More Common Than You Think
You might think this kind of filtering is an aberration, but it's not. Although different content filters react to different triggers, many have similar shortcomings. In fact, some of the most widely deployed content filters operate without any hint of finesse, intelligence or subtlety. For example:

Every time the E-mail version of my newsletter uses the word "format" (such as "file format") I get bounce-backs from filters that assume the plain-text word is really a malicious command ready to spring out and reformat the hard drives of unsuspecting readers.

Likewise, if I tell readers about ways to avoid a specific virus or worm, the mail will be bounced because the filters will see the mere mention of the plain-text name of the virus or worm and assume it's the actual malicious code.

To quote a person whose actual name is "Dick," I'll have to call him "Richard" to get past some filters.

If I want to obscure part of a URL, I can't use the letter X, because a URL with more than two Xs in it also will be blocked. The filters will think it's an "X-rated" site.

<

I could give many more examples, but you get the idea. These content filters are just plain stupid. There's usually no brainpower behind their blocking--they use just the crudest kind of brute-force pattern matching.

Despite all this hypersensitivity, literal-minded filters are ridiculously easy to defeat. That's why you may get spam-mail that talks about "p o r n" or "p.o.r.n" or "p/o/r/n." All anyone has to do is spell or punctuate a trigger word in a slightly nonstandard way, and the word sails through the filter undetected.

Microsoft Outlook and ISP Filters
The above examples are from filters that generate bounce-backs. Many other filters, including the one built into Microsoft Outlook, operate silently. Neither the sender nor recipient may be aware that mail was discarded by the filter.

For example, if the body of an Outlook E-mail message contains the phrase "for free" followed by an exclamation point, it'll trigger Outlook's built-in junk filter, and the mail will be trashed. Even the phrase "dear friend" can trigger the Outlook filter and cause an E-mail to end up in the trash. The sender isn't notified, and the recipient won't know the E-mail was processed unless she looks inside Outlook's trash folder.

Many ISPs have their own content filters, including AOL and MSN. They insert another level of clumsy screening into the E-mail process. Here's one famous example of how clumsy it can be. When AOL proudly rolled out new content filters with version 6, it immediately prevented its customers from accessing sites on subjects like breast cancer research, or joining sexual abuse support groups, because those sites and groups contained forbidden words. (AOL later loosened the restrictions so these valid uses could go on. I'd guess I'd have to say they made a clean breast of it.)

There also are URL-based filters, which can be almost as bad. For example, here's what happened when one IT professional, Scott Taylor, tried to visit the home page of LavaSoft, makers of the free and excellent "Ad-Aware" spyware-removal tool.

First, a little history. There was evidence some people were visiting porn sites at work, so I was directed to look into blocking them. [I chose the] WatchGuard Firebox II firewall. The Firebox offers WebBlocker service, which will filter URL requests against lists of known "offending" sites, on any number of criteria in addition to porn-related (violence, drugs, white supremacy, search engines (!), etc.). I enabled only "lewd, nude, or gross depictions" to be blocked.

So back to today's story. WebBlocker didn't like any of the LavaSoft URLs, and wouldn't let me anywhere near them. Makes you wonder if some Internet commercial interests are in bed with some Internet regulating interests, and the opportunity to keep people away from public-interest sites like LavaSoft is being abused.

More likely, this was simply a case of bad programming. But filters can indeed be subverted and used as a censorship tool, silently eliminating content and site access that the filter-makers or -minders don't want you to see.

For example, AOL was guilty of some outrageous filtering during the last U.S. presidential election. The AOL filters let users visit the home page of the Republican National Committee, but not the very similar Democratic National Committee site. Whatever your politics, and no matter if this was a programming error or someone's deliberate choice, you have to admit that such silent censorship is wrong.

Content filtering is emerging as a national issue in the United States. The Children's Internet Protection Act (CHIPA) is a new national law that sounds great at first blush. Who doesn't want to protect kids? But the law will force public librarians and public schools to use these clumsy, inefficient, and even harmful filters on all public-access computers. You may be surprised to learn that many librarians and teachers are dead-set against this law (see this memo, for example), not because they don't love kids, but because they don't want to have to act as societal thought police. Plus, the law is almost surely unconstitutional. In fact, as I write this, a suit brought against CHIPA by the American Library Association and others is in progress.

For lots more information on all kinds of content filter abuses, see the CensorWare Project.