Big layoffs at Google’s Fuchsia OS call the project’s future into question

Fuchsia lost 16 percent of its employees, while the rest of Google cut 6 percent.

See full article...

 

[jg67379]

Bit of a shame that little seems to be coming out of this project. When was the last time someone shipped a ground-up new OS to consumers? Windows NT? MacOS and Linux both have roots in, what, the 80s? Maybe earlier for the Mac because of FreeBSD. iOS is a fork of MacOS, and Android is a fork of Linux.

I know that OS vendors have been enhancing their core OS as they go along, but I have to wonder what an OS created for today's hardware, with the knowledge gained by the intervening years of computer science research, would look like. And who would bring it to us. There doesn't seem to be a lot of competitive pressure to reinvent the core OS these days.

Linux isn't actually based on code from BSD or Unix, it was just inspired by their design. It was written from scratch though.

 

[Scipio Africanus]

So I'll wait for Fuschia to be rebranded as a replacement for Chat/Meet now right? Or something else as random?

 

[ringofvoid]

I'm still impressed that the Fuschia team was able to deliver a drop-in replacement OS for Nest devices and the update be unnoticeable.

Edit: grammar

 

[Mckillio]

This sounds extremely plausible to me, given the timelines and the fact that nothing came of Fuchsia OS in the intervening (now) six years. It's a settled matter.

My thought NOW was how would they replace the entire Android ecosystem to work with a totally different OS? It appears the answer is, "They're not going to."

Frankly, that would be a huge undertaking, and for those invested in Android, a non-starter for going to a different OS where app variety would be exceptionally limited for quite a while (assuming, of course, that the Android programs aren't compatible with Fuchsia OS to begin with).

As others have mentioned, based on all of this, and these developments, I'd expect to see Fuchsia OS fade out of Google's public discourse in the fairly near future.

Fuchsia is currently on the Nest Hub and Hub Max and it should be coming to the Hub 2nd gen this year and a new smart speaker will be launching with it this year. So something has come of it.

Fuchsia can run Android apps and using Flutter devs can compile apps to run natively on it. So there wouldn't be limited apps. I would imagine that as far as the end user goes, they would get upgraded to Fuchsia without ever realizing it, just like on the Hubs.

At a minimum, Fuchsia seems like a lock to replace Cast OS on all supported devices.

 

[SeanJW]

It would be three different linux distros in one whose support would abruptly end and it would be replaced by YouTubeOS

Google already uses at least 6 Linux distros internally. Oracle Linux, RedHat Enterprise (in Corp), gLinux (based on Debian, for desktop and some Corp servers), the Prod setup that runs every service, and Android, and ChromeOS.

edit: The Oracle Linux/RH is because 3rd party software won't certify or support their products on random distros. Oracle charges less than RH for server support, and Google is very much capable of sending them a bug report with an attached patch saying "this is what's wrong. Fix it."

 

[solomonrex]

That OG PixelBook design is still amazing.

 

[stormcrash]

Quelle surprise. Fuchsia has always seemed to lack a concrete reason to exist beyond some annoyances with Linux and Google being jealous that they don't own their entire OS stack. All the security stuff they've been talking about is nice in theory but it's a concept in search of a problem. It's unclear what hardware would adopt it first and how complete it is towards that goal. The project has a serious risk of falling victim to HURD syndrome, an eternally incomplete OS core that exists for some philosophical reason while not managing to actually solve a use case better thant what's already there

Edit: Appears some of the nest devices are running on it so cool that it actually has shipped but that seems like a pretty niche use that doesn't really leverage all the talk of security focus

 

[stormcrash]

If you mean "major" OS, then that may well be, though odds are it started development roughly the same time as GNU Hurd, which is around 1990. There are also plenty of research OS projects, but if we're talking an OS intended for general purpose consumer use...

Also, NT was based largely on work Microsoft did for IBM on early versions of OS/2, so it's probably a bit older really. And both Sony and Nintendo's console OS are based on FreeBSD. Probably very heavily modified by now, but still FreeBSD is the origin point. The Xbox runs a custom version of Windows while we're at it. Mac OS X/macOS was/is also FreeBSD based. The GUI, however, adopted the OpenStep API early on and used a lot of design elements from NeXTStep. Haven't kept up to see if the OpenStep API has largely been supplanted by something else over the years or if it's still there. It was an ObjC API, and ObjC is the redheaded stepchild of the C-family. Very interesting language, but it tries to mash together two very different syntaxes that can be a little confusing.

