Sunday, December 07, 2008

Wikimedia vs. Internet Watch Foundation

The tech media are all abuzz over how the Internet Watch Foundation has recently added a page on Wikipedia to its "list of pages that contain child porn". This has had the effect of causing all web traffic to any Wikimedia site from any customer of the several ISPs that subscribe to the IWF's filtering service to be forced through IWF's proxy filter so they can block access to the (so far) one page that they've found that contains child porn. A side effect of this is that Wikipedia (the English Wikipedia, at least) has had to block all editing from anonymous persons on the affected ISPs, because they cannot "differentiate different users, and block those abusing the site without blocking other people as well."

Most British ISPs voluntary agree to filter web traffic through their networks for content which violates UK law. Most of them use the services of the Internet Watch Foundation, which is a non-governmental organization with close ties to the Home Office and other law enforcement entities, but not a law enforcement organization in and of itself. I haven't been able to dig out the details of the 1996 agreement that created the IWF, but I get the feeling that there was an undertone of "you do this voluntarily or we will force you to do it" and the ISPs agreed to it because they didn't want the negative PR of appearing to be on the side of child pornographers. Anyway, the IWF provides its clients with lists of URLs that contain "unlawful" content, which its clients are required (by contract) to block. They may not manually review the lists, but must use them exactly as provided by IWF; the IWF is the sole party responsible for the lists.

Wikipedia has an article about a 1970s album by the Scorpions (a German hard rock band) titled "Virgin Killer". The original cover for this album featured a naked prepubescent female with a "cracked-glass" effect covering her genitals; this cover was extremely controversial and was rather quickly withdrawn and replaced by a much tamer cover for most distributions, although it is still possible to buy copies with the controversial cover on it, even in the UK.

Someone (anonymously, presumably) reported the page to the IWF as a "webpage containing potentially illegal content". The IWF reviewed the article and concluded that the image "may be illegal in the United Kingdom" and added the article to its blacklist. Doing so forced the subscribers to IWF's blacklist to force all traffic to any site hosted on the same server as en.wikipedia.org (which is to say virtually all Wikimedia sites) to be passed through the filtering proxy in order to detect and block any attempts to access the article. As a side effect, this caused everyone who goes through the filters to appear to Wikipedia's security systems to appear to be coming from the same IP, which significantly undermines Wikipedia's internal security measures; in addition, at least some of the proxy filters are unable to cope with the load and are reportedly generating random page load failures and other problems in addition to blocking access to the page in question.

Somewhat strangely, IWF did not block the image itself; they only blocked the Virgin Killer article and the "image page" for the image. If the image were to appear on another article, that article would be left untouched, and the image would be displayed. The image in question appears on articles on at least four other Wikimedia projects, none of which were being filtered at last report.

Wikipedia has responded with a cavalcade of screeching about the evils of censorship; a handful of editors who have argued for removing the image have been shouted down and in some cases blocked from editing. Although I have seen no public statement from the Foundation, rumor has it that there have been conversations between the IWF and Sue Gardner (WMF's executive director), and the WMF has sent demands to the ISPs to discontinue the filtering. There appears to be no possibility that Wikipedia will consider removing the image, or even diminish its prominence. Some editors even argued for a hard block of all UK IPs "in order to send a message" that censorship is wrong. In addition, the raw inconvenience that has been imposed technically on UK editors has profoundly upset many UK editors.

There's so much to comment on here. On one hand, we have ISPs imposing filtering that their customers haven't specifically asked for, using filter lists from a self-governing third party that appears to be responsible to no one, so as to require the filtering of an image that is almost certainly not illegal under the relevant law. The filtering is done so badly that it fails to block most uses of the content that is alleged to be illegal, and also blocks content which is not alleged to be illegal (specifically, the text of the article about the album, which is not alleged to violate any law at all). Finally, the technical solution used is apparently not robust enough to handle the traffic caused by forcing most of the UK Wikimedia-related traffic through its servers.

On the other hand, we have the ideologically detached Wikipedia community that has steadfastly refused to contemplate that there are topics, and especially images, which, while they should be available, ought to be placed behind "shields" of some sort so that they are not unexpectedly displayed to readers who are likely to be offended by them. The Wikimedia Commons by now quite likely has the largest collection of free porn on the Internet, and there are just a plethora of articles on Wikipedia about thoroughly prurient topics that are certainly of interest to some, but at the same time are not the sort of things that most people will want children exposed to uncontrollably. Wikipedia's community, however, has steadfastly insisted that "Wikipedia is not censored" and thus steadfastly, even defiantly, almost always refuses to conceal, minimize, or shield prurient, offensive, or shocking images. (Thus, if you happen to be reading Wikipedia from work, absolutely do not hit "random article"; you might well end up with an image of gay sex on your screen.)

Now, I'm a pretty committed civil libertarian; I think censorship is wrong, and I object to the imposition of content filtering on anyone's internet connection for any purpose except at the express request of the customer. So it's very much my wish that this particular filtering scheme be attacked on general principles. However, it's clear to me that this filtering scheme is well-established in Britain (it's been in use on at least one of the ISPs there since 2004, based on one article I read) and there doesn't seem to be much public dissent to its use. Not being British I don't have a really big dog in this fight, and there are far worse offenders on this issue than the UK.

I also take issue with the IWF filtering the article instead of the image. Surely their technical people realize that images within a webpage load independently of the webpage itself, and that they can block just the image without blocking the rest of the article. I question why they did not do this in this case, as it would have been the least intrusive means to accomplish this.

On the technical issue regarding the IP appearance issue, the reports are that the IWF is willing to work with the WMF to sort out a solution that avoids this. If I understand the situation, this is just a matter of them ensuring that XFF works correctly in their proxy, and that their proxies are properly registered with the WMF.

The one thing that would have probably rendered this situation less problematic would have been if the IWF had contacted the WMF when it reached the determination that the content was illegal under UK law. Not only could the WMF have advised the IWF on the least intrusive manner to block the image, but they could also have ensured that the necessary forwarding configuration was in place to avoid the disruption that occurred over the past few days. That said, I suspect that had the IWF attempted to contact the WMF, they would have gotten nowhere; it's likely that the WMF would have either ignored the communication, or responded with hostility instead of attempting to cooperate, or to negotiate a compromise solution.

It's my position that this image should not be blocked. The fact that the exact same image is in use on Amazon UK's site without any action taken against Amazon suggests that there's hidden motivations on the IWF's part here. And it seems quite likely that the WMF's general incompetence combined with the complete lack of community leadership within the English Wikipedia will combined to prevent at sort of compromise solution; the WMF has nobody with the competency to negotiate one, and even if they did there is no effective way to impose it on the community because of the lack of leadership within the community. So I strongly suspect that the IWF intends to stand strong on this one, secure in the understanding that this is one they can win.

Go ahead, Jimmy, Sue, and Mike. Prove me wrong. I dare you.

P.S. Other coverage: Seth Finkelstein, Danny Wool