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.
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.
> 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.
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.
> (...) 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.
Likewise I only started to see Linux really replacing UNIX deployments, around mid-2003, at CERN, transitioning away from Solaris to Scientific Linux (collaboratively designed with Fermilab), and a couple of years later, 2006, Nokia started to transition away from HP-UX to Red-Hat Linux.
Other than that, it was something we dual booted at home by not being able to afford big UNIX.
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.
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.
This is pretty disappointing, but the writing was already on the wall for any desktop-ish version of Fuchsia. I do hope it somehow still replaces the Android core someday, maybe just running the Android Chrome via Starnix.
It's a bummer because Fuchsia takes everything we've learned from automatic software updates, devices that run untrusted and semi-trusted code, proprietary drivers in a OSS kernel, devices whose manufactures aren't compelled to update them, the unsuitability of the user-based security model, containers, and more and applied it to a new kernel and OS.
Fuchsia is more or less exactly the kind of trust-nothing OS architecture I want to run as an end user and as a developer who has to run a lot of random 3rd-party code locally.
It's been clear for a while that Google was de-emphasizing the importance of Fuchsia. Chrome dropping support kind of rules it out as a viable OS for any consumer device. Next up is probably a proper shutdown of the team. Aside from some Nest thingies, did anything else ever launch with Fuchsia? Does it have any relevance for any server side stuff that Google does internally?
My guess is that what happened is that they failed to get OEMs enthusiastic about Fuchsia. At least I assume the likes of Samsung would not be super excited about giving away control of even more of their stack to Google. And I don't think it's a recent thing either; this likely became pretty clear years ago and the project has been slowly dying ever since.
Which it probably would have been, Google regrets giving so much control to OEMs with android and they won't do that again with new development. They locked down wear os a lot more too. But with android it's hard to take back what they've given already. Meaning the OEMs will not be as happy to take fuchsia on board. Not for its technical merits but for the new start under new rules it will imply. The only way for Google to force them would be to discontinue Android which would be very risky. People could fork it, it would put their marketshare at risk of competitors and it would take years to get the app gap with ios closed again as they'd be starting from zero.
As a privacy-conscious user I'd worry more about whether the licence permits something like lineage to exist based on fuchsia. I'm sure it's more secure but more under the control of Google too who I'd put into the adversary camp. I wouldn't use it until it was first cleaned of spyware by projects like graphene or lineage. I'm not sure if the license permits this.
Given the history of layoffs, I am afraid that one day soon we’ll be reading about Fuchsia being dropped. It was always ambitious but it’s not the kind of project modern Google seems capable of committing to.
The IBM of today isn’t what I think the other commenter was referring to.
In the 1970’s IBM was a behemoth. Yet IBM of that time was unable to translate that success meaningfully to personal PCs. In the realm of the personal PC it was Intel and Microsoft that dominated.
Similarly the Microsoft of the 2000’s was a behemoth that failed to translate that success to mobile computing.
I believe the other poster was trying to say that Google of today is like IBM of the 70’s or Microsoft of the 2000’s.
That being said IBM is different today than it was even a couple decades ago and Microsoft has changed a lot in recent years so maybe Google could become innovative again.
Just because they don't sell consumer products doesn't mean that their products aren't marketable.
In the case of research and semiconductor technology, typically IBM isn't out to deliver a finished product, they work with other companies who bring the end result to market. E.g., there's a joint venture between Intel and IBM to advance Intel's fabrication capabilities.
When you think of IBM as more of an R&D consulting company or custom enterprise application company it starts to make more sense. To some extent they also have an off-the-shelf product lineup, but most of the reasons to choose them involve custom solutions.
I understand that. I’m just pointing out that IBM isn’t some fly by night failure of a company just because they don’t make brand name products that every tech enthusiast and practitioner is using.
Their R&D is basically structured like a consulting gig in itself. They help people like Nvidia and Intel deliver on their products.
There are some extremely popular tech stacks like Docker and Hashicorp led by companies that are struggling to be profitable. The whole reason those technologies are popular is because they are free or cheap. Then as soon as the company starts to tighten the screws and monetize their user base, the user base crumbles.
IBM’s Disney contract is about double the size of Hashicorp’s annual revenue.
