Improving coverage of popularity contest data?

[Dan Kegel]

I like the data at
http://popcon.ubuntu.com/
but since it's opt-in, it's only useful for relative comparisons,
and the data might be skewed towards power users.

Are there ways we could increase the coverage of the data
such that we get many more users, especially those who would
not normally bother to install or configure funny packages like popcon?

Two ideas:

1) somehow encourage normal users to turn on popcon, e.g.
by a checkbox during installation of the distribution

2) gather similar statistics from the access logs of the ubuntu repository

What do folks think?