that would be tolerated independently and those that are explicitly meant to be subject to filtering. Likewise, we found minimal blocking for internal political purposes in UAE, but the state
did block a site that essentially acted as a catalog of criticism of the state. Our tests cannot
determine whether it was the material covering human rights abuses or discussion of historical border disputes with Iran, but in as much as the discussion of these topics is taking place
within a broad dissention-based site, the calculation we project onto the censor looks significantly different than that for a site with a different ratio of ‘‘offensive’’ to approved content.
For those states using commercial filtering software and update services to try to maintain a
current list of blocked sites matching particular criteria, we have noted multiple instances
where such software has mistaken sites containing gay and lesbian content for pornography.
For instance, the site for the Log Cabin Republicans of Texas has been blocked by the United
States–based SmartFilter as pornography, apparently the basis for its blocking by United
Arab Emirates. Our research suggests that gay and lesbian content is itself often targeted for
filtering; one might surmise that, even when it is not explicitly targeted, states that implement
related filters are not overly concerned with its unavailability due to overbreadth.
As content changes increasingly quickly on the Web and generalizations become more difficult to make by URL or domain name—thanks in part to the rise of simpler, faster, and
aggregated publishing tools such as Weblogging (blogging) services—accurate filtering is
likely to get trickier for filtering regimes to address over time unless they want to take the
step of banning nearly everything. For example, free Web-hosting domains tend to group an
enormous array of changing content and thus provoke very different responses from state
governments. In 2004, Saudi Arabia blocked every page we tested on http://freespace.virgin
.net and www.erols.com.
29 However, our research indicated the www.erols.com sites had
been only minimally blocked in 2002, and the http://freespace.virgin.net sites had been
blocked in 2002, but accessible in 2003 before being reblocked in 2004. In all three tests,
Saudi Arabia practiced URL blocking on www.geocities.com (possibly through SmartFilter
categorization), blocking only 3 percent of more than one thousand sites tested in 2004. Vietnam blocked all sites we tested on the www.geocities.com and http://members.tripod.org
domains.
Contrast this last example with Yahoo! Groups, which Vietnam appears to filter on a groupby-group basis. We found that the state blocks access to the pages of two groups discussing
the Cao Dai religion in general, but our testers were able to access the page of a California
Cao Dai youth group. Two factors may play a role in this decision. Groups may provide more
‘‘benefit’’ to the censor, due to their interactive nature, and thus implicate the social and possibly economic impacts of the Internet. Groups, too, may have a limited, albeit large, number of possibilities—a single group could, in theory be monitored at the group level where
there is much more metadata about the content contained therein, whereas Geocities could
be grouped by user, but a particular user may offer large numbers of pages on very varied
topics.
Internet Filtering: The Politics and Mechanisms of Control
47
did block a site that essentially acted as a catalog of criticism of the state. Our tests cannot
determine whether it was the material covering human rights abuses or discussion of historical border disputes with Iran, but in as much as the discussion of these topics is taking place
within a broad dissention-based site, the calculation we project onto the censor looks significantly different than that for a site with a different ratio of ‘‘offensive’’ to approved content.
For those states using commercial filtering software and update services to try to maintain a
current list of blocked sites matching particular criteria, we have noted multiple instances
where such software has mistaken sites containing gay and lesbian content for pornography.
For instance, the site for the Log Cabin Republicans of Texas has been blocked by the United
States–based SmartFilter as pornography, apparently the basis for its blocking by United
Arab Emirates. Our research suggests that gay and lesbian content is itself often targeted for
filtering; one might surmise that, even when it is not explicitly targeted, states that implement
related filters are not overly concerned with its unavailability due to overbreadth.
As content changes increasingly quickly on the Web and generalizations become more difficult to make by URL or domain name—thanks in part to the rise of simpler, faster, and
aggregated publishing tools such as Weblogging (blogging) services—accurate filtering is
likely to get trickier for filtering regimes to address over time unless they want to take the
step of banning nearly everything. For example, free Web-hosting domains tend to group an
enormous array of changing content and thus provoke very different responses from state
governments. In 2004, Saudi Arabia blocked every page we tested on http://freespace.virgin
.net and www.erols.com.
29 However, our research indicated the www.erols.com sites had
been only minimally blocked in 2002, and the http://freespace.virgin.net sites had been
blocked in 2002, but accessible in 2003 before being reblocked in 2004. In all three tests,
Saudi Arabia practiced URL blocking on www.geocities.com (possibly through SmartFilter
categorization), blocking only 3 percent of more than one thousand sites tested in 2004. Vietnam blocked all sites we tested on the www.geocities.com and http://members.tripod.org
domains.
Contrast this last example with Yahoo! Groups, which Vietnam appears to filter on a groupby-group basis. We found that the state blocks access to the pages of two groups discussing
the Cao Dai religion in general, but our testers were able to access the page of a California
Cao Dai youth group. Two factors may play a role in this decision. Groups may provide more
‘‘benefit’’ to the censor, due to their interactive nature, and thus implicate the social and possibly economic impacts of the Internet. Groups, too, may have a limited, albeit large, number of possibilities—a single group could, in theory be monitored at the group level where
there is much more metadata about the content contained therein, whereas Geocities could
be grouped by user, but a particular user may offer large numbers of pages on very varied
topics.
Internet Filtering: The Politics and Mechanisms of Control
47