How can we prevent kernel bloat from destroying Linux?

Most of these "bloat" lines are in the growing number of drivers, are they, user? Drivers that are usually built as modules and not loaded unless the hardware they're for is present.

Why though? Take MFM/RLL disk support for instance. It's in the codebase, sure, but it only gets into a build if you turn it on.

when did more lines of code mean less performance? More lines of code could mean the opposite

Install Gentoo

Just download it and compile it with your own requirements

*notices secure microkernel*
OwO whats this?
sel4.systems/
genode.org/

millions of what?

lines of code

Generally slower and less compatible with hardware

why would you want to use linux anyway, the code base is run by fucking SJWs and tranneis trying to ruin it by using libtarded source code

that's just a stupid criteria to conclude that the kernel is bloated, what are they going to check next, how many spaces the kernel has and how superior tabs are?

It's mean less maintainable, and Linux has been consuming more and more ram

Lines of code.

The question is, does it really need to be debloated at this point? It's not even 100MB, at the current rate it will be 200MB by 2030, they stopped making ramsticks that small 12 years ago. The wasted power on loading those legacy modules is maybe 100W per year, globally.

>OwO whats this?
inexistant

>what are they going to check next, how many spaces the kernel has and how superior tabs are?
dont give them ideas user

I want Linux 5.0 to be at least 1GB in size.

I don't give a shit about how big a kernel is.

Can't you compile your own kernel and remove unwanted architecture?

big numbers = takes longer to process

>25m lines of code is alot

There's no hope for Linux. To the trash.
And adopt anything listed here:
microkernel.info/

I'm not a systems developer so bear with me here, but why does Linux include all of the drivers in the kernel? Why not just have the drivers as a userspace package built independently by the manufacturers? I know that userspace drivers will be slower but surely there's a way to fix that?

Attached: 1488587357910.jpg (882x624, 122K)

Remove support for legacy hardware, and move it into linux-legacy or something

so you're gonna alienate half the user base ?

Yes. Get new hardware, or run the legacy kernel and get over it

When did Theo start using Jow Forums?

>I know that userspace drivers will be slower
blog.darknedgy.net/technology/2016/01/01/0/

Attached: 1530865223282.png (1920x1090, 2.15M)

i shouldnt have to spend like 400 bucks to get new ram cpu and mobo. I should be able to continue using my pentium 4 without hassle

>hardware support is bloat
Kys yourself, dumb zoomer.

Get hardware comanies to standardize on a driver for all their new products.

drivers arent a lot of code lmao

ever heard of .config? no you haven't because you're retarded and you should feel retarded.

Drivers are the majority of the kernel user.

>scolds someone for referencing old data
>immediately proceeds to reference old data
Into the trash.

>referencing old data
Referencing wrong data. It's a myth.

its because of the way they are written though

didn't come across as scolding to me, just a reference of information.

>its because of the way they are written
You mean properly?

I'm taking about contents of the article. If it would disregard it's own argument if it could cherry pick supporting evidence right at the start, there's no reason to take seriously anything it's says, or even read it past first paragraph.

no, they can easily fit them all into 100 lines, it would also be more efficient because you dont need to process more lines

it covers myths one by one for a reason.
Many people are as ignorant and as closed minded as you, when it comes to microkernels.

this is terrible bait

Get rid of system alphabet. 2 million unaudited lines right there

have you actually looked at driver source code ? its not an efficient way to write

I haven't, got a link?

github.com/Microsoft/Windows-driver-samples

>XYZ is a myth stemming from outdated evidence, here's some more outdated evidence that says otherwise!
Yeah no get fucked cunt. I'll believe microkernels aren't slow when you can prove this by example. By which I mean when ship a microkernel OS that's not any slower than Linux. I don't need to hear any sophistry and hypothesizing, I take nothing short of hard facts.

Linux takes microseconds to switch context and send a message. seL4 does it in a couple hundred microseconds.
On the same CPU.
How many extra hops would seL4 need by running drivers in userspace to actually make it slower?

convenience. it's easier to just download the source code, enable what you need, and be done with it.
much easier than visiting 20 different domains to get all your hardware working.