From what I'd read David Cutler basically threw out all the existing OS/2 code and started from scratch when designing NT. Can't remember if that was before or after it was changed from being planned as OS/2 NT to Windows NT though. The NT kernel was mostly based on lessons and techniques Cutler brought with him from DEC

 

[DanNeely]

Microsoft made an aborted attempt at writing a kernel written in (extended) C#. Quite the fascinating thing. https://en.m.wikipedia.org/wiki/Singularity_(operating_system)

AFAIK MS was always open that Singularity was just a research project with no intention of being commercialized. They used it to push the boundaries of what could be done with the .net platform when not limited by back compatibility with older versions; but did end up backporting a lot of performance and async features back into the mainstream framework.

 

[stormcrash]

AFAIK MS was always open that Singularity was just a research project with no intention of being commercialized. They used it to push the boundaries of what could be done with the .net platform when not limited by back compatibility with older versions; but did end up backporting a lot of performance and async features back into the mainstream framework.

In other words Microsoft knows what an R&D project is actually supposed to look like. They didn't set out to make a new product, they kept the scope of features and time bound, and then they took the lessons learned to make work to improve the actual product

 

[CenterLess]

That's mostly correct but you need to remember that to keep the drivers closed source, they have to be userspace drivers which is an absolute no no in the Linux kernel mainline. Linux is a monolithic kernel, all drivers are compiled into the kernel at build time. The situation with android was that each device had a forked Linux kernel with all kinds of random modifications to it, not simply linking in a driver binary, so there was no nice boundary of 'Linux kernel' and 'Custom drivers' that you could try and mix and match, it was all a big mess.

Moreover, the end consumer isn't Qualcomms customer and they don't particularly care about you so long as OEMs keep putting their SoCs into new devices, so there's no major incentive for them to support things long term.

I don't think they link all the drivers into the kernel. Only the most common ones. Loadable drivers have been around since 2.4, but you still have to compile against the kernel version you're planning to load into.

 

[edrowland]

Quelle surprise. Fuchsia has always seemed to lack a concrete reason to exist beyond some annoyances with Linux and Google being jealous that they don't own their entire OS stack.

Concrete reasons for Fuchsia to exist: performance, security, smaller O/S footprint.

Security

Linux and Windows both trace their heritage to an age before security became a primary goal of operating systems. Microsoft spent almost a decade retroactively hardening Windows security. Linux did it piecemeal. Both continue to play whack-a-mole.

It's not unreasonable to think that a modern operating system written with a modern eye toward security might do better, even if it hasn't yet been hardened by decades of bitter experience.

Performance

One of the more radical design choices in Fuchsia is to make kernel calls asynchronous by default. The concrete reason for that is that Windows and LINUX have both shown that asynchronous I/O yields significant performance benefits at scale. Windows and LINUX currently contain multiple attempts to retroactively bolt on asynchronous I/O after the fact. Both have significant quirks in their implementation because of that. Fuchsia , on the other hand, does asynchronous I/O by design, starting at the kernel layer (while bolting on a synchronous xNIX emulation layer for use by legacy apps).

The concrete reason to expect gains in performance: message-queue-based notification allows multiple events to be processed without a kernel transition, and avoid the necessity of polling for completion.

It's too early to tell whether that asynchronous I/O performance increment will materialize, and how significant it will be, but it's clearly a primary design goal of the operating system.

O/S Footprint

A primary goal for Fuchsia: to run on small devices.

 

[Signal Chaser 76]

Article takes me back to the heady days of installing BeOS 4.5 and then 5.0 on my beige desktop in oh 2000 or so, marveling at the spinning teacup demo, and then evangelizing BeOS on here and elsewhere, only then to suffer the sadness as Be slowly went under, then! a life raft....Be would be reborn as BeIA, the BeOS Internet Appliance! Of course. Hardware OEM partnership with the Sony eVilla, a portrait display internet appliance, the iMac killer, things are really looking up.

Alas, 3 months later, eVilla is dead.

And so, once bitten twice shy, I've learned never to get excited about new operating systems. They trigger something deep and fundamental in me, something elemental and exciting, like that feeling of opening a box floppies for a new OS from Babbages as a kid!

As the apostle said, when I was young, I was OS-curious, now that I am old, "It Just Works" is the only prize I seek

 

[Easyenough]

The devices running this Fuschia OS, Google/nest speakers and screens, are the least responsive devices I've encountered in 20 years. My Nest hub 2 with sleep sensing can take 10+ seconds to respond to a voice input and several seconds to respond to a touch input. My many Google mini's, pre-update, had barely perceptible lags.

Other variables could be responsible, like using 2002 chips in a 2018 device, but it's surprising to have hub 1 be very responsive on launch and hub 2 to not be responsive at all after the OS update.

