Class struggle on the 'Net

By Scott Bradner
Network World, 5/11/98

I don't understand it. The IETF is working hard to define new technology to bring quality-of-service (QoS) capabilities to the Internet, and people are starting to complain.

One of the most persistent complaints about the Internet and TCP/IP networks in general is that they do not feature any useful way to provide for predictable data delivery. Current Internet technology is known as best-effort delivery.

Some wags have noted that for some ISPs this is indistinguishable from worst-effort delivery.

As you might have noted from all of the hype, there is a big push on now for convergence, which, this time around, means putting everything over IP. Or, as Vice President Gore put it the other day, the theme is "IP everywhere."

But many of these hoped for IP services would be essentially unusable on the current Internet. For example, an IP telephony session through some of the overly congested public ISP peering points, such as MAE-East, would generally be unintelligible.

Pundits have used up a lot of ink in recent years lamenting the state of the 'Net and the World Wide Wait, and saying that the Internet will be useless for real work until the 'Net gets some QoS capabilities. At the same time, these pundits have grabbed at anything that might someday provide some chance for a QoS-capable Internet.

There has been a succession of potential magic bullets, but for various reasons, they have not yet been able to provide the desired functions.

The IETF recently started a new Differentiated Services working group to explore another approach to providing QoS functions for IP - an approach that can support something even the size of the Internet www.ietf.org. This technology is no magic bullet, but it does have considerable promise.

But just as the technology is starting to gel, we are hearing from another group of people - those who seem to want to have everyone get the same bad service and are affronted that someone who is willing to spend a bit more might get better service. For example, the CNN news service last week carried a story titled "Tiered service might lock Internet into class struggle."

You only need QoS controls when there are not enough resources to go around, and QoS controls by definition allocate a scarce resource unfairly. But it is silly indeed to think that we should forgo a system under which users can pay extra to get support for IP telephony and other applications, and instead wait until there are enough resources for everybody. It is like insisting that everyone must eat at McDonald's because some people might be willing to pay extra to eat at Morton's Steak House instead.

It is sad that some commentators seem to make a living out of crying class struggle whenever they see that having money is better than being poor.

Disclaimer: Class struggle is an all too common epithet to Harvard, but the above are my observations.

Class struggle on the 'Net

Related Links