Are micokernels superior?

Are micokernels superior?

Other urls found in this thread:

opensource.org/osd-annotated
youtu.be/jiGjp7JHiYs
youtu.be/wTVfAMRj-7E
lwn.net/Articles/738975/
blog.darknedgy.net/technology/2016/01/01/0/
seL4.systems
harmful.cat-v.org/software/andy_tanenbaum
wiki.sel4.systems/Rust)
twitter.com/NSFWRedditImage

Intel thinks so

Yes.

What's the difference?

Fuck off, Tanenbaum. No one wants your shitty kernel. I hate you so much that I ran ME Cleaner just to remove your dumb kernel module from the Intel ME partition. Kill yourself, you smug big nosed fuck.

Microkernels are much better at security and stability but they are also harder to implement and end up being a bit slower than monolithic kernels.

Yeah, that's why they're only used in embedded shit

mac os is a microkernel :^

Academic's answer:
>Yes
Actual answer:
>Fuck no

If it's not the why intel has it on every new CPU? :^)

QNX uses a microkernel and it's actually pretty good.

No it's not

Yep. If only we had better hardware when the world was picking between HURD and Linux.

The lack of * in major OSes is due to inertia more than anything.

Or if MINIX had been open source back then.

It hasn't been good since Blackberry took over and that's been like a fucking decade at this point

afaik this thing doesnt even have 64-bit or USB drivers on x86.

Also a lot of it nowadays is taken from NetBSD so just use that

But the fact that they take stuff from NetBSD doesn't really matter as it all ends up running in userspace anyway.

>HURD is so shitty you need stonger hardware for it to barely run

It was open source (after all it is the referece OS for Tannenbaum's OS textbook). It just wasn't free software because it had some custom education-only license.

That's not open source then.

>Open-source software (OSS) is computer software with its source code made available with a license in which the copyright holder provides the rights to study, change, and distribute the software to anyone and for any purpose.

and yet every Intel ME works on it without Tannenbaum even knowing that, lol

MINIX con this year was cancelled because not enough people submitted papers.
Literally dead.

You're talking about MINIX 3 which Tannenbaum released under the BSD license.

>not knowing the difference between open source and free

That definition is Free®™ software. Open source just means the source is available alongside binaries.
E.g. game dev kits are open source but you can't reuse that code in an unrelated project.

>>HURD is so shitty you need stonger hardware for it to barely run
I'm talking about the overhead of actually running a microkernel (constant switcihng between kernel and user space). Nowadays this wouldn't be much but back then it was just too much.

HURD despite being le freetard product would actually solve literally every problem Linux was plagued by for years because of it's monolithic architecture.

I know the difference, but clearly you don't know.

>Open-source software (OSS) is computer software with its source code made available with a license in which the copyright holder provides the rights to study, change, and distribute the software to anyone and for any purpose

That's not open source, just look at the fucking definition of open source.

>Open-source software (OSS) is computer software with its source code made available with a license in which the copyright holder provides the rights to study, change, and distribute the software to anyone and for any purpose

>(constant switcihng between kernel and user space). Nowadays this wouldn't be much but back then it was just too much.
entering and leaving kernel mode is still expensive. This is why Linux maps the kernel address space into every process's address space, to avoid a ~10-15% performance penalty from switching.

>Nowadays this wouldn't be much but back then it was just too much.
Maybe not for home use, but no serious enterprise would use it to run their applications and databases because the overhead gets very noticeable once you scale up.

>HURD despite being le freetard product would actually solve literally every problem Linux was plagued by for years because of it's monolithic architecture.
And those problems are...? Unless you list them it's an empty statement.

> Imma keep posting that definition until it becomes correct
It still isn't.
If I make some source code available but licence it such that you can't redistribute it, is it now closed source?

Well duh, wikipedia can't be wrong

It's proprietary, not open source.

opensource.org/osd-annotated

Not the user you're replying to, but that would make it "shared source". It's somewhat used in enterprise and UNIX used to be licensed this way for decades.

Are there any up to date benchmarks about this?

HURD intrigues me. It's clearly a dead project and there's basically no hope at this point of it ever going anywhere, but It's one of those things that I really never knew much about and it would be kinda cool if it ever did become usable
What would it be like? What problems would it solve assuming that it became actually usable?

>And those problems are...? Unless you list them it's an empty statement.
Do you have an entire afternoon to spare?
youtu.be/jiGjp7JHiYs

youtu.be/wTVfAMRj-7E

Even Google is moving away to a microkernel with Fuchsia for Android

no we're brainlets, please summarize.

I'll watch it but that's not what I wanted.
Can you summarize a few points?

I really want someone to pull an "Ungoogled-Chromium" type thing with this. It's "Open Source" but you just fucking know this thing crawling with botnet.

That's OK because smartphone are low-power semi-embedded devices. Hardly anyone cares about performance and even current smartphone hardware is overkill.

