Growth, Form, Function, and Crashes: an article from the Santa Fe institute. It explains scale-free fairly well. The point is, scale-free networks have a few hubs. If you knock a hub out, the network is fragmented. But your chance of knocking a hub out by random is small, since there are so few of them. That makes scale-free networks reliable as well as vulnerable.
Slightly more technical details at PhysicsWeb by the creators of scale-free networks. It also says that if you design a network, it may not be scale-free. But if you let it evolve, it probably will be.