What makes evidence outdated? Is there any new evidence that contradicts it?

As I said, until it's in an actual OS, none of that means crap.

>all these retards who don't realize the kernel can be compiled without the unnecessary drivers
Jow Forums is dead

Attached: 1512326729959.jpg (1200x1000, 174K)

Gee I dunno, the fact that it's from almost 30 fucking years ago?

I guess we can just throw away all that research Galileo did on the scientific method then since it's outdated.

Are you saying that if you run the same code on the same hardware today, you'll get different results?

Havoc and Malice, then

Attached: LoK.png (356x579, 408K)

>still thinking 1st gen microkernels and their downsides are the state of the art of the microkernel approach
>unironically

It was never alive to begin with.

tell that to everyone who moved to lainchan and the cripple's board.

Excellent digits.

But those are Windows drivers...

Yeah? Computers today are entirely different beasts. You need entirely different approach to optimization to get good performance. Such as maximizing cache hits and linearize memory layout, neither of which was of particular importance in the 1990.

drivers between OSs dont change that much since they dont really interact with the OS

>microkernels are superior
>but I won't show you
>here is data I pulled out of my ass
>so just believe me, ok?
The absolute state of microfags.

That's interesting and covered superficially in the speech these slides are from:
archive.fosdem.org/2012/schedule/event/549/96_Martin_Decky-Microkernel_Overhead.pdf
Combined with the microkernels are slow article, this serves as a good introduction to microkernel's potential.
For more current research, seL4 and related papers: ts.data61.csiro.au/projects/seL4/

Is nobody going to mention that kernel devs only focus on the areas that they expertise in and normal people will compile the kernel with just the features they need?

I swear Jow Forums is full of humanities-tier cocksuckers these days.

Laws of physics didn't change since then. Computers - did.

yes they did actually. classical physics and quantum physics make very different implications.

>laws of physics didn't exist before we discovered them
nigga u wot

physics is just how we interpret or understand things

they didn't change but our science and interpretation of them has.

>new hardware
You mean hardware with Intel ME and hardware-level DRM baked in?

>hardware-level DRM
this is a thing?

99% of the kernel is drivers

im not a freetard and I know this

>believes in the Intel ME conspiracy

Attached: 1379388446876.gif (493x342, 393K)

Only thing that comes to mind is HDMI ports that try to block you from connecting to recording devices

>implying it's not confirmed

Name one time Intel ME was used against someone

It's by design. No way to unfuck it

Yes, you can't playback certain commercial 4k content (e.g. Netflix) via the web player (so PC, etc) without Skylake or later Intel CPUs which have hardware DRM. And you also need DRM in your display cable and a compatible monitor/TV. Anyone supporting this by buying and promoting new gen hardware is a consumerist retard.

>black box processes and OS running at a higher privilege than your own OS
>network stack built in
>conspiracy
Goddamn retarded pass user.

Same response: Go ahead. Cite a source for one time this was ever actually used against someone and confirmed as spyware.

Give one reason why they don't document or give source about something that is technically a hardware backdoor?

Because there is no source, since this has never happened.

he means source code, idiot, not academic citations.

Same reason most other software companies don't give their source code. It's their product, and if you choose to buy it, then you're accepting that they have source code that you don't.

Still waiting for ANY instance Intel ME has been used against someone. Go ahead. Prove it's malicious and I'll eat my hat.

>Prove it's malicious
I don't need to. this is literally how Intel advertises it: a device to keep your employees in check
intel.com/content/www/us/en/architecture-and-technology/intel-active-management-technology.html

Employees have no right to privacy on work computers. Prove that the people who set up those work PCs had their computers remotely accessed using Intel ME. Prove that anyone buying their own computer with Intel has been remotely accessed using Intel ME. It can be from anywhere in the world. Please, prove me wrong.

t.intel

*clicks*
Hmm, your link to a news article doesn't seem to be working. Surely there must be some instance this has happened that made news headlines somewhere in the entire world. You're telling me no one has traced their network traffic while using Intel, and caught the government or Intel remotely controlling the computer using ME? What a surprise!

Attached: 1366730772216.jpg (589x564, 179K)