I've not yet read Chris Anderson's book The Long Tail but I know I have to sometime. Its thesis is something to do with how the dynamics of new industry models are dramatically effected by the myriad of niche players that can exist courtesy of the Internet. I'm a sceptic, let's say.
The Times has published an item on a piece of research that claims to dispute the theory. The interesting news is that Anderson runs his own blog and has dealt with this piece of research.
His response is that the consequences of the sales from one retailler does not disprove the thesis. The longer version of this which he posted earlier reveals that the study reveals a lognormal distribution rather than the pwerlaw distribution of the long tail. The only other place I've seen powerlaw distributions referred to was a book by Benoit Mandelbrot (a leading chaos theorist/fractal geometer) called The (Mis)behaviour of Markets: A Fractal View of Market Turbulence.
I'm not sure I fully grasped what was going on in that book, but I think it was that random processes actually behave as if thy are going on in fractional dimension space not integer dimension space. If that is correct it has huge implications for all aspects of economic theory - not least because all econometrics is based on the assumption of errors following a normal (Gaussian) distribution.
Footnote: Other references to the Times article are here (which is where I heard about it from) and here.