Last month, the Libyan government temporarily cut off access to the Internet within the nation’s borders. The goal was to control the flow of information to the public and disrupt coordination among the demonstrators. The shutdown failed to do either, but for a while it threatened to have an odd side effect: impairing the functioning of websites using Libya’s “.ly” domains, including the popular service bit.ly, which millions use to turn long Web links into short ones that can be sent out on Twitter.
As it happened, backup domain servers in Oregon and Amsterdam kept bit.ly running through the blackout. But the incident drew attention to the fact that link-shortening services are now used to share huge numbers of links. New research from the Foundation for Research and Technology in Athens, Greece, and Microsoft Research in Cambridge, U.K., shows just how much is shared this way. And it suggests that the practice may be slowing down parts of the Internet.
The researchers analyzed millions of shortened links by crawling Twitter and generating and testing possible links for two shortening services, bit.ly and ow.ly. They found that a very small number of shortened links account for the majority of traffic, and that use of the shortening services occurs mainly in the United States, Japan, and the United Kingdom. It is virtually absent in India and China. Links to YouTube accounted for more than 10 percent of all bit.ly traffic.
According to Twitter, around 25 percent of tweets contain a URL. Character efficiency isn’t the only reason to use shortened links. Services like bit.ly also offer sophisticated analytics packages, allowing users to see when, where, and how people click on links. This makes it possible to gauge, for example, the popularity of stories by following external links from their Twitter or Facebook accounts. Many desktop or mobile Twitter clients now also offer built-in link-shortening services, which have led to an uptick in the use of such services. Shortened URLs have been embraced by media websites as well.
But inevitably, shortening services increase the amount of time it takes a Web browser to reach a URL. Shortening services including bit.ly, tinyurl.com, ow.ly, and fb.me, the researchers found, introduce a latency of between 50 and 600 percent. This translates to a delay of less than half a second, barely perceptible to most end users. But, the authors say, if usage continues to grow, the result will be “user-perceptible latency and an overall degradation of performance.” They also warn that “alternative shortening architectures for eliminating such overheads may be required in the future.”
Jon Kleinberg, an expert on the study of networks and professor of computer science at Cornell University, says link-shortening services offer a special opportunity for future research. “It has typically been difficult to study usage at this level of granularity, since the information about visits to many different Web pages didn’t necessarily all reside in a single easy-to-access location,” he says. “With shortening services, there is a single point through which the information about accesses will flow, and this makes it possible to undertake kinds of analysis that would have previously been much more difficult.”