Also, in terms of new OSs, Palm OS AKA PalmSourse AKA Garnet OS was 1996 and seems to have been de Novo..

 

[phoenix_rizzen]

From what I'd read David Cutler basically threw out all the existing OS/2 code and started from scratch when designing NT. Can't remember if that was before or after it was changed from being planned as OS/2 NT to Windows NT though. The NT kernel was mostly based on lessons and techniques Cutler brought with him from DEC

[Edit: correction, it was to be OS/2 2.0. Thanks SeanJW.]

NT started out as OS/2 2.0 during development. The original GUI was to be the OS/2 Presentation Manager. After Windows 3.x popularity took off, they switched the GUI to mimic Windows. (At this point, Microsoft split development away from IBM, and IBM had to scramble to start development of a new OS/2 2.0)

They designed the NT kernel to support multiple userland personalities (DOS, Win16, Win32, OS/2) so all they had to do was change the default to Win32 Program Manager before NT 3.x shipped. They also designed it to work on multiple CPU architectures (the HAL was one of the nicest and most ornery features).

NT 3.x and 4.x could all run OS/2 1.x programs.

[Edit: fixed last paragraph with correct versions.]

 

[jhodge]

Linux and Windows both trace their heritage to an age before security became a primary goal of operating systems. Microsoft spent almost a decade retroactively hardening Windows security. Linux did it piecemeal. Both continue to play whack-a-mole.

That's not correct. DOD's Trusted Systems Evaluation Criteria (TCSEC Orange Book) was published in the early 80s, and VMS, which at the very least inspired a lot of Windows NT had at least one version certified at the B1 level. Microsoft may not have maintained best practices with development of the Windows NT family, but security was certainly not a foreign concept during it's development. IIRC, MS even touted a particular WinNT setup as being C2 certified at one point.

 

[darkowl]

Maybe the real problem was that people had trouble spelling Fuchsia correctly?

(The correct vs incorrect spelling of fuchsia was one of the amusing tidbits from XKCD's colour survey, circa 12 and a half years ago: https://blog.xkcd.com/2010/05/03/color-survey-results/ , the writeup is quite interesting and worth a read if you're interested in how we consider colour naming)

 

[bbanditx79]

Hopefully this doesn't spell the end of Fuchsia and is just Google/Alphabet using this as a way to cut some "excess" staff now that Fuchsia is a little more mature.

 

[SeanJW]

NT started out as OS/2 3.0 during development. The original GUI was to be the OS/2 Presentation Manager. After Windows 3.x popularity took off, they switched the GUI to mimic Windows. (At this point, Microsoft split development away from IBM, and IBM had to scramble to start development of a new OS/2 3.0 aka Warp.)

They designed the NT kernel to support multiple userland personalities (DOS, Win16, Win32, OS/2) so all they had to do was change the default to Win32 Program Manager before NT 3.x shipped. They also designed it to work on multiple CPU architectures (the HAL was one of the nicest and most ornery features).

NT 3.x, 4.x, and I think 2K could all run OS/2 1.x and 2.x text and GUI programs. Not sure if they supported OS/2 3.x programs.

They could run OS/2 1.x 16 bit software. OS/2 2.0 was the IBM side of the divorce and Microsoft had nothing to do with it.

And it was killed after NT4; its biggest use was running Sybase that Microsoft licensed to develop MSSQL (both current Sybase and MSSQL are descendants of the old code base)

POSIX lasted longer than OS/2 did in Windows.

Edit: I used that ability to run OS/2 programs to run Lotus Domino on Citrix WinFrame (a licensed fork of Windows NT 3.5 at the time)… because I had a WinFrame test server just sitting there and I wanted to see what the fuss about Notes was….

 

[phoenix_rizzen]

They could run OS/2 1.x 16 bit software. OS/2 2.0 was the IBM side of the divorce and Microsoft had nothing to do with it.

Ah, you're right, it was supposed to be OS/2 2.0.

And it was killed after NT4; its biggest use was running Sybase that Microsoft licensed to develop MSSQL (both current Sybase and MSSQL are descendants of the old code base)

POSIX lasted longer than OS/2 did in Windows.

Forgot about the POSIX personality.

 

[TOPS20-RTE-MPE-Linux-Vipe]

Scalable RTOS for low-cost consumer devices? If I were looking at the Fuschia cost-center I would be asking why is this being done in-house when a open-source RTOS like Zephyr is being shepherded by the Linux Foundation and many users/contributors?

 

[xWidget]

Scalable RTOS for low-cost consumer devices? If I were looking at the Fuschia cost-center I would be asking why is this being done in-house when a open-source RTOS like Zephyr is being shepherded by the Linux Foundation and many users/contributors?

