The government has released the spec documents for its proposed compulsory internet filtering system. The problem? They're a confusing mess of vague waffle.
The specifications documents for ISPs helping to test the proposed "clean feed" could do with some cleaning up themselves if they're going to be used as a basis for any kind of realistic testing.
One of the key reasons for running technical tests before introducing mandatory national ISP filtering is to determine the impact that such filters might have on typical Internet access.
As Senator Stephen Conroy explained during a speech to the Internet Industry Association back in February: "I acknowledge industry concerns that filtering will affect internet speeds or will result in over-blocking or under-blocking. These issues will be addressed through the trial process."
However, if the Technical Testing Framework document issued to ISPs who want to participate in the trials is any guide, no-one has much idea what approaches will be used, what's likely to get blocked, or what results will be deemed acceptable. There's so many appearances of the word "ideally" that it seems that compromise is the only likely goal.
The government gives ISPs the option to test either simple blacklist filtering (using the list of sites already banned by the Australian Communications and Media Authority, ACMA) or a broader range of technologies, such as dynamic analysis filtering, DNS poisoning and checking IP address lists.
However, even the blacklist requirement is rather vague. The testing system is supposed to handle "a blacklist of up to 10,000 URLs". Yet ACMA's own list of prohibited sites contains just 1,300 banned URLs. The numbers will apparently be made up with "internationally available blacklists", though even this is far from certain: "It should also be noted that this is also expected to fluctuate over time."
There are other contradictory elements in the test document. Testing companies will ideally offer "individual user and/or management controls (to switch filtering on/off)", but this isn't compulsory. "While customers may volunteer for the Pilot, during their participation they would not be aware at any particular point in time of whether or not they are receiving a filtered service, depending on the kind of filtering solution being tested," the requirements note.
However, one of the key factors to be considered is "ease of use from a user perspective", which would be near-impossible to measure via surveys if the user isn't aware whether they're receiving the service or not or isn't given the option to switch it off. Consumers are also supposed to be able to offer examples of incorrectly blocked sites or sites that should be banned; this hardly makes sense if they're not sure if the service is switched on or not.
The testing environment is supposed to cover speeds "from 56Kbps through to 12Mbps", which is inadequate at both ends. As any phone owner will attest, mobile broadband speeds can often drop below the prior figure, while the upper limit is well below what the much-vaunted National Broadband Network is poised to offer all Australians.
The contradictory nature doesn't help ISPs decide the vexed question of whether to participate. If they don't, they effectively lose the right to complain if the filter scheme ultimately becomes compulsory and renders their service so slow or ineffective as to be commercially unviable. If they do, they'll have to spend an inordinate amount of time and cost trying to make sense of the vaguely-specified rules.