[stdcli]

"are involved with Kubernetes, etc. None of that stuff probably is even close to working on Fuchsia and I don't think server side is a priority with that."

actually the networking stack in Fuschia, or more particularly zircon the kernel it is optimized around, is written in Go.

Kubernetes is written in go, which is essentially a system for implicitly defining scaled linux boxes on a giant network, and now in the most recent release is allowing for declarative architectures as well, much like how user space services for graphics drivers in Zircon will enable more declarative and optimized use of expensive resources as well (in this case, modular hardware architecture for mobile operating systems, which is what Fuschia was originally envisioned for.)

I'm also pretty sure Google also has experience with mobile operating systems (android), and Fuschia is literally a response of now over a decade of trying to interface android on POSIX in conjunction with multiple chains of monopolized and closed source hardware architectures, being intimately familiar with the evolution of trends in the mobile hardware architectural space, and trying to rewrite a kernel to optimize for that.

"Fuschia will need to play nice...Part of that is supporting linux based stuff; which especially on Chrome..." Chrome had to have an entire team dedicated to resandboxing their tabs to mitigate for spectre and meltdown, which are both results from the unquestioned but growing obscurity due to unquestioned implementations between hardware and Linux integration, which is something that Fuschia attempts to take a step back from, and by doing so simultaneously make it easier for open source development on hardware architectures while optimizing for them.

Still, total site isolation is still an advanced option in chrome that results in a 10% degradation in memory performance so people don't turn it on, if they even know to check for it/what that means (most people don't).

[jillesvangurp]

Crucially kubernetes uses docker, which uses a lot of low level linux and posix stuff. Kind of an issue if that stuff isn't there. Kernel virtualization on fuchsia could of course become a thing (like it is on Windows and OSX) but so far I've not seen much on that topic. WASM on chromium might become a way out as well of course.

But the bottom line is: Empty room. beautiful OS with an empty app store is kind of a non starter in the current market. Windows Phone found that out the hard way. Also, several other mobile operating systems that did not quite make it or continue to struggle. Sailfish, WebOS, Firefox Mobile, Ubuntu Mobile, etc. Even ChromeOS struggled until they added android support and Crostini.

[pjmlp]

ART will run on top of Fuchsia, it is already being ported, you can check the merge commits.

As we always need to remember Linux fans, the fact that Android uses a Linux kernel is irrelevant to userspace.

[jillesvangurp]

I have no idea what ART is but I doubt it addresses all concerns I listed above. On Android, there are plenty of native libraries and apps these days as well. I'm pretty sure these don't work as is without a compatibility layer that essentially replicates a lot of linux/posix stuff, which Fuchsia does not implement.

In any case, it wouldn't be the first time that Google walks away after putting lots of development in something. In general, I think they are closer to merging Android and Chrome OS than they are to replacing either with Fuchsia (not to mention convincing OEMs like Samsung to actually use it).

[pjmlp]

ART is the Android RunTime, basically the large majority of Android's userspace.

Those native libraries are forbidden to use Linux specific features, only these APIs are sancioned.

https://developer.android.com/ndk/guides/stable_apis

Since Android 7, Google has been claping down NDK users that try to used anything that isn't part of that list.

Since Android 8, APKs are only allowed to reach for their own internal filesystem and use SAF for anything else. Something that will be further enforced on Android Q, so no luck trying to peek into /dev, /share and similar.

[return_0e]

As for virtualization, Fuchsia already has its own KVM equivalent called Machina which so far can run Debian on top of Zircon and with several compatibility changes for supporting the ART runtime in Zircon already merged in, it should be also possible to run Android apps with this.

But perhaps the reason Fuchsia won't struggle unlike the other OSes you mention, is that it is possible that it will be compatible to run all the Android apps in the play store from day one, thanks to Machina; allowing a smoother transition, a process similar to what Apple did with the PowerPC to X86 switch but in Google's case, its for a completely different OS.

[stdcli]

there is also rkt, and anyways, while docker containers are great, they are just an abstraction of cgroups and namespaces, yet you forget that cgroups are a relatively recent concept in Linux and docker containers didn't even have namespaces in its first, second of third iteration, yet you act like docker relies on the immutable principles of posix.

Anyways, docker is a good example of how current linux systems are not optimized for modular sandboxing and containerziation. Still, people are so uneducated even in tech on how important this idea of only working with bare bones (I started in C so allocated bytes as I needed them and always considered how not to use them first, is a far cry from npming an express server and seeing the endless train of dependencies that are invoked) that still they do not secure their containers, and the number of ubuntu18.04 std base images I see running a docker container that simply contains a python app or something equally trivial, live in production at some of the top tech companies, which you can google and download a rootkit for, with no linux hardening whatsoever, is the terrifying norm of centralized web application companies today. I really am not going to buy into this idea that docker contiainers baring full replicas of the operating systems they sit on top of are a justification for POSIX.

If you want increased modularity for security, sandboxing and running different application, look at QubesOS, which is already far along and has it's own baremetal hypervisor, which is much like how docker works in userspace but optimized all the way to bare metal hardware. Fuschia takes a similiar approach when looking at optimizing modularity in mobile computing hardware architecture.

" Also, several other mobile operating systems that did not quite make it or continue to struggle."

This is true, but this is coming from the same company who has experience designing both software and mobile hardware architecture. Just because something is not already popular and widely adopted isn't a reason not to do it. I'm always an anti monopoly person myself, especially in the world of technology.

You can read my other comments and see the justifications around the need for this. As someone coming from the hardware architecture design space for qualcomm snapdragons all the way to 14nm iphone architecture, there is a need to remodularize kernel for advanced execution and increased competition in this space. POSIX is not sustainable looking 10-20-40 years into the future of hardware computing, particularly in the next ten, and android game developers who make a living off of candy crush do not really seem to care about this impending doom, only that they will have to traverse yet another learning curve if the platform gains adoption or become competitive in the space, which sucks, but it's not as bad as you'd think.

Besides, forcing people to continue to traverse learning curves keep the market competitive and keeps people from becoming to religiously entrenched that google's current android API is an unattested god. Yeh, its hard to make money sure, it's competitive sure, but we so easily forget android was a first stab and open source response to the iphone. The first motorola android phone came out my sophomore year in college. Now people can't imagine how we would survive without their 6yr olds going to school without an iphone 7. We often assume we need things to be the way they already are, and are not accustomed to change, but I can tell you Fuschia is needed in the space, operating system competition is needed in the space, and in the next decade we will be evolving to think different about excessive use of memory, dependencies and latency and be looking for something like this, luckily it will be about a decade into development at that point, about what android is now, and people say it's unreasonable to consider anything else..