states use a combination of media, telecommunications, national security, and Internetspecific laws and regulatory schemes to restrict the publication of and access to information
on the Internet. Most states that filter require Internet service providers to obtain licenses
before providing Internet access to citizens. Some states place pressure on cybercafe ´s and
ISPs to monitor Internet usage by their customers. With the exception of a few places, no state
seems to communicate much at all with the public about its process for blocking and unblocking content on the Internet. Most states, instead, have only a series of broad laws that
cover content issues online. The effect of these regimes is to put citizens on notice that they
should not publish or access content online that violates certain norms and to create a sense
that someone might be paying attention to their online activity.
Our global survey of Internet filtering in 2006 turned up instances where the Internet is not
subject to online filtering, but where the state manages to dampen online dissidence through
other means. In Egypt, for instance, the Internet is not filtered (reports suggest that Egypt at
one time blocked the Muslim Brotherhood site, but did not appear to do so during our testing),
13 but security forces have detained people for their activities online.
14 The perception
that the online space is subject to extensive state surveillance leads to a broad fear of accessing or publishing information online that may be perceived to be subversive—though bloggers
and other online activists cross the perceived lines with regularity.
15
The Spectrum from Manual to Automatic Filtering
Most of the filtering regimes we studied, with the exception of parts of the Chinese filtering regime, appear to rely on the preidentification and categorization of undesirable Web sites. As
the Web grows in scope as well as form, it is likely that states with an interest in filtering will
attempt to develop or obtain technology to automatically review or generalize about the content of a Web page as it is accessed, or other Internet communication as it happens. The Web
2.0 phenomenon only makes this challenge harder, as citizens have the ability to publish online content on the fly and to syndicate that content for free.
The job of the censor in a Web 2.0 world might or might not be accomplished by looking for
certain keywords in the title of the page or on its link—its URL. While URLs are clearly not as
determinative of content as the name of a television channel or a newspaper, they may be in
some situations (consider what generalizations one might draw about a URL of www.google
.com/search?q=tiananmen). In others, a URL may serve as an adequate proxy—URLs containing particular obscene words are more likely to have obscene content. If the goal is to
block all content coming from a particular state, the top-level domain structure makes this remarkably simple. On the other end of the spectrum, however, are blogging sites or generic
free Web-hosting sites like www.geocities.com, where the presence of a page within the general site provides little information about the content or authorship of other pages on the site.
Despite the obvious imperfections of filtering via URLs, we have found little evidence that
the states in which we tested, with the exception of China, attempted a dynamic assessment
36
Jonathan Zittrain and John Palfrey
on the Internet. Most states that filter require Internet service providers to obtain licenses
before providing Internet access to citizens. Some states place pressure on cybercafe ´s and
ISPs to monitor Internet usage by their customers. With the exception of a few places, no state
seems to communicate much at all with the public about its process for blocking and unblocking content on the Internet. Most states, instead, have only a series of broad laws that
cover content issues online. The effect of these regimes is to put citizens on notice that they
should not publish or access content online that violates certain norms and to create a sense
that someone might be paying attention to their online activity.
Our global survey of Internet filtering in 2006 turned up instances where the Internet is not
subject to online filtering, but where the state manages to dampen online dissidence through
other means. In Egypt, for instance, the Internet is not filtered (reports suggest that Egypt at
one time blocked the Muslim Brotherhood site, but did not appear to do so during our testing),
13 but security forces have detained people for their activities online.
14 The perception
that the online space is subject to extensive state surveillance leads to a broad fear of accessing or publishing information online that may be perceived to be subversive—though bloggers
and other online activists cross the perceived lines with regularity.
15
The Spectrum from Manual to Automatic Filtering
Most of the filtering regimes we studied, with the exception of parts of the Chinese filtering regime, appear to rely on the preidentification and categorization of undesirable Web sites. As
the Web grows in scope as well as form, it is likely that states with an interest in filtering will
attempt to develop or obtain technology to automatically review or generalize about the content of a Web page as it is accessed, or other Internet communication as it happens. The Web
2.0 phenomenon only makes this challenge harder, as citizens have the ability to publish online content on the fly and to syndicate that content for free.
The job of the censor in a Web 2.0 world might or might not be accomplished by looking for
certain keywords in the title of the page or on its link—its URL. While URLs are clearly not as
determinative of content as the name of a television channel or a newspaper, they may be in
some situations (consider what generalizations one might draw about a URL of www.google
.com/search?q=tiananmen). In others, a URL may serve as an adequate proxy—URLs containing particular obscene words are more likely to have obscene content. If the goal is to
block all content coming from a particular state, the top-level domain structure makes this remarkably simple. On the other end of the spectrum, however, are blogging sites or generic
free Web-hosting sites like www.geocities.com, where the presence of a page within the general site provides little information about the content or authorship of other pages on the site.
Despite the obvious imperfections of filtering via URLs, we have found little evidence that
the states in which we tested, with the exception of China, attempted a dynamic assessment
36
Jonathan Zittrain and John Palfrey