Metcalfe's "law" doesn't work as it is just an approximation for the number of connections when N is large. I recall reading that a better fit for large N is kN.logN, where k is a network-specific constant, thus slightly different for, say, FB and Reddit etc. The reason being that users tend to cluster into user-groups.
Thanks, will follow the links.
I did point out that it wasn't accurate.. I take it you were just driving that point?
The best law is the one that fits the actual growth after the growth has been accomplished. (i.e. hind sight is 20/20).
I wasn't replying to you, but to @geekgirl - sometimes can be confusing if replies are all indented yet at the same level.
However, just as Beckstrom has attempted, it is not enough to criticise the (ab)use of Metcalfe's idea, but to discover a more accurate metric.