MMS Friends

Tuesday, January 31, 2006

In Defense of Google

So i'll admit that when i first heard of Google's recent concession to the Chinese government, i thought Google had finally become evil. Just goes to show you how much faith one can put in the Palo Alto Daily News. (You will notice that i am not including any direct quotes from the Palo Alto Daily News here, but that is only becuase i cannot find any way to access the paper online.) When i read that Google was censoring search results available to Chinese users, i assumed it would work like this: The Chinese government would supply Google with a list of Chinese IP addresses, Google would filter every request coming into its US-based servers, and Google would return a restricted set of results if the IP address matched the Chinese list. In fact, i can think of no other effective mechanism to censor the search results available to Chinese users. And if Google used this mechanism, that would indeed be evil..

Thanks to this week's issue of The Economist and an NPR broadcast last week, i've learned a little more about the arrangement though. Google is not blocking its US-based servers at all. What is happening instead is that Google is setting up servers in China itself, and my understanding is that as a condition of this arrangement Google is required to adhere to certain guidelines. According to several articles i've read, this has been standard Google practice in France and Germany for quite some time now. These local servers are put in place to augment the US servers and speed up certain searches, and to my knowledge there is no effort being made by Google or the Chinese government to block access to Google's servers in the US. So ultimately Chinese access to Google is not restricted at all; on the contrary, it is enhanced..

Not only is there a net advantage to Google's placement of servers in China, there is no loss anywhere in the process. Many searches (the non-controversial ones) are sped up, while the banned searches that must go through US servers are NO SLOWER than they were before. Of course, one possible disadvantage to this arrangement is the risk that someone doing a questionable search would stop at the Chinese servers, thinking the censored results were complete. However, Google has taken care of this problem by alerting Chinese users whenever search results have been censored. Once such a notice comes on the screen, the Chinese user can search on Google's US-based servers for the information that was censored..

The only efficiency problem this poses is a waste of time for the individual who goes to the Chinese server first only to find out she must subsequently search on the US server. Okay, let's run some numbers on this. Assuming that the time required for a search is independent of the topic of the search, there is going to be some threshold percentage of censored searches for a user beyond which it is more efficient to use the US servers for all searches. Let's assume that the Chinese servers are x times as fast for the Chinese user as the US servers would be. The break-even point is then (x-1)/x*. In other words, if the Chinese servers are four times as fast, more than 3/4 of the searches must be censored before there is a net loss of efficiency for the Chinese user who unilaterally uses the Chinese servers first..

Of course, this is all assuming the Chinese user is completely oblivious to the issue of whether her search terms will engender censored results. But i submit that well before the threshold is reached, any reasonable user will have developed some schema for determining a priori whether her search results will likely be censored. And even if the user is unable to predict that some of her searches are more likely to be censored than others, by the time she reaches the threshold she will know to run all her searches on US servers anyway. Combine this with the observation that Google would not be placing servers in China in the first place unless they represented some significant increase in speed (measured by x in the previous paragraph), and it is inevitable that even for the users who have to search twice to get their censored results Google's decision to restrict content in order to place servers in China is beneficial..

* i got this figure by assuming it takes on average t seconds to carry out a search using a US server and therefore t/x seconds on a Chinese server. i let p be the proportion of time a result is censored, and i compared the average times required to search based on two strategies: the first where the search is always done on a US server, the second where the search is always started on a Chinese server and then a US server is used if the results are censored. The equation i get is: t = p(t+t/x) + (1-p)t/x. The t's drop out, and p can be solved in terms of x..

0 Comments:

Post a Comment

<< Home