While the human rights aspects fuel my passion on this issue, I think that since I’m a literate geek it’s more my role to help explain the “technical difficulties” (!) of what’s being proposed.

I’ll read your comments on your own site as I head north on the train later this morning.

In particular, I offer a more extended explanation of why overlay networks such as those established by Tor make ISP-level filtering irrelevant and hence futile.

It doesn’t matter if they censor content that I am interested in. The only thing that matters is that they have assumed the right do it if they choose. That’s what ticks us off, and until they understand that, they will continue to be deluded.

This would probably be better phrased as:

Whether or not they actually censor material we are likely to read is not the issue. The issue is that they are attempting to build a censorship mechanism at all. Viewing child pornography is already a crime, those wishing to view it can and will subvert any filter. The correct response to child pornography is and continues to be law enforcement. In the age of the Internet, any attempt at censorship is futile at best and most likely counter-productive since it simply encourages widespread adoption of covert channels.

As Gabriel notes, censorship is futile and dogged attempts to implement a futile idea are stupid. And as Michael notes, stupid attempts to implement futile ideas in the face of well-intentioned and well-informed criticism are enough to enrage even the mildest of (idealistic) geeks.

Again, as Michael says, I think we do owe it to the pro-censorship lobby to explain the inner workings of Tor hidden services so they understand how intensely futile censorship is. Sites that advertise their perverted wares via a Tor hidden service (or something equivalent) are completely invisible as far the rest of the Internet is concerned. A consumer of child pornography can download stuff from a Tor hidden service and not have a clue what the physical IP address of the server is and likewise the provider will not have a clue what the physical IP address of the client is. With technology like that available, you could block every single pornography site on the web and still not impede access to a child porn site one single iota. The only way to defeat Tor (and Tor-like services) is to physically unplug everyone’s DSL routers and smash their wireless routers. It really is that futile.

The difference between those who support this plan and those who don’t is simply a matter of who understands the technology and who doesn’t.

@Michael Meloni: I have no control over where in Crikey my pieces appear. However I’ll re-publish it here the following day.

Because there is a general rule of thumb that can be applied to doing things that don’t work… it’s stupid.

Hitting my car with a hammer to make it run faster doesn’t work, therefore hitting my car with a hammer in an attempt to make it run faster is a stupid thing to do. Attempting to block child pornography by running automated filterbots over everyone’s internet connection doesn’t work, therefore running automated filterbots over everyone’s internet connection in an attempt to block child pornography is a stupid thing to do.

The difference between those who support this plan and those who don’t is simply a matter of who understands the technology and who doesn’t - in the same way that various “flat-tax” plans seem to garner support among those who don’t have much of an idea about economics and hostility among those who do. Fortunately for the economy, the number of people who understand how a flat tax is a bad idea comfortably outweighs the number who don’t. Unfortunately for the country’s IT&T sector, that ratio is rather more precariously balanced when it comes to understanding filtering.

Filtering traffic against a known black-list of illegal sites is one thing. What the government is proposing is some sort of automated technology it believes will automatically be able to differentiate between legal and illegal content ‘on-the-fly’ (as it is being requested by and delivered to a consumer). Even limiting such technology to filtering words (leaving aside pictures and videos for the moment) is not overly effective. It has almost become a cliched argument in anti-filtering circles that if you try and screen out the word “breast” then you run a very real risk of denying people access to information regarding mammograms or how best to grill chicken.

The English language has a finite (large, but finite) list of words, only a few of which are ‘naughty’ enough to warrant ending up on any reasonable person’s filter list. Running a query across a page of text to check for naughty words or phrases is a fairly simple process. However, even aside from the ‘false positive’ example of the innocuous use of words like ‘breast’ there is an almost certain chance of acquiring ‘false negative’ results. A false negative is where the filter gives the page a clean bill of health even though it ought - by the principles of the filter - to have been blocked. Blocking words like “cock” and “cunt” won’t pick up euphemistic phrases such as “skin flute” or “bearded clam”.

Extrapolating such difficulties out to filtering photographs presents challenges that are an order of magnitude greater than when dealling with text. Digitised photographs are represented as a two dimensional array of coloured dots. Unlike words, coloured dots impart absolutely no meaning whatsoever as to the content of the photograph. Depending on the lighting and angle of the shot, a gynaecological closeup could be represented by much the same array of coloured dots as a photograph of a person’s hand, or even a pig on a farm. Filtering software attempts to define ‘patterns of dots’ that - as a rule - depict things like “large expanse of flesh tones” - or even with cutting edge technology - “human face”. Expecting an automatic filter to examine a photograph and determine whether it depicted an innocent shot of dad on the beach, or a ‘model’ putting her ankles behind her ears is simply not within the realms of technological reality.

