Companies providing possible infrastructure -- from telecommunications carriers, to network equipment makers, to device makers, to operating-system creators -- are hot because of the prospects.

"What I like here is that the devices are starting to do things for you without you having to fuss with them."

Making all this happen -- or even just significant pieces of it -- will require certain technological underpinnings to be in place. That's what Vint Cerf pays attention to at MCI WorldCom (WCOM), where he is senior vice president for Internet architecture and technology.

Cerf co-created the Internet's TCP/IP in 1973 with Robert Kahn, and continues to help steer the Internet's evolution at MCI/SPRINT (FON). Adrian Mello, Galen Gruman, Cameron Crotty and Chuck Lenatti spoke with Cerf about what it would take to make the Internet-everywhere ambitions feasible.

Making wireless real

UpsideToday: There's a lot of buzz around wireless IP right now, but there's a lot of skepticism about how real it is and how useful it might be.

Cerf: It's quite real. I was at Telcom 99 in Geneva, and everywhere you went it was WAP [wireless application protocol]. And you could see an awful lot of utility in being able to get wireless access to information. I think that we're going to see a lot more of the wireless access, and it'll show up in a bunch of different ways. The third-generation cell phone stuff sounds really exciting. It's on the order of 2 megabits per second in burst data rate.

UT: There are issues of access with wireless technology -- you need a line-of-sight to the transmitter and so on.

Cerf: Oh, there's all kinds of issues associated with this radio-based stuff. But think about what it's like to be wireless. I have a wireless LAN on my machine; I have a wireless LAN at home. So I just sort of wander into the house and I'm on the Net. My machine figures out that it's in a new radio link and it resyncs and reallocates an IP address. It's all pretty much transparent.

Imagine a really smart radio machine that just automatically paid attention to its radio environment, and it picked the highest bandwidth thing it could find?

First it asks, "Is there anybody running Bluetooth?" And then it says, "Well, is there anybody running a radio LAN?" and it looks for that. And if you can't find that, then it says, "Is there any Ricochet out there?"

And if that doesn't work, then its next step probably is try to get a digital signal out of the equivalent of a cell phone band.

UT: But you're talking about three or four different transceivers running three or four different protocols and all talking to each other. There has to be one controlling authority for the radios, in a sense.

Cerf: Well, yeah, that's your PC software. It's that kind of intelligence and awareness of environment that can make Internet-enabled things pretty nice to work with. What I like here is that the devices are starting to do things for you without you having to fuss with them. This automatic configuration is really attractive.

UT: Sun Microsystems (SUNW) has been pushing that, of course -- quite hard -- and there are some other efforts like that. How solid are those efforts?

Cerf: I've seen some pretty good applications come up with the wireless access protocols. I like the one with the camera where you take a picture and the image is transported to a Web site -- all wireless, and all automatic.

Is there enough bandwidth?

UT: In the past you've been concerned about the availability of bandwidth on the Internet. Is that still a concern of yours?

Cerf: It continues to be a concern. I still have my little morning ritual in which I get up and say a prayer.

UT: Why? This isn't something that we hear a lot. In fact, very few people seem overly concerned about bandwidth. They're saying, "We've got tons of dark fiber, so what's the problem?"

Cerf: The issue is not the fiber. That's not the problem. The problem is the switches. That's the big challenge. And for those of us at MCI WorldCom, particularly the guys who are trying to keep UUNet ahead of the demand curve, this is an enormous challenge.

They have had to grow the network capacity by roughly a factor of 10 every year for the last three years, and you don't do that by just throwing more routers at it or lighting more fiber. You have to rearchitect the system to change the way traffic is managed and the way in which it flows through the Net.

UT: So why are the switches the problem?

Cerf: Let me just ask you to do a thought experiment. We recently announced a trial of a 1.6-terabit-per-second link in Texas. And if you think about what it would be like to do packet-switching at that speed, a 1,000-bit packet would take half a nanosecond or thereabouts -- a little more than that: sixth-tenths of a nanosecond, at one point 6 terabits per second.

Now, six-tenths of a nanosecond is not a very long time to decide what to do with the packet. And that's what routers do. They have to get at least the target address, they have to look up in a table where it should go, and if there are any quality-of-service or other kinds of features on that packet, the router would have to make decisions about that, too.