Sundar Pichai, senior vice president of Chrome

SAN FRANCISCO--Google expects Chrome OS to be a success. But it's chosen its terms for success very carefully.

Google shares with many of its rivals a natural, reasonable ambition to measure success by market penetration. This week at the Google I/O conference here, the company was quick to tout that there have been 100 million activations of Android devices, that 310 different Android devices have gone on sale so far, and that Android users have downloaded 4.5 billion apps to date.

Though data-obsessed Google doubtless will count how many Chromebooks are sold, that isn't the measurement at the top of the priority list, Sundar Pichai, senior vice president of Chrome, said in an interview.

"Our goal, our main criterion, is [that] I want really high user satisfaction amongst consumers, businesses, and schools, independent of our quantity. I want people who pick up and buy one to be very happy with their purchase," Pichai said.

Recognizing that a Chrome OS laptop isn't for everybody, though, the company is restricting sales to online channels only, where the people most likely to buy a Chromebook are those who are actively looking for it.

"Part of the reason it's not in physical retail is our goal is not to push a lot of these," Pichai said. "We want people to know what they're buying. Online gives a check. In the physical world you might accidentally walk out with a Chromebook. I don't want that to happen."

The strategy makes sense, given that Chrome OS is such a departure from existing computing technology. Apple is the company that likes to talk about the "post-PC era," but iPads and iPhones are, architecturally speaking, very similar to PCs. A touch-screen interface brings a direct physical mode of interaction, but fundamentally it's still an operating system running software on its processor and storing data on its storage system.

Chrome OS, on the other hand, is inextricably linked to the Internet. Although a traditional operating system--an embedded version of Linux--is under the covers, the applications on the system run within its Chrome browser. They're Web applications, using Web languages like JavaScript and Web interfaces like AppCache to store data, and WebGL to show hardware-accelerated 3D graphics. Though properly written applications will be able to run while a Chrome OS laptop is disconnected from the network, cloud computing is mandatory.

That's probably fine for a lot of people. You can look up a recipe, check in with your Facebook friends, answer your e-mail, enter customer information into a Web form, order something from Amazon, watch YouTube videos, and plan a budget in a Google Docs spreadsheet.

One big problem, though, is what you can't do: run Microsoft Office, play Portal 2, make a photo book in iPhoto. Or, perhaps more to the point for people considering a supplement to the PC they probably already have in their homes, play the wealth of games on an iPad.

But Google plans to start small and grow. Pichai thinks Chrome OS will appeal to a lot of people, but evidently recognizing that it will take a lot of time to win over most folks, the company is aiming initial products at the enthusiasts who are predisposed to like it. From that seed, Google expects a tree to grow.

"You build a great experience, and you continually improve it. A few people get on board. As long as you delight them, they serve as messengers. Then somebody else hears about it, it breaks out, you market it," Pichai said. "You have to earn it step by step."

It's a strategy that worked well for Android and Chrome. Android launched with a single phone, on a single carrier, in a single market. Chrome launched in beta on Windows only, missing many features. Both grew, and Google improved them steadily.

"What's important is the pace at which you make progress. This is why we decided to shorten the [Chrome] release cycle to six weeks," he said.

It's not clear, though, how universally the strategy works. Apple's first iPhone began from modest beginnings--it had no support for 3G networks and no ability to run third-party applications, for starts--and grew into a tremendous success. Gmail, too, began with a small group of enthusiastic early adopters.

But sometimes an early kernel of fans isn't enough, even with Google's fast-iteration ethos. Google Buzz failed to catch on widely, and Google Wave was largely scrapped.

And, while Pichai asserts that the Cr-48, Google's developer-oriented Chrome OS notebook prototype, was well received, it left reviewers underwhelmed when it initially arrived.

Pichai blamed expectations that were too high for a system Google said was not done, but that people judged as a finished product nonetheless. "We knew things were broken there. People got very upset about trackpads. Perhaps people weren't ready for beta hardware," Pichai said.

A large part of the Chrome OS sales pitch is that, unlike people's PC experience, a Chromebook will get better with time as Google constantly upgrades the operating system--silently, in the background, with no user intervention. Even those initially weak trackpads in the Cr-48 got better with new software, he said.

"If tomorrow Brian makes WebGL great, suddenly Angry Birds works faster. Say you had a Cr-48 for six months. You just open your computer, and things just work faster," Pichai said. "We keep updating Chrome. All the GPU benefits working their way though Chrome make Chrome OS faster. We're going to offload more stuff to the GPU."

Google has high hopes for another technology called Native Client that, if the company can convince programmers to adopt it, could endow Web applications with the speed of native ones while not compromising security. Importantly, Google plans to build Chrome itself on a Native Client foundation.

Linus Upson, vice president of engineering for the Chrome team, believes Google will start that rebuilding process later this year, beginning with one piece of Chrome, the built-in PDF reader. Pichai was more cautious.

"Linus forgets not everyone codes as fast as he does," he said. "We have to make sure Native Client proves its way. After that we make sure we get Chrome running inside it."

But it is on the list, in part because Google is powerfully interested in Chrome security. And Chrome OS comes with a verified boot process, an encrypted file system by default, and plug-ins that run in a restricted sandbox, Pichai points out.

"Down the line, when we talk about Chrome running inside Native Client--a double or triple sandbox--that's what gets us excited. That's the kind of project that gets the best engineers in the world," Pichai said, positively glowing at the prospect.

Google's fast-changing Chrome can cause heartburn for some Web developers who already must constantly test their sites with an expanding number of browsers. Brian Rakowski, director of product management, said Google tests new versions of the browser to keep incompatibilities from encroaching.

"We're careful with compatibility issues," Rakowski said. "The [Chrome] dev channel is good feedback for what's broken, at least in bigger sites," and Google adds new tests if it finds a site that stopped working so that incompatibility won't go unnoticed again.

In any event, Pichai clearly won't let up on the pace of Chrome change. The mission is to rebuild personal computing with the Internet deeply integrated, not patched on at a higher level.

"The benefits from a security standpoint, getting new APIs out, and just pushing the platform," outweigh the problems of keeping up with Chrome. "If you don't do this, I think the Web will fall behind the native platforms pretty quickly."