Isn't Zephyr just like a fancy Arduino set of libraries for more hardware? It's been awhile since I properly looked at it but I don't think it's an OS in the same sense that something like Linux is, with multiple users, loadable drivers, etc. You just compile it all into your program and flash it.

IIRC it has a kernel in the sense that you can have multiple tasks running that the kernel manages, but that's about it.

 

[Mckillio]

The devices running this Fuschia OS, Google/nest speakers and screens, are the least responsive devices I've encountered in 20 years. My Nest hub 2 with sleep sensing can take 10+ seconds to respond to a voice input and several seconds to respond to a touch input. My many Google mini's, pre-update, had barely perceptible lags.

Other variables could be responsible, like using 2002 chips in a 2018 device, but it's surprising to have hub 1 be very responsive on launch and hub 2 to not be responsive at all after the OS update.

Also, in terms of new OSs, Palm OS AKA PalmSourse AKA Garnet OS was 1996 and seems to have been de Novo..

Neither the Hub 2 nor the speakers run Fuchsia.

 

[thadec]

Scalable RTOS for low-cost consumer devices? If I were looking at the Fuschia cost-center I would be asking why is this being done in-house when a open-source RTOS like Zephyr is being shepherded by the Linux Foundation and many users/contributors?

More to the point: there is no way for Google to monetize a low-cost consumer device other than their own. It is up to the Fuschia team to come up with their own product to justify their ongoing existence. Otherwise Google is going to mothball it. Google isn't doing much with Dart, Carbon or Flutter these days because the teams that developed them never made a product or case for using them.

 

[thadec]

Neither the Hub 2 nor the speakers run Fuchsia.

The next Nest speakers will
https://chromeunboxed.com/nest-speaker-third-gen-fuchsia-out-of-the-boxas will the speaker that will be the Pixel Tablet's dock. Even if the Hub 2 or the existing speakers never get updated to Fuchsia - I have read where they will be eventually - CastOS is dead and all new speakers and smart displays will have Fuchsia. That is, presuming that Google continues to make speakers and smart displays, which cannot be presumed at this point.

 

[Mckillio]

Google went through a phase where they allowed the western tech media to make them ashamed of Android (which the Chrome team at Google took advantage of to have them push ChromeOS on tablets and other large screen devices instead of Android, and Chromecast instead of Android TV). Years later after basically NONE of the big cloud initiatives panning out - Google Assistant, Google Cloud platform, Chromecast, Stadia etc. either petered out or flopped big time from the start - they are back to Android which - despite its detractors (all Apple and Microsoft ecosystem users anyway) - mostly just does what it is supposed to do as well as it needs to on every platform. And oh yeah, has over 1.5 billion daily active users and brings Google a ton of revenue.

So, Fuchsia's potential as an Android replacement is dead. It also can't really replace ChromeOS except for low end browser-only devices used in education and as kiosks anymore because the Linux functionality keeps expanding (KVM and Parallels for enterprise users, Crostini LInux and now Steam). That means that the Fuchsia team needs to create its own product to justify its existence. It is neat that it replaced "CastOS", but Google has silently stopped building smart speakers into their routers, hasn't done much with their Nest speakers since releasing the Mini in 2019 and Audio in 2020 (Amazon really hasn't done much with Echo either) and rumor has it that the next Pixel Tablet will replace/compete with at least the Max version of the Nest Hub (though it is said that the dock will still run Fuchsia).

The Android-based IoT platforms - Android Brillo and its replacement Android Things - both failed mainly due to lack of support. A lot of the interesting things with Android that Google was doing when Sergey Brin and Larry Page were running the ship were abandoned by Sundar Pichai, who preferred the cloud/cast/Chrome team. So maybe Fuchsia can take its place as an IoT platform, where a microkernel can be of use as it takes up less storage and CPU resources than even a Linux distro. (Note: examples of microkernels include the RTOS used in a lot of very cheap fitness traackers). But if that is all that they are going to do with it, then it doesn't require a large team. But - again - create a product to justify their existence like - like the team that Pichar favored admittedly did with Chromecast and Chromebooks and obviously as the Android team did ... to the point where Pixel was once a ChromeOS brand and now Android phones and tablets have usurped it - and they can get their headcount back.

There's a lot wrong in your post but the Nest Audio came out in 2020 and their routers never had a speaker, just the Point and it's still for sale.

