LWN: Comments on "Fuchsia: a new operating system"

Fuchsia: a new operating system

walex — Wed, 10 May 2017 12:38:12 +0000

I understand that every new generation of researchers like to reinvent the wheel, but perhaps if Google had simply bought for cheap the rights to the Chorus OS, or MUSS, or BeOS, or TAOS, or MESHIX, or COYOTOS, or VSTa even?
In particular Chorus was a fully finished, debugged, working product.
There is an amazing amount of great research and products in the past decades that has become obscure if not totally forgotten.

Fuchsia: a new operating system

Wol — Tue, 11 Apr 2017 10:19:36 +0000

> Binary drivers are not going to go away. Qualcomm, Imagination Tech., MediaTek, Samsung all rely on closed binary drivers (and this will continue to do so due to patents).

Except it looks like the next big bun-fight over patents could kill them (the software version) completely!

One of the biggest pro-software-patent Supreme Court Judges seems to have come to the conclusion that software patents are illogical and cannot be justified. I don't think he quite got there in the last judgement, but it seems as though he is trying to think things through logically, and when dealing with software patents he's suddenly realised he's got his logic in a mobius twist.

Cheers,
Wol

Fuchsia: a new operating system

nix — Thu, 06 Apr 2017 00:48:12 +0000

Honestly, Unix has handles and something like capabilities: indeed, they're the same thing, and are the most fundamental thing about Unix.

We call them file descriptors and/or (for the instance of an fd owned by a specific process) file descriptions (a term that only POSIX could love). Those are precisely handles and also serve as unforgeable capabilities.

Fuchsia: a new operating system

aryonoco — Tue, 04 Apr 2017 02:40:46 +0000

Good luck enforcing that copyright infringement in court when most kernel developers don't think binary blobs are a GPL violation.

Fuchsia: a new operating system

pizza — Mon, 03 Apr 2017 21:44:27 +0000

> Binary drivers are not going to go away. Qualcomm, Imagination Tech., MediaTek, Samsung all rely on closed binary drivers (and this will continue to do so due to patents).

You forgot to add "despite this being a violation of the GPL"

Fuchsia: a new operating system

aryonoco — Mon, 03 Apr 2017 21:41:38 +0000

The information available points to a group of few dozen rockstar OS developers that Google has poached from Apple, Blackberry, Android and Chrome specifically to work on Fuchsia.

This is no one's hobby project. This is a long term bet by Google, but a very serious one, similar to RISC-V.

Fuchsia: a new operating system

aryonoco — Mon, 03 Apr 2017 21:38:11 +0000

You very well might, but Google, and the over a billion people whose Android phones are running outdated software with known root vulnerabilities due to the inability of chipset makers to update their drivers, would very much like to be able to use a system that can be properly updated, binary drivers or not.

Binary drivers are not going to go away. Qualcomm, Imagination Tech., MediaTek, Samsung all rely on closed binary drivers (and this will continue to do so due to patents). The question is not binary drivers vs open source mainline drivers, it is between binary drivers blobs in a kernel that cannot easily be updated, or binary blobs in a kernel that can be.

And Google, rightly in my opinion, is coming to the conclusion that the latter option is better.

Hence Fuchsia.

Fuchsia: a new operating system

kingdon — Mon, 03 Apr 2017 03:21:03 +0000

If I had to guess, it would think it is more 20% project than anything else at this point. But that's just speculation (and "Google doesn't know yet" is a perfectly valid answer).

Fuchsia: a new operating system

pabs — Mon, 03 Apr 2017 00:41:20 +0000

I'll take the current system over proprietary drivers any day of the week.

Fuchsia: a new operating system

aryonoco — Mon, 03 Apr 2017 00:16:55 +0000

Let's look at some of the people we know are working on Fuchsia:

Travis Geiselbrecht, one of the main developers of BeOS and NewOS. An original member of the the iPhone team and the OS for the Danger Hiptop/T-Mobile Sidekick.

Brian Swetland: Core developer of BeOS and Danger Hiptop One of the original developers of Android.

