Wednesday, December 30, 2009

A curious case of government pressure


News today of an attempt by the Canadian government to get a German ISP to censor a political parody site that was critical of the Canadian Government.

I'd be wary of making too much of this in relation to the Australian debate for several reasons.

The Canadian Government agency does not have jurisdiction over the German ISP and there is no suggestion, as far as I can tell, that they asked the German Government to assist with the censorship action. In other words: the Canadian Government wasn't using its monopoly on the use of force on Canadian soil to force the German ISP to act in a certain way.

Secondly, the collateral damage caused to the 4500 sites is most likely the result of incompetence on the German ISPs part, rather than the Canadian Government's part.

Yes, it does illustrate the dangers of a Government that thinks it can throw its weight around, but it doesn't really have too much to say about the Australian debate since it wasn't a case of the state acting against a commercial entity within the state. It was a case of an agency of a state asking a commercial entity of another state for assistance. One can argue that the German ISP should have told the Canadian Government to bugger off, but that is a different argument.

1 comment:

Stilgherrian said...

Two issues stand out for me.

One is that the German ISP decided to respond to a request from a foreign government. OK, perhaps they asked nicely, but it's interesting that they didn't put their own customer's interests first.

More interesting, though, is the collateral damage.

Sure, just a mistake, but it does demonstrate the risk to 4500 unrelated businesses and other entities from a single error. A highly networked world with big data centres means one mistake can have massive consequences.

When constructing policies, the risk assessment of "getting it wrong" needs to consider this scale of potential impact, not just that one business or entity could be affected by the mistake. What error rates are actually acceptable in this environment? Very, very tiny ones, I'd suggest.