4 ms·

I worked at Google (on the Chrome team) at the time Fuchsia was started and for many years after that. The team had some of Google's most brilliant engineers w

I worked at Google (on the Chrome team) at the time Fuchsia was started and for many years after that. The team had some of Google's most brilliant engineers working on it. Technically, they had some really impressive stuff. However, I don't think anyone was ever able to articulate to me its path to success. In other words - let's say you build an amazing, powerful new operating system. How do you actually get anyone to adopt it? What frustrated me is that there wasn't even an answer for Android or Chrome OS. Why build a new operating system that wouldn't even be relevant or useful for those, the two platforms that Google had total 100% control over? If I had been in charge of the world, I would have taken a more pragmatic approach. Chrome OS was the obvious first target - it was built on Linux but it didn't have to be. The operating system really only has a tiny number of native binaries that need to run on it: * Chrome * The Chrome OS GUI (which shares a lot of code with Chrome but is distinct) * The Android VM * A Linux VM If you port those four to run on another operating system, then you're done. You could start shipping Chrome OS devices based on that new OS. I think it also would have been reasonable to have a goal to port Android to run on it, basically replace the Linux kernel. Probably it would have been necessary to add some Linux compatibility layer to enable Linux drivers for some number of years, with the goal of eventually shipping devices that don't require Linux. But the Fuchsia team wanted to build a complete operating system from scratch, including a whole new GUI, a whole new developer SDK, and more. Why??? When they were finally forced to ship something useful, they shipped the Nest Hub, which in my opinion was a big mistake. The Nest Hub based on Fuchsia does absolutely nothing that the previous one didn't do. It basically took a relatively simple, easy product and increased its maintenance burden by 100x while providing no value. If they had chosen Chrome OS instead it could have actually provided some value in terms of performance, security, modularity, and developer APIs. It just seems like such a waste of talent to try to reinvent everything with no plan for adoption, rather than building more incrementally and actually shipping.

spankalee 3y ago
I think they may have thought that if they built something great that "management" would make the other teams adopt Fuchsia at the core somehow. Of course other teams just develop a defensive stance and guard against that politically and, as Android has done, by adopting their ideas. The later is good actually, since incremental changes are so more likely to land and stick.
- snitch182 3y ago
  If you build it they will come
jonathanyc 3y ago
Thanks for sharing your honest thoughts. One question I have though—you say the Nest Home on Fuchsia didn’t have any new features enabled by Fuchsia. What specific features enabled by Fuchsia could have been used in Chrome OS? As I understand it almost all of the isolation there is provided by applications running inside of the browser sandbox, which is provided by Chromium.
jokoon 3y ago
I think the goal might be to have more control on android devices to make more money. Although I don't really see why they can't do that with the linux kernel, maybe the linux kernel sort of prevents them from collecting more royalties. But I agree that making a new kernel from scratch is not really a very good idea, unless it brings big improvements, which I doubt.
csdreamer7 3y ago
> If they had chosen Chrome OS instead it could have actually provided some value in terms of performance, security, modularity, and developer APIs. I have not worked at Google. My assumption was that they wanted to avoid the GPL and avoid Linux's strick userspace backwards compatibility to get things upstreamed.
- squarefoot 3y ago
  The GPL is what forced initially hardware companies to publish drivers source code, helping Linux to run on a huge number of architectures and platforms. Its viral nature, although may have some drawbacks in some contexts, to me has been crucial in helping Linux to jump from a hobby project to the critical mass of users that later forced businesses to take a serious look at it. I still hear the words of a (quite incompetent and unpleasant) PM at a place I worked at in 2000-2001; we were having some problems (glitches on network traffic triggered by a GFX card interrupt) with Tru64 on Alpha machines and since we had a lot of hardware at hand we installed Linux (RedHat if memory serves) for comparison tests and the problem disappeared, so we jokingly told "well, then just ask the client to install Linux, they'll also save a lot in licenses since it's free". His reply: "Linux? Nah, that's an hobby project, the customer uses professional software". Two years later the PM had been sacked and the customer was starting to migrate everything to Linux. Too bad I already left the year before, but that's how life goes.
  - foofie 3y ago
    > (...) His reply: "Linux? Nah, that's an hobby project, the customer uses professional software". Two years later the PM had been sacked and the customer was starting to migrate everything to Linux. (...) I recall using Linux around 2000. Calling it "hobby project" might be derogatory, but it was definitely very rough on the edges. For perspective, the infamous SCO lawsuits, which to me marks the moment Linux was officially deemed a competitor and a threat by established companies in the industry, only took off nearly a decade after this date. Back in 2000 I wouldn't have risked my career by suggesting my company's product line should depend on it to run. Also, a lot changes in two years, specially in IT circles. I also recall back in the early 2000s people rushed to install new development versions of distro releases because those would finally add critical feature/fix an egregious bug even at the expense of introducing regressions. Knoppix was only released late 2000, which goes to show how untested this domain was back then.
6510 3y ago
I could never afford it of course but when I thought about a new OS for maybe 30 min I could quite easily think of a few dozen things that sounded exciting and would make it worth the effort. The goals chosen by popular OS's are very good but there is plenty of fruit slightly further up the tree. For alphabet there are many more options.
hulitu 3y ago
> But the Fuchsia team wanted to build a complete operating system from scratch, including a whole new GUI, a whole new developer SDK, and more. Why??? CADT ? It is, since 20 years, a trend in SW development: Delete and start over.

