What's an OS for? Thanks to Rupert Murdoch, Google has spent most of the time it probably wanted to be talking about Chrome discussing the value of paid content on the Web; but that's actually an awful lot of what Chrome is really for. It's a big fancy browser on custom-built hardware, which is another answer to the question I asked Google's VP of engineering (and ex-Microsoft general manager in charge of evangelising Windows as a platform) Vic Gundotra at the Google IO developer conference this year: when Google does its own OS, how will it deal with drivers and peripherals? Do what Apple does and lock down the hardware so you have hardly any drivers to worry about and stick a Web server in everything else so you can use inefficient but ubiquitous Web protocols instead.
At the time he told me the question was at least 14 years old (which makes me think we really should have a good answer by now), quoted Marc Andreesson in his Netscape days, saying he thought Windows would become nothing more than a set of poorly debugged device drivers and then challenged the assumption that "peripherals will always require native device drivers". He has a digital photo frame that he logs into on the Web, and one of the smart power meters Google is working on (power consumption is indexable data that Google can aggregate and extract value from just like Web sites); "even with peripherals," he told me, "their connectivity to the Internet is where their value derives." I'm not certain about that; with the speed of our DSL, I'd rather have my photos go onto a photo frame from our network drive than from Flickr - but it would certainly be easier to read our electric meter over the Internet while we're away. So does that mean you can wave aside the driver question after 14 years? Only if you want your devices to cost more.
You have to have smarts at one end of the connection; if they're not in the OS, they have to be in the peripheral. You can't have thin client at both ends. Every peripheral that you can't treat as external storage (cameras, media players) is going to have to have an embedded OS of its own to run the Web server for the driver-less OS to talk to. Smarter devices can go online and sync to the driver-less OS as peers - perhaps through a Google service? The rest of the ecosystem can do a little more work, and Google can do a lot less.
Talking of who's doing the work, I've seen some speculation about whether hardware OEMs will want to make low-end Chrome notebooks. (Google calls them 'slightly bigger than a netbook'; so just call it a notebook! They may have SSDs instead of hard drives; so does the ThinkPad we're toting and it's a notebook.) Certainly, the success of netbooks took the shine off Dell's profits this year and PC makers have been quick to push 10 and 11" netbooks with graphics acceleration like Ion and usable keyboards instead of 7" margin killers. With no premium hardware, would getting the OS for free be enough to tempt them back after the way Linux netbooks came flooding back as returns? How about less than free?
Remember that while Chrome OS will be open source, it will be open source in the same way Android is; the apps like Google Maps, with its turn-by-turn directions aren't. Google controls those and makes enough money from them that it can offer a share of ad revenue to manufacturers; if HP made a Chrome OS device, they'd make money on all your searches. Venture capitalist Bill Gurley nailed this when he christened it "less than free". To compete, other OS manufacturers would have to not only drop their licence fee but pay the hardware manufacturer. The followup question is how much of that a company with a seven to one share of the search market (compared to Bing) can do without incurring an anti-trust investigation.