Internet Buzz with Richard Wiggins | 3

An Interview with Vint Cerf

Will there be different classes of packets on the Internet?

Certainly in the network that we're building we're quite serious about creating multiple classes of service in the network. That's being primarily motivated by our interest in doing Internet telephony over the system. You can well imagine that delay is the enemy of good interactive telephonic communication. In order to achieve low delay, we have to be able to mark packets as priority, and they have to be serviced ahead of others that might just be carrying e-mail or something else that doesn't require the same interaction.

In addition to that, there are one-way services like radio and video, where you are essentially multicasting or broadcasting, and for that we need additional facilities in the Net to replicate packets where needed in order to deliver them to multiple targets, and again they need to be delivered with reasonably low delay. The most demanding application, however, isn't telephony. The most demanding one turns out to be the securities trading environment, where everyone has to hear the same information within a very small amount of time to avoid the possibility of arbitrage. You can appreciate the level of stress here …

When you have multiple classes service on one network, how do you choose which packets get the priority? There's a couple of economists – Hal Varian and Jeff Mackie-Mason -- who said back in '93 that the only way to choose is dollars.

In fact, you will find that when we get classes of service implemented in the network, those classes of service that consume more resources than the others will cost more. It's the only sensible thing to do.

Yes, I agree with that. This is really about cost recovery. But that actually has two implications: it means that applications that don't require a lot of bandwidth get cheaper, and the ones that require more get more expensive….

Let's be careful here, because it may be that bandwidth is not the only metric. Performance is probably the more general way of saying it….

I agree…Going back to your example: e-mail ought to cost less than a phone call – not because e-mail necessarily uses less bandwidth in the network … a big e-mail can use a lot – but because it's not as urgent, and it can be sort of...

…stuck in there!

Right. My prediction actually is that you're going to see a sort of environment in which things like e-mail are as free as they are today; you'll pay for a connection but that sort of service will disappear into the woodwork, but when you want to do a videoconference from Paris with your wife just to see how the family is doing, that will be a little more expensive than sending an e-mail.

More...