let me see
If I read this correctly, they have applied the filter to a few 1000 users, however the internet users number in the millions, as do the illicit websites, thus extrapolating the results from this trial is like seeing a hole in a wall, and guessing how did it get there.
I've been involved in too many trials, where you do the 10, 50, 100, 200, 500 tests, and everything appears linear, and people then assume you can extrapolate to 10 000, 100 000 because it was linear until 500, there it must be the same from there on in.
Unfortunately, in the limit of small numbers, an exponential curve may indeed be linear, for the sample, however once you exceed small limited numbers, the load and delays often go exponential.
So say you do a small test with 1000 sites, and 1000 users, nxm is only 1000 000 matrix.. However extrapolate to 1000 000 000 sites x 1000 000 users, and hmm, 1 E+15 in the data set now.
This is going to be interesting..
Though I still cannot fathom that someone can even consider the idea that they know better what OTHER adults should be allowed to see, than the adults them selves ?