We have been detecting which websites are blocked in China since February this year. Using our tool and contributions from our visitors, we've created the most complete database anywhere of what's blocked by the so-called Great Firewall of China, including 2200+ blocked websites and 200+ blocked searches. But how accurate is our data?
When a website is tested, it's always done by comparing the results of trying to access it from A) a computer outside China and B) a computer inside China. This way, a website that is simply not working is not labelled as blocked. A website that works when accessed outside of China but fails when tested from inside China, on the other hand, will be labelled as blocked.
This makes our data more accurate, but not perfect. Another difficulty is that the Great Firewall often changes temporarily. A given website that was previously accessible may be blocked for, say, 20 minutes. The opposite is also true. Blocked websites are sometimes temporarily accessible. Examples include Facebook (March 28, April 25) and YouTube (July 31, Aug 15 etc).
To tackle this particular challenge, our scripts always re-run testing of a website once more, after it's status has changed. In most cases, temporary changes are quickly detected and the websites are labelled correctly again. In some cases, however, if the change lasts somewhat longer, it may take up to a week for our system to re-label the website correctly. So, for example, Paypal was tested to be blocked twice on Aug 22. It's now accessible again. But it will take several days before our system automatically tests it again. Meanwhile, it's labelled as blocked.