 

[CajunMoses]

Sounds like the author desperately wants Google to promote and market Fuchsia. Meanwhile, it seems like Google is intentionally incorporating Fuchsia in products in such a way that the consumer is totally shielded from directly interacting with the OS. So the consumer can be and is oblivious to the underlying OS. As long as that remains the case, there is no point to promoting and marketing Fuchsia. Google's approach to Fuchsia may be that the less the consumer is aware of the OS, the better.

 

[aerogems]

Linux isn't actually based on code from BSD or Unix, it was just inspired by their design. It was written from scratch though.

You could call this semantics, but Linus' aim with Linux was basically to create a Unix for 386 CPUs. Linux has always adhered to POSIX and aimed to be as Unix-like as possible. Take FreeDOS or ReactOS. Technically they're novel implementations, but they're trying to emulate another OS as closely as possible. So, IMO, that means they may not be a direct descendant exactly, but it doesn't really make them "from scratch" either since they're emulating something else.

 

[aerogems]

As the apostle said, when I was young, I was OS-curious, now that I am old, "It Just Works" is the only prize I seek

Man does that ever resonate. There was a time when I was using Gentoo Linux on my personal desktop and I'd go in and tweak the build scripts sometimes as well. This was also still in the dark days when you would frequently have to manually tweak your XF86Config file and DRI was a brand new thing that only kinda worked if you stood on your head while looking in a mirror, drinking earl gray tea from a vintage Disney little mermaid teacup, on the night of a full moon, as the clock struck 8 of 12 at midnight. Kids these days don't know how much of an accomplishment it used to be to get Linux working at all, but I digress. I enjoyed the challenge in my younger days when free time abounded.

These days, I have less free time, so I don't feel like spending it fucking around with whatever random thing may go wrong with Linux, or dealing with software that is forever in "almost there" status. If I have a couple free hours on the weekend, I want to be able to just fire up Steam, load a game, play for as much of that couple hours as I can, and then rinse and repeat the next day I have some free time. Just give me something that works. Windows ain't perfect, but it's closer than Linux will probably ever be in my lifetime.

 

[flunk]

I think Fuschia is a backup plan in case Google gets sued by a bunch of open-source developers and they can't continue to distribute Android or Chrome OS. That doesn't mean the design isn't neat with the Flutter-based UI, Flutter itself being cross-platform so writing an app for Fushia means you've also written an app for Android and iOS. The opportunity for better integration between parts and a ground-up kernel designed for low-power devices and the ability to run the same OS across everything from internet of things devices to desktops.

 

[jwbaker]

Sounds like the author desperately wants Google to promote and market Fuchsia. Meanwhile, it seems like Google is intentionally incorporating Fuchsia in products in such a way that the consumer is totally shielded from directly interacting with the OS. So the consumer can be and is oblivious to the underlying OS. As long as that remains the case, there is no point to promoting and marketing Fuchsia. Google's approach to Fuchsia may be that the less the consumer is aware of the OS, the better.

Indeed, and the article also ignores, as the author long has, the possibility that Google wants Fuchsia as the OS for its internal datacenter hardware, whether running on the main hosts or embedded in the roots of trust, smart NICs, etc. This site has always had a massive blind spot for reasons why Google would want a non-Linux operating system.

 

[davidlars]

Weeell, there's always TempleOS if you want to have good laugh.

Have a look at Serenity OS while you're at it. A complete, mostly POSIX compliant OS (kernel+userland, including a web browser from scratch!). Originally built by one person (Andreas Kling) who has vlogged his process on YouTube. Last I checked ~1500 people had contributed to the project (all for the fun of it). Fun fact, Linus Groh, one the contributors who work mostly on Ladybird the web browser for the OS (and home made dependencies like LibJS) recently got invited to TC39, the group that develops the JavaScript standard.

Update: and yes I'm impressed by how much can be done by a bunch of motivated hobbyists. Linux also started by a motivated hobbyist. But I'm not equating SerenityOS with Fuchsia or any other OS project by for profit companies. The priorities are obviously very different. I wouldn't expect security to be a high priority for example. The journey to whatever this could become in the future is more exciting to me than most of whatever shows up in the tech news cycle.

 

[Oliver Weichhold]

Google has become the Microsoft at the end of the previous Millennium ... slow, bloated, evil and incompetent. Only resting on their previous merit .. search.

 

[macosandlinux]

So now they have 336 people on it?
That still seems like a team large enough to keep the work going, whatever Google decides that work to be.

The only problem is they have 3 teams of 100 people each working on 3 competing messenger apps for Fuchsia, so the OS has just 36 left to go /s

 

[Verisimilitude]

Does this OS have a foreseeable fuschia?

 

[Panashi]

By these lay-offs by the major tech companies, are they really planning to do something bigger or better?