I don't think startup founders dream of those metrics. But IBM has a healthy profit as well. Still it is an empty shell of its former self, selling (among many venerated B2B products) snake oil like Watson and design thinking. Beyond the husk of its brand, it is a generic, irrelevant, boring IT company.
Do not delude yourself - IBM is surviving off of grifting big legacy IBM Technology and public sector contracts through their Consulting arm (...which they also bought, off of PwC) and near-/off-shoring.
Most "Quantum Ambassadors" and similar are funded thanks to benevolence/nepotism of local management since they do not create billable hours or maintenance contracts. Adding to that, IBM's rolling out suffocating utilization targets for their employees in more and more countries which will take care of that as well.
The weird thing about Fuchsia is that it's been public for at least eight years. In that time Google has gone through several messaging apps - like, launched them, got bored with them, launched a successor, and killed the original, multiple times over. All for an OS they only shipped on one weird voice-assistant-with-a-screen appliance.
In general, yes, but I was thinking specifically about how hard that division was hit last year. This kind of thing can become self-fulfilling and since it’s not tightly coupled to revenue I’d be worried if I worked there, which is never good for productivity.
I feel like the AI debacle with Google having no good answer to ChatGPT was only the tip of the iceberg and it's becoming clear that the company does not have a good leadership structure in place.
Some blame Sundar, personally I have no idea if it's true but it's sad to see a company that seemed so bleeding edge slowly decay like that.
We’ll see how it compares but also a single technical success does not make a successful business, and it doesn’t replace a strategy.
OpenAI has a ton of interest right now but is also losing a lot of money annually, and a lot of AI products are hitting pain points around reliability and security which are going to require significant advances to overcome. Google appears to be flailing around trying to find something which will work, and given how much of this technology they invented that is really a strong argument for better management.
By now it is clear that Fuchsia will eventually follow Brillo and Android Things footsteps.
It is a shame, Fuchsia has a lot of cool ideas, I guess by having Binder, Android IPC for Treble drivers with a microkernel like approach, and Rust integration, kind of keeps some of Fuchsia's ideas on Android, leaving little else for Fuchsia, other than replacing the Linux kernel.
Yeah, it's really sad. I feel like Elon Musk is fighting that downward trajectory towards mediocrity but he's punished daily for it so it must be hard to resists. I guess when you get as big as Google you get tangled in economic and governmental obligations. It's a real Letterman-Chappelle situation.
They have a laundry list of anti-trust investigations lined up against Apple for "anything Apple do better than Google" so it appears that they will be resting on their browser, search, video, and mapping monopolies which are funded by their ad duopoly, and using the US government to force competitors to hand their products over to Google.
I really don't know how business works at that high of a level. I'm not sure how much of it is for show and make believe at a federal level and how much of it is an actual dirty fight. I've never found a coherent book on the subject.
What is a "staff retention project"? Companies don't generally spin up 400-employee divisions just so they can keep hold of employees.
This was clearly intended to eventually replace Android and ChromeOS. The layoffs are likely because someone high up freaked out about how many people were working on a very very long term strategic project.
I would guess dropping native Chrome is because it's a ton of work and they don't need it, at least not for a long time. They can use Flutter for embedded devices, and if they support Android they can just run Chrome for Android.
> Companies don't generally spin up 400-employee divisions just so they can keep hold of employees.
Back in ZIRP days, Google did. I was there and saw it firsthand. Were you? I know the initial team that started this. Very smart folks. But there was never a shred of future there. When they tried to convince me to join the team, there was no answer to my question of "who would use this and why?". Android compat would never happen without a vm or container (in which case all the fancy privilege control is useless since all the sensitive data is INSIDE the container that MUST act linux-like and not fancy-priviledge-separation-like). And nobody would rewrite all the apps of the world to fit the new paradigm, as seen by chrome dropping this target finally, and by the failure of much-better-sponsored windows phone OS. Given that even the team had no clear answer to what they were doing and why, yes, the conclusion is that they were permitted to do this just to keep them off market to competitors, and available should they be needed elsewhere at Google. This also explains why there were layoffs in the team once the free money ended, and it expains this article here too.