Chris McKillop, a member of the original iPhone team and the original WebOS team who also worked on the QNX. He also worked on the Danger Hiptop.

Adam Barth, a longtime member of Google's Chrome and Flutter teams. He also built an operating system for Raspberry Pi called Tau.

And probably many more that we don't yet know about.

These are not a bunch of interns; Google has hired/poached some of the OS developers in the world and put them to create Fuchsia. These are early days, but I expect to see a lot more from these guys in the next couple of years.

Fuchsia: a new operating system

yoshi314 — Sun, 02 Apr 2017 15:50:00 +0000

speculation is that they aim to provide a way for better hardware support by not having them rely on shifting abi of the kernel.

kernel maintenance for android (and availability of system updates) is pretty much a nightmare, with vendors on one end and kernel developers on the other.

Fuchsia: a new operating system

michaelgugino — Sat, 01 Apr 2017 03:04:34 +0000

Downloaded the build tools and compiled it. My first impression was not great compiling on an AMD system. Was able to get the OS to boot after hacking apart some of their bash scripts and changing a couple of the qemu options. Pulled up the terminal, I was unable to get much of anything to work following the instructions for some of bits that are already included. Launching an application seemed to immediately crash the system.

Ships with a slightly modified musl libc implementation, IIRC. That part is interesting.

I'm not a kernel or embedded C developer, but this looks like the work of a couple interns or students during summer of code. This might be a v1/POC of something that they are building behind the scenes, but it doesn't look like a serious effort from Google at this point.

Fuchsia: a new operating system

ortalo — Fri, 31 Mar 2017 10:22:14 +0000

Red vs. pink has always been a debate... hence the compromise I guess, with that green natural touch which is currently so trendy.
Anyway, at least it will change from cats: https://fr.wikipedia.org/wiki/Fuchsia#/media/File:Fleurs_...

As a side note, I like the idea with capabilities ; but I wonder how it will go up through the entire software task and reach the user.

Fuchsia: a new operating system

Cyberax — Thu, 30 Mar 2017 17:10:24 +0000

They didn't write it: http://www.drdobbs.com/windows/a-brief-history-of-windows...

Fuchsia: a new operating system

smitty_one_each — Thu, 30 Mar 2017 15:42:03 +0000

Offloading any unwanted architectures, legacy APIs, and associated personalities may be a motivator.

Fuchsia: a new operating system

k3a — Thu, 30 Mar 2017 13:46:43 +0000

There are enough OSes already... Google probably just needs something special for their application.

Fuchsia: a new operating system

dgm — Thu, 30 Mar 2017 09:53:07 +0000

Wrong choice of color?

Fuchsia: a new operating system

xav — Thu, 30 Mar 2017 09:04:34 +0000

My guess is on the license. Maybe one day you'll see a VM for android bytecode on fushia.

Fuchsia: a new operating system

Cyberax — Thu, 30 Mar 2017 06:38:46 +0000

All other OSes have a fatal flaw.

Fuchsia: a new operating system

roc — Thu, 30 Mar 2017 04:18:40 +0000

My big question is "what is the point of Fuschia?"

Is it better security? Then why use C++? Alternatively, why not start with seL4 or a derivative?

Is it the non-copyleft license?

Is it full employment for Google engineers?

Fuchsia: a new operating system

Fowl — Thu, 30 Mar 2017 03:58:21 +0000

Mmmm... Handles, objects, capabilities and C++. Sounds like that other famous "microkernel" everybody loves to hate ;p

Fuchsia: a new operating system

kjp — Thu, 30 Mar 2017 02:31:22 +0000

Looking at the wiki for magenta kernel - it's so spartan and clean. They're so proud of their handles and jobs and c++ objects and .. I don't know there's not a lot else. Now compare that to the average lwn kernel article - some megacorp wants to save 1% on battery/power usage, another wants to save 1% on TLB flushes, and yet another wants to shave 1% off of epoll wakeup scalability. All of which take a ton of hacky code, and don't really care about security.

I guess the saying is true - if a product still works, it hasn't added enough features yet. See Windows for the end game.