Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

As usual I get quite hopeful when I hear about a new operating system. We so absolutely need paths to move away from what we have today.

I didn't even know what "mixed reality operating system" was so that made me hopeful. May include some nifty quantum processing.

Alas, it is just stuff built on top of Android, which is built on top of the Linux kernel.

Perhaps "Meta Horizon SDK for Android" would be a better label?`



Android is already quite cool compared with what traditional OSes happen to be.

For starters it is what Microsoft has failed to do with Longhorn, because contrary to having internal sabotage from the kernel team, the Android team doubled down on their efforts to make a managed userspace actually work in a mainstream product.

Dalvik sucked balls against Sony and Nokia J2ME implementations, with its DEX interpreter, yet they managed to create ART with quite cool mix of JIT/AOT/PGO, and a very modern GC.

Then it follows up, that for all the praise Plan 9 gets on hacker circles, Android's approach to user space closely follows on Inferno with Limbo, Plan 9's successor that many of its praisers tend to forget.

Followed by being able to encapsulate Linux kernel into a micro-kernel like stack, where most drivers are process based using Android IPC to talk to the kernel, and can be written in a mix of Java, C++, and now Rust.

If only they handn't screw Sun, and stepped away when the ship went down, maybe they could have saved themselves some headaches, however Google isn't really a language designer company, so maybe that was for the greater good nontheless.

Ah, and using AIDL is so much better developer experience than dealing with the primitive COM tooling on Windows.


Limbo and Inferno are partially alive in Go and plan9/9front. On J2ME, it worked well in legay devices; but for modern usage, Android's Java-like API was a godsend for the developers, as they could seamlessly transition from one Java-like lang to another one. Heck, J2ME interpreters for Android do exist, and I think lots of them just use shims...


> May include some nifty quantum processing.

While we're at it they might do a bit of dabbling in reversing entropy and solving the halting problem - it's only impossible in flat space-time after all.


I was amused and interested when I found out that HaikuOS is not Linux, nor even the Linux Kernel, but a reimplementation of BeOS (concepts mostly) on the NewOS kernel.

I've been dwelling in some nostalgia recently and miss some of the earlier days when everything wasn't so homogenous. This is all rose-colored glasses though, as I don't miss when everything was also extremely incompatible. And slow. And expensive.


> I've been dwelling in some nostalgia recently and miss some of the earlier days when everything wasn't so homogenous.

You are not alone in this. I see announcements like this one from Meta and others like Oxide - companies out there having a swing in a different direction and it makes me glad that there are people out there thinking outside the monoculture that computing feels like it's become sometimes.

Sure, standardisation has been good in some respects, but I feel like the incentive to really innovate (as opposed to incrementally improve) in the operating system and compute space has really dried up.


> Perhaps "Meta Horizon SDK for Android" would be a better label?`

Yes, the same as Android SDK for Linux. But it became just Android.


What about "Meta Horizon SDK for Android SDK for Linux?"


Let's just cut to the chase and call it the HDK.


> We so absolutely need paths to move away from what we have today.

Why though? What necessitates our move away from what we have?


Diversity = freedom


Nope


"Nope" doesn't really contribute to the discussion. Why does diversity not equal freedom?


>Alas, it is just stuff built on top of Android, which is built on top of the Linux kernel.

Look into HarmonyOS, Fuchsia, LionsOS, Managarm, Haiku and Genode.


And Qubes OS.


Qubes relies on Linux, huge TCB.

Refer to Makatea[0] for Qubes reimagined on a less flawed architecture.

0. https://trustworthy.systems/projects/makatea/


It doesn't rely on Linux, it relies on Xen and hardware virtualization: https://www.qubes-os.org/faq/#why-does-qubes-use-xen-instead...


What unexplored space do you think is ripe for a totally new operating system?


Ironically, general purpose computing. Like servers, smartphones, etc. The premise of Fuchsia, as a prominent example, is to build an OS that operates more like a sandboxed/capabilities-based microservices platform, with structured IPC across processes. All major platforms both cloud and clients (even browsers) have gone to great lengths to deliver precisely that on top of existing OSs, with many expensive layers and hacks. It would be a lot nicer to have it built into the OS.

