite hard. To unlock the Internet, the carriers' aspirations have to be stomped on.
What policies could be enacted to make this happen?
Policymakers don't want to kill the goose that lays the golden egg. They don't want to legislate the Internet where it creates hostility with investors. I think legislators really, really hope that those carriage folk get the message and go to IPv6 so they won't be forced to legislate. There are no clear answers here, which is why I say, in all the news about IPv6, there is still room for pessimism.
What advice would you offer a room full of U.S. CIOs about IPv4 depletion and IPv6 adoption?
IPv6 is in everybody's interest, including the enterprise. If I was an enterprise network manager, what I would do is make sure that any equipment I'm buying from here on through is protocol neutral. I would want equipment that is not just paying lip service to IPv6, but can demonstrate IPv6. I would be doing my security infrastructure and my firewalls with IPv6. I'd be doing that now. I'd be enabling my public services -- Web, email -- with IPv6. I'd be doing that now because if I wait I'm going to be in the middle of a rush, and the price is going to be 10 times higher for those who wait.
The age of the laptop is over. The enterprise infrastructure is going to be LTE devices, iPads and smartphones. Work is going to go with the employee; the employee isn't coming to work. The rational enterprise is going to see that [trend] and will be quietly building up capability in IPv6. Mobility is changing business in the same way that the fax, telephone and even the PC changed business. The next wave is just as dramatic and just as fun. IPv6 is going to be fundamental to that change.
What would it take for you to be optimistic about IPv6?
If we get to 20% usage in the U.S., this is all over. If China did it, watch out. Or if Germany did it, watch out. My metric is not number of routes, number of domain names, number of Autonomous System Numbers, not even traffic. I want to see it in users' day-to-day activities. If 20% of the users in America use IPv6, the future of IPv6 is assured.