>wants a "summary" of 6hrs of technical details
go back to management Rajesh

Its a hybrid kernel.

lwn.net/Articles/738975/
This isn't quite the same thing, it's about switching page tables instead about actually switching to kernel mode. (switching the page tables would mean the kernel doesn't need to be in process address space, which is stronger isolation, and makes ASLR much more effective, among other things) But doing that gives a performance regression ranging from 5-30%, depending on how many system calls a process makes.

There's another article I remember about the actual direct issue (the time it takes to switch from user mode to kernel mode and back) but I can't find it right now. The gist of it was that that latency of that is still a significant performance concern that they try to minimize.

Hybrid kernels are superior. Some things don't make sense in userspace, some are cumberstone
in kernel. Put it where it makes sense.

I thought that was more because they wanted something that isn't GPL.

hybrid kernel is a microsoft marketing term with no substance whatsoever
stop believing in memes
also please define what are these "some things"

>lwn.net/Articles/738975/

thanks

Remind me why people use GPU instead of BSD?

I feel like GPL exists need to make more open software, but to make less closed one like it's some sort of Zero sum game.

Does Wind River Systems use a Microkernel?
Does SCO use a microkernel?
Does IBM use a microkernel?
etc, etc.

>Remind me why people use GPU instead of BSD?
because BSD can't play gayms

>I feel like GPL exists need to make more open software, but to make less closed one like it's some sort of Zero sum game.
stop phoneposting

>microkernels are slow
:^D
blog.darknedgy.net/technology/2016/01/01/0/

They do wonders for security in embedded, at least.
See: Sony PlayStation Vita, Nintendo 3DS and Nintendo Switch

I can't post from PC.

>3ds
My hacked 3ds says otherwise.

here's a nickel kid, buy a better computer

None of which use a microkernel.

The key thing is that BSD doesn't require someone who forks the software to use the same license, while GPL does. Because forks of GPL stuff are forced to also use the GPL, they aren't allowed to make it proprietary aka botnet.
BSD meanwhile doesn't care what license you fork it under, meaning someone could create a proprietary botnet version of a BSD licensed program.

BSDfags claim BSD is "moar free" because it lets people do whatever the fuck they want.
GPLfags claim GPL is "moar free" because it ensures nobody can take away your four essential freedoms to run, study, redistribute, and distribute modified versions.

Take it to the license wars thread. This is the microkernel thread.

I wish this user would summarize that thing about HURD or some other microkernel solving "literally every problem Linux was plagued by for years"

I personally agree with BSDfags.
But GPL is so much more popular for some reason.

I blame RMS. He is kinda like Steve Jobs of Linux world. Really good at making people think he made something. (See GNU + Linux).

Okay.

>2017
>morons keep talking about a failed design based on a 1st gen microkernel
Newsflash: 1993, Liedtke, L4.

>I feel like GPL exists need to make more open software, but to make less closed one like it's some sort of Zero sum game.
That's the whole idea. It's why critics call it a virus.
When you make something great (Loonix) and licence it copyleft, people who want to use and modify it also have to make their derivitives copyleft.
So in a sense, GPL isn't truly free. BSD-type licences are more free from a libertarian perspective as you can do virtually anything with them. But GPL and other copyleft licences encourage more copyleft software to exist, whereas BSD, MIT and friends have no such in-build mechanism.

Sorry we're getting in the way of your 300 quality microkernel posts.

>I blame RMS. He is kinda like Steve Jobs of Linux world. Really good at making people think he made something. (See GNU + Linux).
Except he really DID make something. Emacs, GCC, etc.

>Thinking microkernel refers to the size of the device

This.

Redox is literally our only hope for a secure OS: microkernel and written in a safe language. My dream computer system is Redox running on a POWER9 Talos system, with fully free firmware on the whole hardware stack.

The whole point of microkernels is security. Privileged drivers are just a bad idea, and reality has confirmed this (how many bugs in Linux, Android, iOS are due to drivers? Have you read Project Zero's writeups about Broadcom vulnerabilities?). And it's not just academic masturbation: security critical systems use microkernels (seL4). So yes, microkernels are superior even if there is a performance tradeoff.

>Thinking microkernel refers to the size of the kernel
Literally #1 in

Well, that's fair, except it has nothing to do with Linux.
That said, I like the virus analogy of .

The only thing I know about his software is that it's bloated, non-standard and if you implement something based on it, you will have troubles using anything else (bashisms, inability to compile kernel with clang and so on).

Some distributions run with virtually no GNU software.

Why is it so popular then?

People who make the software have nothing to gain from using GPL instead of BSD.

The situation with libreboot developer seems like a good deterrent from using anything RMS related.

>Redox is literally our only hope for a secure OS: microkernel and written in a safe language.
"safe language", sure, but seL4.systems is proven, and written by the best people in the field.

>GPL and other copyleft licences encourage more copyleft software to exist
This is why I prefer GPL. I understand that BSD makes more sense from a libertarian perspective, but I dislike the botnet so much that I personally don't give a shit.