I was at Google then too. I had the same questions when I was asked to work on the team or even just collaborate with the team on Chrome support.
I think it would have been fine to just throw the Android runtime in a container. That wouldn't be useless - it was still a great opportunity to reinvent the kernel and the Linux driver model, and all of the other parts of the system.
I find the "staff retention" thing confusing. Sure, I understand talented people want some freedom and flexibility, but I don't think that having a team of 400+ people go crazy for years with no plan is the only way to retain. Why not let them go crazy for a while but then force them to downscope, pick a real need and solve it with their tech?
Or, why not make it more incubator-style? Instead of a team of 400, have 10 teams of 40 and give them a year to build something amazing, then turn only the top 1 or 2 into products and send the others back to the drawing board?
You worked at google. Would you say that serious top-down leadership with a vision existed in the age of Sundar? I spent time in ChromeOS and Android teams. Almost 8 years of time. The actual teams had some vision and plans, but there was NOTHING from above. So yes, it is quite believable to me. Imagine: fuchsia team wanted to fuck around for funsies. Nobody above stopped it. It is so google 2012+
> Android compat would never happen without a vm or container (in which case all the fancy privilege control is useless since all the sensitive data is INSIDE the container that MUST act linux-like and not fancy-priviledge-separation-like).
VM-per-app has very real possibilities. You'd need some new logic for things like keyboard apps, but ...
> Replacing Android and Chrome OS would have made so much sense, but they were definitely NOT trying to do that. I never understood why.
Are you sure? In my experience people often say that project X isn't intended to replace project Y when it absolutely is, in order to avoid having to fight the people working on project Y.
An open source example of this is WASM. When it first appeared there were countless "wasm isn't intended to replace JavaScript" claims. Of course it 100% was (for a number of use cases anyway), but people had to say that to placate the JS fanatics.
It looks like Google potted ART to Fuchsia at one point.
Can't speak for management, but they could've seen it as funding open-ended research. Give a bunch of nerds freedom from shipping dates, deadlines, adoption plans, customer demands, and other such things that they always complain about, then see what they come up with.
It's expected that this kind of strategy will yield nothing most of the time, but eventually you get a positive Black Swan.
I don't think having a great liberally-licenced microkernel-based OS is as good for the world in a practical sense compared to forcing every Android and Chromebook vendor to publish the changes they made to a copyleft monolithic kernel. I can totally understand being worried that the rise of Fuchsia would result in a dark age for running alternative open source OSes on cheap consumer smartphones, SBCs and ARM64 laptops.
Some do. If the current state of things is what happens when vendors are required to publish their source by law, imagine how much worse it would be if they weren't - there'd be nothing. What little we have of FOSS OSes on cheap consumer devices would be gone. Imagine no LineageOS, no community-supported postmarketOS devices, no installing mainline Linux on old Chromebooks etc.
This is true in some abstract sense but the value really goes up dramatically if more than a few people actually use it. There’s a strong feedback loop around having drivers or being able to run common code, and that’s why many people were hoping that Google had a strategy since there aren’t that many projects with the resources to push through the early iterations of those problems.
The article keeps talking about the Chrome Browser on Fuschia being discontinued & the end of Workstation.
> At first, the shift from workstation to workbench seems like a simple name change, but the subsequent discontinuation of Chrome for Fuchsia all but confirms that Google does not intend to release a desktop/mobile device directly powered by Fuchsia in the foreseeable future.
But then it turns around and says what seems to be: oh, but the web view will still be maintained!
> Importantly, this move has no implications for the Nest Hub lineup, which will continue to receive updates and support web-based features as normal.
Giving up on Chrome as an app, a full browser with user experience - that's one thing. But if the web engine is still alive & maintained, that feels like a fairly reversible choice. It's hard to imagine Fuschia persisting without a web engine, but that doesn't seem like what the pile of inferences in this article is saying!
Pathetic. The last remaining gasp of any chance at anything innovative ever coming from Google again, over a decade of work, being tossed out to juice some quarterly numbers.
And this is coming from someone who cheered for Google until I guess 2012, then gave them up but defended them until maybe 2017 - 2019 somewhere and now just hope they are never recovers.
Let them go the way of IBM: continue to exist without threatening general purpose computing.
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.