These are histograms from a PolyMix-4 run with Squid. Size values are taken from the content-length headers, as logged in Squid's store.log. Each URL is represented once in this data -- repeated requests have been filtered out. These histograms have variable bin sizes. The bin size is proportional to the logarithm of the size (x-value).

Type	# unique objs	% unique objs	PGL specification	Actual mean
download	30778	0.39	logn(300KB, 300KB); 0.5%	312KB
other	1580604	19.8	logn(25KB, 10KB); 19.5%	25.4KB
html	1245638	15.6	exp(8.5KB); 15%	8.96KB
image	5134496	64.2	exp(4.5KB); 65%	4.61KB
all	7991516	100		10.6KB

exp() is the exponential distribution. logn() is the log-normal distribution.