The thing I found particularly fascinating about this one is the mining of public data, and the occasional slip, to come up with a defensible, if not plausible, estimate of Google's capacity.