Product release!

[nathanjw]

The product I've been working on at Google since I joined up in March, ChromeOS, had its major public debut today:

http://www.google.com/chromeos

In addition to showing off the current version of the OS, and the integration with new toys like the Chrome Web Store, there's an actual piece of hardware, the "Cr-48" Chrome notebook. Unfortunately, it's in a "limited pilot program", so you can't all go out and buy one, but you can apply for the pilot and/or open new tabs in Chrome a lot and hope to get the golden ticket gift box.

(for what it's worth, I mostly worked on 802.11 authentication and roaming and related connectivity issues).

Comments

[jered]

Dumb question, but given that it's a linux PC how much 802.11 work was necessary? Is the existing stack pretty broken, or is there Chrome-specific functionality?

[nathanjw]

Some of both. There are quite a few things the existing stack doesn't do well, or didn't do well with our exact hardware (a relatively new Atheros chipset); changes there have generally been upstreamed into the kernel/mac80211 projects or wpa_supplicant (look for patch messages starting with the subject "CHROMIUMOS"). 802.11n support was kind of unreliable when we started as well. We're making heavy use of dbus under the hood, and that turned out to not be a well-tested corner of a few of the nominally-standard tools we're using (wpa_supplicant and dhcpcd, for example). Some things that are probably generally useful but not heavily banged-on elsewhere were things like speed from system resume to network availability, and a bunch of tuning of background-scanning frequency and techniques.

Other parts are local glue, particularly in our network manager software (a fork from the connman project), which was not a great piece of code in the first place, to cope with things like making 3g/wifi transitions automatically and without freaking out, and the necessary plumbing stacks for getting network information into Chrome's UI and letting it control us in return. I spent a bunch of time, mostly with existing pieces, to make 802.1x EAP-TLS work, and play nicely with the TPM-based local keychain we use. You can think of a bunch of this as the system design/integration work necessary to create a new distribution.

[jered]

Got it. Interesting!

I'll definitely be watching to see how this works out. I applied for the "beta hardware," although I don't know if I'm an ideal fit. I checked the "I will use this as my primary machine" honestly, but it will be interesting to see if that's actually possible. About 50% of my work life is in email, but the other 50% is in PowerPoint and I need to collaborate with others so it needs to be really PPT compatible, not "oh, I can mostly view this PPT someone sent me" compatible.

The other interesting area will be usability, fit and finish. This is where Apple traditionally wins. I was playing with jesse's rooted nook color this morning, and while the hardware feels nice the default Android fonts are awful. On the hardware side I'm sure the Cr-48 is very nice, but I worry that the commercial devices will be the "race to the bottom" that Windows/Linux netbooks are...

[ketzie.livejournal.com]

Did I even know you were at Google? In any event, congrats on the accomplishment!