So, Cisco (CSCO) was very reluctant to spend a lot of time and energy trying to get the production of V6 going until they heard an ISP saying, "I need it, I will pay for it, I won't buy your stuff anymore if it doesn't have it."

So I finally went over to Cisco, not too long ago, at the invitation of their CTO, and we just sat down and we all talked about what the situation was, where VG6 was, what the pressure points were, and why this was an issue for me. I came away feeling as if I had officially raised the attention. They said, "Nobody's asking for this," and I said, "I am. I'm an ISP and I'm asking."

UpsideToday: What must happen for V6 to be deployed?

Cerf: It's deployed in experimental form or sort of a test form in a number of places. We run it in something called the VBNS, which is a very high performance backbone network service. We built that for the National Science Foundation in October 1994. That network caters to mostly the R&D and university communities.

So we run V6 in that network using Cisco software. A number of other vendors and ISPs are participating in something called the Six-Bone, which is sort of a spin-off of the M-Bone. The M-Bone is the multicast backbone and the Six-Bone is the IPV6 backbone. So a bunch of us have been running V6 for a while.

But it's one thing to run V6. It's something else to figure out a transition strategy that actually lets people run in mixed modes, lets V4 and V6 interwork, and lets you spread more and more V6 as time goes on.

I've asked my engineers to start working on transition strategies and interoperability tests to see how well we can actually handle the V4 to V6 exchanges. Much to my surprise, in June of last year, the Internet Architecture Board invited a group to look at IP Version 6 and to see what the scenarios might be in the future.

UT: Is one service provider going to have to step forward and say, "We're going to do this. We're going to go to V6."

Cerf: I think that there's going to have to be one of those. Everybody's standing around waiting for somebody to jump in the pool, and I think that the pressure will build.

The thing that's going to push us over the hump, I think, will be cable set-top boxes and cell phones that are Internet-enabled, you know. When Nokia starts cranking out 80 million cell phones, and then somebody else starts cranking out 80 million set-top boxes, and they all want to be IP-enabled, where are they going to get the address space?

Even if they don't run out, the guys who hand out IPV4 address space are really parsimonious about it, because the IETF community has given them a set of guidelines that allows them to be parsimonious, so we don't run out of address space too quickly.

So what's the natural reaction? Well, ICANN [Internet Corp. for Assigned Names and Numbers], the organization that oversees this whole domain name and IP address allocation process, has officially turned on IPV6 allocations, so now you can go and get a V6 address space allocation, which we've done, for example, for VBNS. So the process has begun.

There's a group called the IPV6 Forum that is trying to sound the alarm, so to speak, to say it's time to get busy preparing for IPV6. So at some point the vendors will hear that, and, I hope, will start releasing production-quality V6 code.

And I hope that Microsoft will go beyond its current experimental implementation of V6 in an NT framework and go all the way to Windows, I don't know what, 2001 or 2002 production implementation of V6. And of course many other people will have to do the same thing. I believe that it's inescapable that we have to do something.

UT: Will people wait until it's more difficult than it need be?

Cerf: It's hard to say. I think that if you wait, you just increase the long period of time over which you have to do translations back and forth.

UTe: If all the ISPs are competing for address space, won't ISPs that want to offer always-on DSL or cable modem have a business incentive to go to V6?

Cerf: That's a good point, and I hadn't thought about making that point to the ISPs in fact, but that's a good one. You know, generically it's clear that you use up more address space with all these on. But it didn't occur to me until you just mentioned it that maybe that would be a good way of persuading other ISPs to go to V6.

UT: So, hypothetically, if one of them says, "We've made a commitment to V6, so we're in a better position to support broadband customers...."

Cerf: Yeah. It's fair to point out, though, that if you've got a V6 address and you want to be able to talk to anyone on the Internet, somewhere the V6 address has to convert into V4. So the ISP may be motivated, as you say, but the customer may have to buy or the ISP may have to supply an NAT [network address translation] box to be compatible with everybody else.

In time there'll be enough V6 out there that at some point it will make sense for a backbone network to flip from running V4 to running V6, and when they do that, the same NAT boxes will be needed to help the V4 guy communicate with the V6 world. None of this is going to be easy, and getting stuff going back and forth between the V4 and V6 networks is going to be an interesting challenge.