Rediff.com« Back to articlePrint this article

Google's invisibility cloak

December 09, 2008 14:18 IST

Google likes to portray itself as a company that does everything in the open. But it appears that at least some of its employees are harboring a secret.

Web researcher Net Applications recently discovered that between 11 to 30 per cent of traffic streaming out of Google's Mountain View, Calif., office is stripped clean of the usual identifying information that accompanies such traffic. That begs the question: What secret is it that Google doesn't want the rest of the Web to know?

The finding, first reported in InternetNews.com, quickly sparked online chatter about a Windows challenger in the works. "I'd be shocked if Google wasn't developing its own operating system," says Vince Vizzacarro, Net Applications' executive vice president of marketing. "They clearly want to ride online services without using Microsoft."

The Aliso Viejo, Calif.-based firm made the discovery after adding a new feature to its analytics software that pinpoints the source of Internet traffic down to a specific company. The technology, which tracks various trends in Internet usage by analyzing traffic to more than 40,000 Web sites around the world, can also detect a computer's browser, IP address, referring search engine or search term, default language and screen resolution.

The new data showed that a percentage of Internet users in Google's offices (principally based in the company's Mountain View headquarters) are using an operating system that essentially shields itself from detection by stripping traffic of identifying information. Vizzacarro describes this data, known as a user agent, as a string of information that a computer uses to identify itself.

Removing it (possibly via a proxy server) means that outsiders like Net Applications can't tell which operating system a particular Web user is using. (Net Applications uses other methods, like a Web site's JavaScript to detect other information about a user and determine that the traffic is coming from Google.) About 11 per cent of Google's Web traffic currently shows up like this. The level fluctuates daily, Vizzacarro says. A few days ago it was around 30 per cent.

Traffic from Microsoft employees, in contrast, can be parsed in detail, leading some to wonder why Google is taking pains to cover its tracks. "It's not a natural process.

Google is the only company we've seen that does this," Vizzacarro says. Google employees not using the secret OS are employing various versions of Unix, such as Linux or Ubuntu, and some older operating systems*, he says.

Google declined to say whether it is planning an operating system. A spokesman noted that the company doesn't comment on rumors or speculation.

There are plenty of reasons why Google might be toiling on such a project. Two recently released products--browser Chrome and mobile operating system Android--have the potential to be spun into something broader. Chrome, combined with Google's open-source browser extension Gears, could be fashioned into a Web version of Windows. "Google could be gearing up for some sort of Chrome-centric, Linux-based OS," Vizzacarro says.

Android has officially debuted on just one product, the T-Mobile G1 cellphone. But fans and tinkerers have already ported it to devices ranging from Nokia's Internet tablet to a garage door opener. Google would like Android to be distributed even more widely.

Rich Miner, Google's group manager of mobile platforms, told Forbes.com earlier this year that the Android platform could power anything from a mini-computer to a set-top box to a portable navigation device. Google released the entire Android platform source code on the Web in October to encourage developers to do just that.

Net Applications could be detecting something other than a secret OS, of course. Its algorithms may be picking up on Google's robots that crawl and grab content from the Web, for instance. Or the veiled traffic could be a holdover from Google's two-year-long, super-secret incubation of Chrome. Since the "company of origin" feature is new, Net Applications can only trace data back three weeks in time.

Since Net Applications doesn't yet separate mobile traffic from desktop computer searches, the results could also stem from Google employees surfing the Web on Android phones. But Vizzacarro says that the first scenario would likely produce far more traffic than 11 per cent, while the latter two would probably produce less. "We know it's not their spider," he says. "Unless they have a group of engineers working on a project, it doesn't seem to add up."

Another explanation could be that Google is employing an OS that simply doesn't provide user agent detail. In that scenario, the missing information could be viewed as an optional field that was left blank.

Net Applications plans to release more detailed results to clients in a month. That could include companies like Hewlett-Packard, Dell, Adobe and Microsoft. Tech firms will likely want the data, Vizzacarro says, because knowing what companies like Google are using internally could help them sell to or partner with them.

Until then, the mystery of Google's invisibility cloak is likely to linger.

*Editor's note: An earlier version of this story included a reference to software that wasn't relevant.

Elizabeth Woyke, Forbes
Source: source image