Australian ISP level filter gets technical green light

A colleague sent me this link today to an article claiming that “ISPs give clean feed filter a technical green-light“.  After I stopped screaming and chewing at the carpet I thought I’d have a closer look before the mainstream tech media beats it to death:

More than half of the Internet service providers (ISPs) taking part in the Federal Government’s ISP filtering trial have reported minimal speed disruptions or technology problems.

Of the nine participating ISPs, iPrimus, Netforce, Webshield, Nelson Bay Online and OMNIconnect told ARN they had seen no slowdowns in Internet speeds or problems with the filtering solutions in place.

Of the remaining four ISPs, Tech2U and Highway1 were unable to respond by time of publication while Unwired and Optus refused to comment.

iPrimus Australia CEO, Ravi Bhatia, said his company’s ISP filtering trial, which must be opted into by its customers, had “probably involved a few thousand users”.

“The users have not experienced any problems, they haven’t experienced any service degradation so it’s been a pretty good experience,” he said.

So on first glance, the thing that jumps out is that it’s a bit hard to discern any actual data from this article. While there are numerous references to “no slowdowns” and “minimal speed disruptions or technology problems”, there is no detail as to:

  • What constitutes minimal? 1%? 10%?
  • How were slowdowns/disruptions measured and reported? Benchmarking? Subjective experience? Magic 8 ball?

And just how representative was the study? Any basic course in Statistics will tell you that in order to get a representative sample of any population, it is necessary to select randomly – and it would seem that much of the participation in the current trial was done on an opt in basis. This introduces several obvious potential skews to the results -with people opting in potentially being ideologically skewed towards supporting (or not supporting)  the the filter, and potentially less capable of measuring (or less likely to experience) the effects.

Even if you take this rather dubious announcement at face value, several major issues with the filter still have not been addressed:

  • None of the results reported measurement of  the effectiveness of the filter (either in blocking unwanted content or in avoiding trivial circumvention)
  • The ethics of censoring based on a secret list with no transparent mechanism of appeal, redress or checks to prevent misuse.

Despite the trial’s lack of actual objective or representative data (regarding slowdown or efficacy),  real discussion of the ethics of the situation  or any process or procedure remotely resembling science, Stephen Conroy plans to use it as a major input into governmental policy.  I’m not really surprised, just disappointed.


About this entry