Also, I thought we agreed to save this for another thread?

>it's bloated, non-standard and if you implement something based on it, you will have troubles using anything else
wat?

Chromium is also written by the best people in the field and has its source routinely fuzzed with cutting-edge tools, and it still has tons of bugs. It is simply not feasible to write safe software in unsafe languages and we should just give up on that assumption as soon as possible, for our own sake.

>People who make the software have nothing to gain from using GPL instead of BSD.
If someone decides to fork their software, they get to use the forked version if it's better, or merge whatever improvements have been made if it isn't. This is guaranteed with GPL(assuming whoever forks it doesn't break the license terms, at least). With BSD there's no such guarantee. So if he actually plans on using his own software, I would say that's at least a potential gain.

That's a good argument, thanks.
Perhaps something that allows to sell as closed but disclose any modifications?

Different needs for different projects.
If you want your project to be adopted everywhere, including commercial sphere, you want to use permissive licenses. ssh is an example of this and even RMS agrees it is ok in such cases.

>It is simply not feasible to write safe software in unsafe languages
Explain seL4, then.

>Different needs for different projects.
Precisely. But if you're a nobody slapping together something for the sake of convenience, there's no reason to use BSD license over GPL, while GPL has at least potential benefits that could help you directly.

It's also written in C which nobody wants to write nowadays. Even if you can write provably correct code it's still a horrible language.
The reference operating system is just that, a reference. It's not something you can use.

Meanwhile redox os is a somewhat usable operating system with a complete graphics stack and it's very close to being self-hosted.

>It's also written in C which nobody wants to write nowadays
Rust soybois unironically believe this.
>Meanwhile redox os is a somewhat usable operating system
Unironically comparing a popular microkernel that's being used in the wild in high reliability, fault-tolerant applcations with a toy os, and favoring the latter.

>it's still a horrible language.
why?

Weak type system. Requires external tools that impose restrictions on what you can do in the code to prove memory safety.

>Andy Tanenbaum hasn’t learned anything
> - Microkernels are the way to go
> False unless your only goal is to get papers published.
> Plan 9's kernel is a fraction of the size of any microkernel
> we know and offers more functionality and comparable
> or often better performance.

harmful.cat-v.org/software/andy_tanenbaum

AFAIK, seL4 is verified under a certain set of "assumptions" (ARM version under a certain configuration), excludes several portions of code (like hand-written assembly), virtual memory, etc. and verification still requires some manual actions every time it's updated. I mean, seL4 is certainly a very remarkable feat and is almost certainly mostly bug-free compared to other similar-sized C projects, but it would probably be saner and more secure to outright use a safe language (and it seems they are exploring that: wiki.sel4.systems/Rust) rather than have an entire parallel complex infrastructure just to check your unsafe-language code doesn't do funny things.

Everyone I've ever met who likes raccoons was a greasy smelly autist.

Are you a greasy smelly autist OP?

...

>rather than have an entire parallel complex infrastructure just to check your unsafe-language code doesn't do funny things.
Uh, you misunderstand both the scope of the verification (it's far wider than you seem to think) and what it verifies (it's nothing to do with the language, but whether seL4 does what it is supposed to do).

>2017
>an argument against tanenbaum in 1992 is used to against microkernels
>L4 didn't even exist yet.
tanenbaum=microkernels=bad!!!11

>But GPL is so much more popular for some reason.
Maybe GPL is popular because people have a greater incentive to contribute back?

Redox is based on sel4

is it? couldn't find anything that confirms this

The author of Redox made that statement during a live interview. Of course, Redox is 100% Rust but it borrows its design from sel4

So it's not.

I use the linux kernel because I hate hybrids

The lead Redox dev jackpot51 works for System76. Buy from them if you want to make sure he keeps getting paid.

>I eat dog shit because I hate human food

They want something that isn't monolithic so they can move Android to defined-ABI userspace drivers. Right now they're getting assfucked by the embedded hardware community because they can't update the kernel on any device without source for all the drivers. Project HAL is the carrot and Fuschia is the stick. There's literally ONE MANAGER at ARM that's getting in the way of them open sourcing the Mali GPU drivers and thus allowing the vast majority of Android devices to run 100% Free code, so Google needs to persuade him around somehow. Honestly I think it'd be cheaper if they just sent him a pair of escorts for a week.

You're wrong. They don't need to move from Linux to do this. Else explain:
>project treble

I mentioned that as the carrot.
>look, all you have to do is make your shitty kernel builds NOT BREAK USERSPACE and work with this HAL
>if you can't do that we replace Linux with Fuschia and force you to rewrite all your drivers as userspace servers that have to obey the HAL anyway

No one ever asked him to install MINIX in ME

Imagine that’s going going to change next year

That was so depressing. They need to get 3.4 out, not more rc garbage.
And switch to time based releases. Immediately.