Bring the difference between ‘legal’ and ‘illegal’ down even further and the task being asked of a machine is beyond that which even the human brain is capable. As a society we use the magic number 18 as a yardstick for all sorts of things, including the age at which it is legal to appear in a pornographic photograph. At 216 months, a woman is deemed legally able to be photographed naked in exchange for money, at 215 months she is not. Picking people’s ages is not an easy task. I have a hard time picking young people’s ages at anywhere between 16 and 24 these days, I certainly wouldn’t be able to ascertain someone’s age to the month from simply looking at a photograph. It is easy to see how unrealistic the expectations that people have of filtering software are.

Even if we could engineer a software routine that could determine the age of the subject of a photograph to the level of granularity required to “just block child porn and not block anything that isn’t child porn”, the amount of processing time and effort required for such a routine to run would be immense. Consider that a digitised video is a series of digital images displayed one after the other at a rate of 30 images per second and we would have to run the same process (once for every frame of the video) nearly three thousand times just to cover a minute and a half of digitised video.

I would be very surprised if there was any true ‘hostility’ to the notion of eradicating child pornography from either the internet or the community at large. The percieved hostility is not directed toward the aim of making the world a safer place for children, it is directed at the wholesale ‘magic thinking’ that seems to have gripped people regarding the capabilities of filtering technology. There is simply no way, using contempory hardware and software, that the internet can be automatically filtered of child pornography. And there is certainly no way that it can be even attempted without impacting both the availability and speed of access of perfectly legal material.

Crikey will be running her letter today. I’m writing a reply as we speak, essentially a refined version of the discussion we’re having. Both should go online around 2pm Sydney time. You’ll get a mention.

[Update 12.45pm: My response has been held back until tomorrow's Crikey, but I'm told it'll definitely run then.]

I hope you understand soon Bernadette why the majority of people with technical know how, especially those with internet knowledge, disagree with mandatory filtering. It’s simply a bad idea that benefits no one except moral crusaders and politicians. Children certainly aren’t helped.

- Michael

1. Two completely different problems are being conflated. One is preventing the distribution of already-illegal child pornography. The other is preventing children from viewing an extremely broad category of material which is deemed adults-only, but which everyone else is permitted to see. The different problems might well need different solutions, but they’re being jumbled together in what many people, myself included, see as a political act.
2. People are asking for technical solutions to social problems. While faith in The Power of Technology is wonderful, technology isn’t necessarily the solution to society’s problems.
3. Technically illiterate people are demanding a specific technical solution: filters. Those in the know are already several pages ahead of the story, and already know this won’t work. Geeks get angry when they feel their knowledge is not being respected.

You’re quite right when you say that we, geeks in general, need to explain ourselves more clearly. The old-fashioned BOFH approach of attacking the lusers is well past its use-by date!

@Neil H: Thanks for the pointer, wouldn’t have seen that article otherwise. It helped me find another one.

@Michael: Exactly. The anger comes from point 3 above.

@rene: Thanks for the BT pointers, and welcome! Yes, the message from real-world experience always seems to be that filters only block the casual user from accidentally stumbling across something they don’t want to see. Those who are specifically trying to find the material will always find it. The filters work just well enough that politicians can be shown a “convincing” demonstration.

@Jon Seymour: That is a very fine summary of the sources of anger.

On this point…

In a recent post I asked, somewhat tounge in cheek, if proponents of censorship would be willing to trade their right to privacy if this helped in the battle against child pornography. If they are not prepared to do this, what right do they have to ask that we live in a state that has the mechanisms to censor content at will?

… I suspect one problem is that many of the proponents of the filters would say that, yes, they are happy to lose their privacy. Many seem to believe that old adage that “if you’ve got nothing to hide you shouldn’t be afraid”, in the rather naive assumption that those doing the looking will only ever have the same moral and social values as themselves, and wouldn’t ever be people who’d try to do them harm.

[P.S. I've slightly re-formatted some comments and corrected spelling and punctuation errors. I just hate my web pages looking ugly. Keep those comments coming, though! I'm still putting together a formal response to Ms McMenamin's letter.]