In our 2005 testing, we located 115 Weblogs within 3 large blogging domains (blogsky,
blogspot, and persianblog) that were blocked in Iran. This blocking corresponded to only 24
percent of all blogs tested within those domains, and our testing was designed to locate
blocked sites. Clearly, Iran desires to block access to some blogs, but has not seen fit to
block all blogs. Our empirical data do not help to explain why filtering authorities in Iran
made this decision, but it clearly was the result of a deliberate action. Also note that the site
for www.movabletype.org, an application designed to allow blogging to take place on any domain, was blocked. Perhaps this indicates a policy of containing the blogs by restricting them
to the large blogging domains, where they can then be reviewed and potentially filtered on a
one-by-one basis.
China’s response to the same problem provides an instructive contrast. When China became worried about bloggers the state shut down the main blogging domains for a period of
weeks—much as they have, periodically, for Wikipedia. When the domains came back online,
they contained filters that would reject posts containing particular keywords.
30 In effect, China
moved to a content-based filtering system, but determined that the best place for such content evaluation was not the point of Web page access but the point of publication, and possessed the authority to force these filters on the downstream application provider. Most of
these providers coded these restrictions into the software provided to bloggers. This approach is similar to that taken with Google to respond to the accessibility of disfavored content via Google’s caching function. Google was blocked in China until a mechanism was put
in place to prevent cache access.
31 In the fall of 2005, Saudi Arabia was reported to have
blocked access to all blogs on the Blogger network, which plainly represented an overbroad
set of blocks. These examples make clear the length to which regimes can go to preserve
‘‘good’’ access instead of simply blocking an entire service.
Alternate approaches that demand a finer-grained means of filtering, such as the use of
automated keywords to identify and expunge sensitive information on the fly, or greater manual involvement in choosing individual Web pages to be filtered, are possible so long as a
state is willing to invest the time and resources necessary to render them effective. China in
particular appears to be prepared to make such an investment, one mirrored by choices
made by the Chinese state in the context of traditional media. For example, China allows
CNN to be broadcast within the country with a form of time delay so the feed can be temporarily turned off when, in one case, stories about the death of political reformer Zhao Ziyang
were broadcast.
Filtering and Underbreadth One of the primary surprises in our data over the past several
years is the infrequency with which plainly sensitive pages were blocked within otherwise acceptable sites. For instance, we found no cases where specific articles were blocked on major
news sites, except in China. In fact, the regimes in which we tested very rarely made an attempt to block www.cnn.com, www.nytimes.com, http://bbc.co.uk, or others. (Exceptions to
48
Jonathan Zittrain and John Palfrey
blogspot, and persianblog) that were blocked in Iran. This blocking corresponded to only 24
percent of all blogs tested within those domains, and our testing was designed to locate
blocked sites. Clearly, Iran desires to block access to some blogs, but has not seen fit to
block all blogs. Our empirical data do not help to explain why filtering authorities in Iran
made this decision, but it clearly was the result of a deliberate action. Also note that the site
for www.movabletype.org, an application designed to allow blogging to take place on any domain, was blocked. Perhaps this indicates a policy of containing the blogs by restricting them
to the large blogging domains, where they can then be reviewed and potentially filtered on a
one-by-one basis.
China’s response to the same problem provides an instructive contrast. When China became worried about bloggers the state shut down the main blogging domains for a period of
weeks—much as they have, periodically, for Wikipedia. When the domains came back online,
they contained filters that would reject posts containing particular keywords.
30 In effect, China
moved to a content-based filtering system, but determined that the best place for such content evaluation was not the point of Web page access but the point of publication, and possessed the authority to force these filters on the downstream application provider. Most of
these providers coded these restrictions into the software provided to bloggers. This approach is similar to that taken with Google to respond to the accessibility of disfavored content via Google’s caching function. Google was blocked in China until a mechanism was put
in place to prevent cache access.
31 In the fall of 2005, Saudi Arabia was reported to have
blocked access to all blogs on the Blogger network, which plainly represented an overbroad
set of blocks. These examples make clear the length to which regimes can go to preserve
‘‘good’’ access instead of simply blocking an entire service.
Alternate approaches that demand a finer-grained means of filtering, such as the use of
automated keywords to identify and expunge sensitive information on the fly, or greater manual involvement in choosing individual Web pages to be filtered, are possible so long as a
state is willing to invest the time and resources necessary to render them effective. China in
particular appears to be prepared to make such an investment, one mirrored by choices
made by the Chinese state in the context of traditional media. For example, China allows
CNN to be broadcast within the country with a form of time delay so the feed can be temporarily turned off when, in one case, stories about the death of political reformer Zhao Ziyang
were broadcast.
Filtering and Underbreadth One of the primary surprises in our data over the past several
years is the infrequency with which plainly sensitive pages were blocked within otherwise acceptable sites. For instance, we found no cases where specific articles were blocked on major
news sites, except in China. In fact, the regimes in which we tested very rarely made an attempt to block www.cnn.com, www.nytimes.com, http://bbc.co.uk, or others. (Exceptions to
48
Jonathan Zittrain and John Palfrey