Thursday, September 18, 2003

 

Today's Optimization, Tomorrow's Bottleneck

One of Mile O'Dell's aphorisms that has stood the test of time is, "Today's optimization is tomorrow's bottleneck."

This BBC article, Goodbye to a flat rate for broadband?, says:
"'The last four or five years has been about building this infrastructure of a high-speed network, providing a dumb access,' says Milind Gadekar, vice-president of P-Cube, whose monitoring system is being tested by service providers in Europe. Now, he says, service providers need to make their networks 'intelligent' so they can identify users and the applications used."
Just what I want -- a network that identifies the application I'm using!

I can imagine
*** ERROR 4XX: UNAPPROVED APPLICATION ***
*** PERMISSION TO USE NETWORK DENIED ***

Nicholas Negroponte has been talking about how some bits are more valuable than others for years. For example, we'd pay a lot for the "Is it malignant?" bit or the "Got the promotion?" bit. Is it far-fetched to imagine that an application aware network operator would charge a premium for such bits?

It is certainly plausible that an application aware network operator might charge differently for, say, telephony than for, say, streaming audio. This is fine as long as we don't mind that some third party is encouraging certain applications and discouraging others (with motivations that might not be transparent). It is also fine if we already know what all the applications are on our network, and if the encouragement/discouragement process doesn't make it harder to add new ones. In other words, NOT fine.

Andrew Odlyzko has a well-thought-out paper that points out that sellers often have strong motives to do "price discrimination," that is, to charge different prices to different users at different times (e.g., airline yield management, or even end-of-season sales). He observes that economists think price discrimination improves economic efficiency, but that customers often hate it.

We need to accept that price discrimination happens, and focus on where it happens, for what and to whom. According to the End-to-End Principle, if you have a choice to put a function at the edge of the network or in the middle, you should put it at the edge. Price discrimination in the middle of the network is a risk to new app discovery and to free speech. We should keep the network stupid -- and put the "for what" and "to whom" of price discrimination at the edge.

Comments: Post a Comment

This page is powered by Blogger. Isn't yours?