spankalee 3y ago

I think they may have thought that if they built something great that "management" would make the other teams adopt Fuchsia at the core somehow. Of course other teams just develop a defensive stance and guard against that politically and, as Android has done, by adopting their ideas. The later is good actually, since incremental changes are so more likely to land and stick.

snitch182 3y ago
If you build it they will come

jonathanyc 3y ago

Thanks for sharing your honest thoughts. One question I have though—you say the Nest Home on Fuchsia didn’t have any new features enabled by Fuchsia. What specific features enabled by Fuchsia could have been used in Chrome OS? As I understand it almost all of the isolation there is provided by applications running inside of the browser sandbox, which is provided by Chromium.

jokoon 3y ago

I think the goal might be to have more control on android devices to make more money. Although I don't really see why they can't do that with the linux kernel, maybe the linux kernel sort of prevents them from collecting more royalties. But I agree that making a new kernel from scratch is not really a very good idea, unless it brings big improvements, which I doubt.

csdreamer7 3y ago

> If they had chosen Chrome OS instead it could have actually provided some value in terms of performance, security, modularity, and developer APIs. I have not worked at Google. My assumption was that they wanted to avoid the GPL and avoid Linux's strick userspace backwards compatibility to get things upstreamed.

squarefoot 3y ago
The GPL is what forced initially hardware companies to publish drivers source code, helping Linux to run on a huge number of architectures and platforms. Its viral nature, although may have some drawbacks in some contexts, to me has been crucial in helping Linux to jump from a hobby project to the critical mass of users that later forced businesses to take a serious look at it. I still hear the words of a (quite incompetent and unpleasant) PM at a place I worked at in 2000-2001; we were having some problems (glitches on network traffic triggered by a GFX card interrupt) with Tru64 on Alpha machines and since we had a lot of hardware at hand we installed Linux (RedHat if memory serves) for comparison tests and the problem disappeared, so we jokingly told "well, then just ask the client to install Linux, they'll also save a lot in licenses since it's free". His reply: "Linux? Nah, that's an hobby project, the customer uses professional software". Two years later the PM had been sacked and the customer was starting to migrate everything to Linux. Too bad I already left the year before, but that's how life goes.
- foofie 3y ago
  > (...) His reply: "Linux? Nah, that's an hobby project, the customer uses professional software". Two years later the PM had been sacked and the customer was starting to migrate everything to Linux. (...) I recall using Linux around 2000. Calling it "hobby project" might be derogatory, but it was definitely very rough on the edges. For perspective, the infamous SCO lawsuits, which to me marks the moment Linux was officially deemed a competitor and a threat by established companies in the industry, only took off nearly a decade after this date. Back in 2000 I wouldn't have risked my career by suggesting my company's product line should depend on it to run. Also, a lot changes in two years, specially in IT circles. I also recall back in the early 2000s people rushed to install new development versions of distro releases because those would finally add critical feature/fix an egregious bug even at the expense of introducing regressions. Knoppix was only released late 2000, which goes to show how untested this domain was back then.

6510 3y ago

I could never afford it of course but when I thought about a new OS for maybe 30 min I could quite easily think of a few dozen things that sounded exciting and would make it worth the effort. The goals chosen by popular OS's are very good but there is plenty of fruit slightly further up the tree. For alphabet there are many more options.

hulitu 3y ago

> But the Fuchsia team wanted to build a complete operating system from scratch, including a whole new GUI, a whole new developer SDK, and more. Why??? CADT ? It is, since 20 years, a trend in SW development: Delete and start over.