Internet Buzz with Richard Wiggins | 4

An Interview with Vint Cerf

This raises an interesting issue for us on the design and implementation side. It's one thing to allow someone to connect up to the network and essentially subscribe to a higher quality of service and pay for that. It's something else to let someone dynamically say "Well, now, I'm sitting on a Web site, and looking at this page, and now I'm going to download a 10 gigabyte file and now I need more service." Dynamic selection of service is a much tougher proposition than a subscription-based thing. The other thing you notice is that if it's dynamic, before you grant the access to a higher-priority service, you have to make sure that whoever's making the demand is actually a paying customer who is willing to compensate you for that.

So you get into access control questions; you get into admission control policies. Can I actually support the service that I just said I would offer? Now, those are all pretty complicated engineering questions that are still the subject of research in some cases and very advanced development in others.

….That's really why the Internet2 project exists, and why the NGI effort in the Federal government exists, because we have to have places in the Internet world where we can do the experiments and try out these various solutions and not just see how they work in the laboratory, but also see how they work with several hundred thousand users. We discovered in the NSFnet period that until we do that, we really don't know how well new technologies will work. There are issues that have to do with scaling that are very important.

We've taken advantage of that in fact; the academic community has been the guinea pig for a number of advanced services, including tests of quality of service or classes of service in the vBNS network – which experience we now hope to translate into the publicly accessible system.

There's an apocryphal statement attributed to you, Dr. Cerf, as to the number of IP addresses the network would need….

Oh, yes…The maximum number of addresses you could have in the current system is about 4 billion – 32 bits, and we don't have that many assigned right now. We probably have about 50 million assigned. So we haven't run out yet, and the predictions are that we won't run out of those until perhaps 2005 or so. The problem is that we can see new applications coming, that involve Internet-enabled appliances – video cassette recorders, television sets, telephones, and the like – and if that happens, there may be a hundred different devices for each person that requires an Internet address. That causes me to want very much to move to IP Version 6, for other reasons as well, but partly to get this big address space.

Now, the apocryphal quote might have been that I took some license in suggesting that with a 128 bit address space, every electron in the universe could have a Web page if it wanted to. The fact of the matter is that 2¹²⁸ is only 10³⁸, and the number of electrons in the universe is probably more like 10⁸⁰ or 10⁸⁸, so I'm only off by 50 orders of magnitude.

[laughter] But it was a pretty good one-liner, so I decided to leave it in there... [laughter]

Continue on to Part 2 of our Interview with Vint Cerf