The paradox arises from the meaning of “best.” If “best” meant, "generate the most cash for the network owner," there would be no paradox. But if we accepted this meaning of best, we'd have to be content with the tightly-controlled, relatively thin stream of bits that the telephone companies currently grant us. Communications networks have a more important job than generating return on investment—their value comes from their connectivity and from the services they enable. Therefore, the best network delivers bits in the largest volumes at the fastest speeds. In addition, the best network is the most open to new communications services; it closes off the fewest futures and elicits the most innovation.
Designing a network that is intelligently tuned (optimized) for a particular type of data or service—such as TV or financial transactions—inevitably makes that network less open. As software engineers say, “Today's optimization is tomorrow's bottleneck.” Thus, the best network is a “stupid” network that does nothing but move bits. Only then is the network truly open to any and all services that want to use it, no matter how innovative or how unexpected. In the best network, the services live at the edges of the network and use the network to transport bits; they do not rely on any special characteristics of the network itself.
Via How to Monopolize the New Network [1], “The Paradox of the Best Network [2]”
This, along with Rise of the Stupid Network [3] explains why the Internet has gained such a significant role in the world's infrastructure since it was commercialized nine years ago. The network doesn't care what traffic it carries, only that it does; it's up to the edges to add intelligence, which is easier to do than adding intelligence in the network itself.
[1] http://isen.com/archives/030304.html
[2] http://netparadox.com/netparadox.html