Does that mean it’s truly ripe? I don’t know.


Dumb, probably overly broad Q: I'm looking to build a hardware project off a flutter app. Linux on Raspberry Pi is the straightforward option. But I want an excuse to build on Fuchsia.

Let's pretend Fuchsia is just as easy to deploy on (I'm 99% sure it won't be).

Does Fuchsia buy me any cool stuff long-term that isn't possible on Linux?

Like, what's my excuse for going with Fuchsia other than "(handwaves) it'll be more secure"? (to be clear, im teasing myself and my understanding of Fuchsia, not your explanation)


> Let's pretend Fuchsia is just as easy to deploy on (I'm 99% sure it won't be).

To stress this beyond a doubt: don’t do this unless you’re willing to become an early-masochistic-martyr adopter. My knowledge is outdated by years though. I don’t know how far you’d get today.

> Does Fuchsia buy me any cool stuff long-term that isn't possible on Linux?

I don’t think so. Fuchsia shines with capability-based multitenancy, and basically anywhere you would have different processes from different vendors communicating together (like a mobile OS with multiple partially trusted apps that may need to talk to one another). If you own/audit all code on your device, especially in a single app, you don’t gain anything from that.

Some auxiliary stuff like content addressable file system and OTA deployments may be attractive, but I have no idea if those things are actually supported or even around these days.

Oh, one thing that may be worth tracking in your domain is how fuchsia is progressing in terms of power management. If you’re on arm and battery power, and they prioritize it, it may beat Linux & android in terms of low power devices. But that’s pure speculation.


Unikernels. Throw away the OS completely, and just run the server application on the hardware/hypervisor. That would be my answer for servers. You kinda don't need an OS for them.


For what it’s worth, I am doing this with FreeBSD right now … with jail.


Dreaming of a world where normal PCs could have an OS on the quality level of MacOS with the software ecosystem but Linux is just too fragmented that you can't build anything coherent or reliable on an interface level in the way MacOS is.

This isn't saying Linux is bad, its great at what it does and being modular but how can something like MacOS spell check where it works on any text field coherently across any app work in a modular world like that.

Windows just feels like bad decisions made decades ago hold it back, it can't even open a folder of 20 thumbnails without choking, MacOS can handle 10,000 like a knife through butter. Can't even find a file opened yesterday when you search for it in the start menu by exact name, MacOS manages in a fraction of a second. Maybe the right team could rewrite all the things causing this jank but you'd essentially have to replace so much of the company that caused it anyway that it doesn't seem feasible.

Still use all 3 for different tasks but MacOS is the only one that feels like an operating system should feel in 2024.


A space where operating systems exist to service the operator and not the OS vendor.

One that isn’t just a vehicle to push ads and subscriptions to <blank>-as-a-service or otherwise act as a siphon to send telemetry up to the mothership.

Yes, I know *nix exists but it’s so fractured and every little variant has its own quirks making it difficult to be a general purpose OS that the masses could adopt, IMO.


> Yes, I know *nix exists but it’s so fractured and every little variant has its own quirks making it difficult to be a general purpose OS that the masses could adopt, IMO.

Yes, but to be clear: the main hurdle for adoption is fragmentation across GUIs/installs/package management across distros, which is out of scope (for better or worse) from the POV of Linux. Linux is a technological marvel, and if “they” could sort out these issues it would be the shortest path to mainstream appeal, by far. I’m certain this is technically feasible but also extremely challenging to pull off from a leadership perspective. Unfortunately, I don’t think there is a solution the majority would embrace. There is a ton of flame wars to overcome, hills people will die on.


It seems like if you created a new operating system to solve these problems and it gained some traction, you'd fracture the landscape even more.

Unless your operating system happened to be superior to all the existing solutions in all aspects with no tradeoffs.

Relevant xkcd: https://xkcd.com/927/


“Operating system” is being used to refer to the UI for running and managing apps (not the original under-the-hood sense of the word for how app share resources, manage security, etc.). For better or worse, that is how consumers understand it.


And you seriously expect an ad company to improve anything? What's Google missing?


What we have today is Absolutely Awesome. Even Windows.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: