The Opportunity for Microkernel Operating Systems

Sup, fags?
I decided to make this bread for discussion about microkernel OSes becoming more mainstream in the future.
From what I can tell, there's two that are coming up right now that look like they have potential.

Redox is a microkernel OS being written in Rust (for better or for worse).
It claims to be inspired by SeL4, MINIX, Plan 9, and BSD.
Here's the website:
redox-os.org/
Here's the "book" documentation. It's unfinished, but should provide some info:
doc.redox-os.org/book/
Here's the source code:
github.com/redox-os/redox

Fuchsia is a microkernel OS being made by Google (for better or for worse).
Its kernel is based on littlekernel, which was made by the guy who made Haiku's kernel.
Here's the wiki article:
en.wikipedia.org/wiki/Google_Fuchsia
Here's the source code:
fuchsia.googlesource.com/

Other urls found in this thread:

kernel.org/doc/ols/2007/ols2007v1-pages-251-262.pdf
bsdcan.org/2015/schedule/events/597.en.html
gnu.org/software/hurd/contributing.html
github.com/redox-os/redox/issues/350
theregister.co.uk/2016/01/27/unikernels_dos_era/
draketo.de/english/free-software/howto-hurd-140-chars
doc.redox-os.org/book/introduction/why_free_software.html
wiki.sel4.systems/Hardware/IA32
wiki.sel4.systems/Hardware/sabreLite
twitter.com/ticki_
hypervisor.org/
twitter.com/SFWRedditVideos

Fuchsia is going to replace Linux as Androids kernel and Linus Torvalds will be on suicide watch as Linux' market share plummets

Both are cuck licensed and will therefore in practice almost rxclusively be used for proprietary garbage. Enjoy your FreeBSD with a ukernel.

The reason for that is that they want to reduce the attack surface of their DRM BTW.

what I was thinking is that there would be a potential for an "ungoogled-fuchsia" in the same way that there's ungoogled-chromium, other such chromium forks, and the various more privacy-respecting Android custom rom setups.

any microkernels written by white people?

What do you expect, a microkernel that phones home?

No they're written by niggers and pajeets. That's what you want to read, right? Because you can't google and find out yourself. Because you're lazy.

You know what, just write a microkernel yourself. Do it, faggot.

...

Is the symbol on the right supposed to represent a hernia?

GNU Hurd is still moving at a snail's pace.

If only there was a complete POSIX-2008 built on sel4.

yes, or if not the microkernel, then the OS built upon it.


Yeah the cuck licenses are a shitty part of this, but Redox isn't gonna be POSIX-compliant? I was under the impression that they were.


It's supposed to represent AIDS


Yeah that project is basically dead at this point. Kind of a shame because I would have liked to know what it would be like to use if it were more developed, but it doesn't look like that will happen.

If only you knew what Hurd can do. Imagine reading a CD ISO image through the Internet with an FTP connection through a TOR network connection, through one specific network interface in your machine that has three physical network connections. You don't have to pre-download the whole ISO, you can actually read the ISO image through that specific connection. The access speed will be slow in practice but the idea is that such a mechanism is not provided by some special purpose TOR FTP ISO program but it is built by the standard Hurd translator stack. It's fantastic.

wow that actually sounds pretty incredible! Again it's a real shame that HURD is dead. These microkernels and OSes based on them seem really interesting, and I hope they actually go somewhere with this new line of them like the ones mentioned in the OP, or sel4

Anyone else think the idea of microkernels sounds more compliant with the Unix philosophy? The whole "Do one thing, and do it well"?

I mean with a monolithic one, all of the kernel-type stuff is in one big process that does all of it, whereas with a microkernel, it gets split up into separate servers that each handle one type of component.

I'm not the only one who sees this am I?

You are correct that it takes the unix philosophy to it's logical extreme. I just hope it's obvious to you why it never caught on.

performance issues? Is that really a big deal in today's computing environment though? I mean MINIX is able to run inside of every single Intel CPU just fine, and Fuchsia is being targeted at running on everything from embedded shit to phones to laptops to desktops.

Can FUSE accomplish that?

cuck licenses got us Intel ME nigger

You could but such a software system would not be modular in the same fashion as Hurd. You would need to build in TOR support into your FTP file system. I'm not quite sure how to configure the system such that all this happens on a specific network interface instead of letting the system choose one automagically.

...

it's called a tun/tap interface and iptables

i would like to see Hurd though just in case linus dies and get's replaced by a kike and/or sjw which is bound to happen eventually. linux is fucked when linus goes, who is going to replace him. same but to a lesser extent with stallman, he's spread his message, GNU isn't going anywhere when he eventually succumbs to obesity and toe fungus

microkernels are harmful
rust belongs in the trash
(((google)))

Performance is more of a symptom of some unsolved microkernal problems. There's actually a fair number of microkernals used that you never notice and they can be shockingly fast. In fact, when you look into some of these uses, you realize that *nix as it is, is really kinda shit if you want to use it for everything. QNX and OK4 are example of microkernals doing an amazing job.

The problem with microkernals, is scaleability. In nearly ever instance you see microkernals microkernals being used to good effect is when it's doing a very limited set of things on hardware that doesn't change. This is because the engineers adding modules to the microkernal can have expectations in how things will behave. The moment you want to create a fully functional and adaptable OS out of them you have to start answering technical questions that nobody has come up with a practical answer to yet. These questions must be answered or you'll have trouble making microkernals stable, fast and able to work across all the hardware that you could stick a *nix or windows or whatever else on.

These issues may or may not be a fundamental and insurmountable set of problems. It's just that it's really easy to make monolithic/hybrid kernels work well enough so nobody in their right minds (i.e. with money) is going to burn time and cash to try to remove the theoretical brakes from what is in practicality a university toy or microcontroller project. So this means is you only have university people doing occasional and disjointed research and GNUheads with a lot of free time making slow advancements.

The problem with microkernals is essentially the same set of problems with the monitor industry and why we don't have Phillips Zeus flatcreen CRT's everywhere.
Here's a paper from 2007 that you can read that glances over some of the problems that make Linux as it is much easier to deal, with than if it were a microkernal: kernel.org/doc/ols/2007/ols2007v1-pages-251-262.pdf

Also, do look into OK4, it's fairly impressive.

But user, microkernels follow the Unix philosophy. Using monolithic kernels is how we got to systemd.

You missed the point. It's certain that you can achieve the same outcome in Linux or Unix. The point is in the mechanism that you need to achieve this scenario. In Hurd, it is trivial to "string together" or "pipe together" the pieces of technology in a concept known as "Hurd translators". Hurd is supremely flexible as consequence of the translator and microkernel technology. To achieve this outcome in traditional Unix often means developing a special purpose service (whether in the form of active demons or fully integrated into a single app) to get these technologies to work together.

This. We need a backup plan. I heard from some anons that Linus's successor is already picked out and isn't going to take any shit, but who knows how things will go down when it actually happens.


Interesting stuff. So it sounds like microkernel OSes are really excellent when developed for one specific purpose, but fall apart when it comes to making a universal, scalable OS on the level of GNU/Linux or BSD.

I find this interesting for two reasons. One is because it's kinda true with regards to the main microkernel a lot of people think of, MINIX. It's used a lot in some embedded stuff, but you're not gonna see it on a Desktop or a Server. Same goes for that QNX and OKL4 you mentioned. They're targeted at one specific purpose basically.

The other reason is because of a few examples that contradict this idea. One is Microsoft NT. While it's technically referred to as a "Hybrid Kernel", many anons have claimed that this is mostly just some marketing BS/general BS, and that it's basically a microkernel. I don't know how true that is, or how this should be classified, but Windows is used both on Desktop, and to a lesser extent Servers, and has had some (albeit failed) attempts at entering the mobile market as well. The other is Fuchsia, which as previously stated, is being made as a universal OS. We'll see how that plays out, but yeah

christ, i had no idea why hernias would be such a big deal

wtf I hate monolithic kernels now
not really, but damn you know exactly how to appeal to my inner autism.

Redox is garbage and it's community is worse. Fuscha will go the tizen route and only ever existed as a means of putting corporate pressure on the linux devs. Hurd is our only hope.

Will we ever see microkernel linux?

I don't know but but there is a microkernel NetBSD now.

bsdcan.org/2015/schedule/events/597.en.html

...

I forget, could you please explain for OP why the rust community is so terrible?

Why would you not learn from the mistakes of almost 50 years of unix? Imho "everything is a file" worked well when filesystems were the new hot shit, but nowadays, the filesystem is just yet another namespace just like the network or PIDs or UIDs. This fixation blocks innovation and leads to retarded ideas like "everything is a url". On modern multi user systems, all namespaces should be easily and securely segregable, which POSIX does not offer. Cgroups and Jails are a step in the right direction, but are flawed to maintain POSIX compatibility. Why do executables need to be read from a filesystem and not a pipe? Can't we do better than stdin, stdout, stderr? So many examples come to my mind.
Disclaimer for the prejudiced, the narrow-minded, and the retards here: blah blah poettering blah muh muh

Rust community vs redox community. With about 10 developers from Holla Forums you can btfo the SJWs from redox. You just need to get off your lazy asses.

they don't need, you can pipe shit into bash or python or another interpreter of something, etc.

shit, fucked up formatting. I hope you understand.

ok so since you clearly know everything, tell me how your system would work.
If it's not "everything is a file" or "everything is a url", then what exactly will it be? How will your system work from a user or administrator perspective. Are you sure that your system will not introduce unnecessary bloat?
What else is genuinely needed other than stdin, stdout, and stderr?
If you're going to call the current system crap, you need to explain why it's crap, and what would be better.
also,
aside from Lennart Poettering's last name, this was complete baby-gibberish.

I'm talking about binaries. So what system call is that?


Will do when I'm back from work.
That's the level of retardedness we're dealing with here when you post controversial ideas. I'm sorry. I'm not advocating any of poettering's ideas or solutions.

the same (exec)
you need to write the executable to a filesystem, but it may be tmpfs (i.e. not really on disk), so it'll be the same from performance p.o.v.

Windows 10 is a microkernel OS.

Do planks. Keep those muscles strong.

I've heard that a lot
If that's true, then that makes me hopeful for the success of microkernels in the future.

gnu.org/software/hurd/contributing.html

Stop talking and start working faggot.

That’s not how free software works, faggot.

Why are you still talking, faggot?
Get to work.

gnu.org/software/hurd/contributing.html

Microkernels have literally nothing to do with Unix. The first microkernel was invented by Per Brinch Hansen and it's older than Unix. Dijkstra's THE system also had some similarities to microkernels. This is another attempt by Bell Labs assholes to steal credit from other people, like they already did to Multics innovations.

To be fair, I don't think any of the assholes from Bell Lab in particular try to take credit for this one. They insist that the kernels of Unix and Plan 9 are just small, not microkernels.

The most used microkernel in the world is made by a jew with cuck license. The main point of microkernels in practise is so the manufactures can deploy their shit infested proprietary drivers without having to rely on OS vendor/community. There is no excuse to do it this lazy way, it only helps manufacturers to fuck the consumers with their proprietary malware infested garbage.
We are lucky enough that hardware was shitty enough in the past so monolithic kernels made enough importance so manufacturers are now forced to work with the kernel. Now we at least have the option of someone gatekeeping what shit we have to eat.
Stop shilling micro kernels, the benefits are negligable and downsides too great.

Work takes time, and time is money. It’s not Free software unless I get it FOR FREE.

Still can't update without a reboot. It's much closer to the monolithic though.

Next to nobody use minix, but the processor for the IME.
That's next to meaningless in the embedded world, where microkernel has the fame and success with symbian and QNX.

Only well spent with some mining on cuckchan.

At least the Redox is not a huge fag about it. Even goes as far as to say codes of conducts are useless github.com/redox-os/redox/issues/350
He is based

wtf I love Rust now

You are such a faggot. Every thread someone mentions Bell Labs, you freak out.

I really like the idea of a memory safe kernel, but I don't particularly like microkernels beyond their ability to be implemented quickly as a proof of concept. Bryan Cantrill brings up a couple other really key points. I'd really suggest giving it a read, and given he is talking about unikernels here, but everything he mentions applies to microkernels to a smaller degree as well.
theregister.co.uk/2016/01/27/unikernels_dos_era/

You mean redox
Rust is still filled with idiots and has a retarded CoC.
If you want to get started on rust don't become part of their community.

What is there to get involved in besides an OS? Rust is one of the languages with the greatest project abandonment ratio.

OP here

exactly. I wasn't saying that microkernels were invented by bell, or that they only exist in unix.
I was merely suggesting that a unix or unix-like operating system with a microkernel seems to fit the philosophy better than one that has a monolithic kernel.


calm down. I don't think microkernels have anything in particular to do with proprietary software. I seriously doubt HURD would have had any proprietary infestations. Meanwhile we live in a world where Linux, the GPLv2-licensed monolithic kernel, has binary blobs.
I'm not saying to take the tinfoil hat off, but please loosen it up a bit.


wow it sounds like Redox is *shockingly* using Rust to actually get shit done. Instead of just circlejerking about identity politics and how safe and perfect their language is, they're actually shutting up and using it to make something important.
I think I'm starting to like redox now.

Not even once

I don't like plagiarists and liars. If it was Java or whoever people taking credit for inventing things they didn't invent or saying their "philosophy" was responsible for ideas that already existed, I would be saying the same thing. For some reason, only Bell Labs Unix assholes do that.

Jeremy is actually pretty funny in a somewhat awkward way. I like him.
sage for eceleb post

I don't know how to shut that off, but I know from fucking around with it that you can switch TTYs. TTY1 prints out some kernel info, TTY2 is a regular command line interface, and TTY3 is the flatshit.

hehe, ancaps aren't anarchists.
He will never make a good OS if he keeps licking those boots.
I place my bets on seL4 still.

No they don't. Everyone from Bell Labs seems to always pay their respects to the people who inspired their ways of software development and their ideas. You're just being a contrarian asshole for the sake of feeling smarter than every other *nix user.

Is there any OS with that which I can try in a VM or whatever?

I use Debian GNU/Linux with the Linux-libre kernel. No proprietary software on my machine, and I'm eventually going to make the switch to GuixSD.

Why? Debian's default kernel is unblobbed.

A free society also needs to address the problem of spirituality, something that is not even on the horizon right now. I bet nobody in here even understands why "free spirituality" is even important. I'm willing to bet there's people here right now who even believe spirituality is detrimental to society. That's how bad the situation is.
But I wouldn't like to derail the thread.

Free software is also indispensable.

What does that mean? You can't just invent a term and expect it to have meaning without explaining it.

Redox is garbage and the people making it too.
Why ?
Cuck license
Bullshited people about it
Refused to allow people to retort nicely because "trolls"
Plus all the cult of the code of conduct behind rust.


The numbers of email in the mailing list basically doubled in the past six month but whatever.


This guy gets it
draketo.de/english/free-software/howto-hurd-140-chars


Part of the said problems were resolved but I am now persuaded that it's only at the hardware level that needs to change to reduce the numbers of syscalls.
Thus it's not only a "kernel" design problem but a hardware one too.
I'll need to experiment on a fpga one day (when I'll have the money).


Besides being a corporate cuck that encourages the use of a language that encourage cult mentality.

They day they put the license of the software under GPLv3 they'll have more of my attention.
There's no point of a "free" software if it can be transformed has a black box at the will of anyone.
They'll have my support the day they go for Dlang who's more mature instead of surfing on the rust hype/meme.

Well the problem with communities like rust is that even when you don't want to get involved you get involved.
Why ?
Because it's a fanatic cult.
They'll use anything to show how successful they are, thus if a project like redox gets a great success then they'll go all frenzy showing how much rust is a great success (even tho statistically and technically a failure) and that's how you get involved even tho you haven't asked shit, and that way they'll get more of the same crazies participating in the project and thus ultimately subverting the project at some point.

Either you're retarded or this is bait.

I think he meant spirituality in terms of the doctrine of fascism.
I don't know because he's so vague. Probably something about true freedom and how that differs from freedom in context of the original pdf.
*jk, just shitposting*

Yeah this is the bad part about Redox and Fuchsia. The license is shit. Although Redox does claim that it is committed to being compliant with the GNU Free System Distribution guidelines, and has a whole page about their support for Free Software. On that page, they seem to imply that the main reason they went permissive was because they could then incorporate GPL-incompatible Free Software into their project, such as native support for ZFS.
doc.redox-os.org/book/introduction/why_free_software.html

But I don't think that's worth it when you consider the issues with permissive licensing as you mentioned.


I have no idea what the fuck any of you are talkin about, but it sounds really gay, and not in a good way either.

Oh and still waiting on an answer for this.

I don't think so. I'm not sure if was the first two boards that they have put seL4 on was native or vertualized.
wiki.sel4.systems/Hardware/IA32
wiki.sel4.systems/Hardware/sabreLite

Free software can't be transformed into a black box. Everybody is free to study the software and share knowledge about it.

gpl is THE cuck license

bait

Not the user you responded to.
Let me explain the thing is about the free software movement is that there's a whole people jerking themselves about the philosophy that it create/recreates for the whole population.
Not that it doesn't have any importance but like you reacted most people sadly don't give a shit about the universal values that it transmits, 90+% of people in general only want their shit to work the way they want it to work, it's preposterous I know, but that is one of the second major reason (the first reason being MS mandatory installation on all PCs) why it's god dam hard to make people migrate to another platform.
The only leverage that Gnu/linux had in the past is the server market and it's now gaining a very slow pace into normieland because a few of them have gained consciousness of the botnet, aka they are reacting out of fear and not because of the philosophical symbols that it also represents.
If you want people to migrate to a platform you need to first evaluate their needs, see if the needs match Gnu/linux compatibility, make tests, make them realize that, show the benefice if they migrate, and make a gratis lifetime support for them that they would never had with any other operating system (companies will pay for that tho).
If you don't do that with normies it's useless they'll bitch and brag you to install back windows.
If you're lucky you'll find people who can read and you'll make documentation (still gratis) for them but you won't be wasting time.

Nope, it's hybrid. It's not close enough to be called a microkernel.

Niggerlovers.

And because it's full of passive-aggressive emasculated cynical good for nothings that take any chance they get to either power trip or to try to project and spread the white/male/whatever guilt they feel because they were never taught the value of feeling self-respect as children. And they feel that in this way they are fighting *something* they don't quite understand but that makes them redeemed.


Not exactly. I meant it as the opposite of materialist reductionism or positivism. If not for anything else because it has historically been a counterbalance to those forces, even when abused, and I believe it is very needed right now. I'm not a fascist at all. Yes, I believe true freedom involves among other things free software and hardware.
And also the acknowledgement of a mental quality of life, because we live our lives as much inside our minds as we do within our physical environment. By negating this, humanity is slowly castrating themselves, their ability to dream and create, to come up with alternative ways of thinking and doing. We're turning into robots (and soon we might begin doing so in a literal sense, ironically).
It may not seem so at first glance, but the fact that we shun spirituality and sometimes even simply art more and more goes hand in hand with the fact that everyone talks through Facebook services, mails through Gmail and uses Apple products. Perhaps not as a consequence, but as a symptom of the homogenizing effects of materialism.

As if knowing that thoughts and emotions are chemical reactions in our brain somehow made them less valuable and ultimately pointless in some sort of postmodernist nihilism.
I don't think we should throw ourselves in the arms of magical thinking, religion and pseudo-scientific charlatanism, nothing like that. But there is definitely a place and a function for mysticism in human society, as there is a place for true art and the development and enrichment of our mental universes.

I think spirituality might be beneficial, but I wouldn't be able to get myself to participate in anything that's not honest about what it is. It'd have to be constructed to acknowledge it doesn't represent any true deeper meaning, and is purely a psychological hack.
Does that exist?

I am entirely uninterested in OS development. OSes do not matter now. We must start talking about CPU microcode, open hardware and hardware development, system BIOS, and device BIOSes.

Operating systems are not important. They are essentially user-land code in terms of the real code operating on your hardware. The entire BSD and Linux ecosystems have been undermined and subverted through the modern PC architecture and firmware. Talking about them, their development, and their use are a inconsequential sidebar to what really matters.

Security is not everything. It's important, but it's not the only relevant aspect of operating system design. Operating systems matter even if the hardware is compromised.
Avoid tunnel vision. If something is vitally important that doesn't mean nothing else matters.

ALL "Hybrid" Kernels can be called microkernels. But they can also be called monolithic, hence why they're just called hybrids

For those wondering, the actual microkernel part of NT is ntoskrnl.exe that's loosely inspired by Mach according to it's original developer Dave Cutler

Likewise the hybrid kernel used in Mac OSX is XNU while the microkernel bit is Mach

...

Chaos Magic?

...

Sounds like the way some people do it matches what I'm looking for:

I can literally tell that rust shill apart by the way he writes.

It's a monolithic kernel. They only call it part Microkernel for the "security" marketing point.
Admit it you got bamboozled by PR.

Hybrid kernel is a marketing term invented by Microsoft.
The difference between monolithic and microkernel is what runs in kernelspace.
In the case of NT, network stack, filesystem stack, parts of GUI, and a webserver run in kernelspace, so that makes it monolithic as fuck.
Same for XNU, sans the webserver and GUI.
Fuck off already you retard.

Well, it's not such a stretch, when both products in question are effectively maintained by the same company.
He didn't say Fuchsia would replace Linux everywhere; just as the Kernel for Google Approved OSes, such as in Android.

That said, this probably isn't how things will go, considering Fuchsia still doesn't support running usual Android apps, and would require devs to port all of their shit over.

You are confused. When we say "free software" this means software that respects users' freedom and community. Roughly, it means that the users have the freedom to run, copy, distribute, study, change and improve the software. Thus, "free software" is a matter of liberty, not price. To understand the concept, you should think of "free" as in "free speech," not as in "free beer".

Hybrid microkernels kernels are a hybrid between strict microkernels and traditional monolithic kernels. Strict microkernels have no functionality beyond what is strictly necessary for a microkernel to work. Hybrid kernels include functionality that works beyond that minimum level. The normal reason why hybrid kernels exist is to improve the performance over strictly minimal microkernels.

rustfag once again responding to his own posts to necrobump his rust shill thread.

GET AWAY

please explain

Minix (ouside of embedded and the ME) is dead.

Minix3 is still actively developed, last release was 6 months ago, but even if kernel progress is slow they use NetBSD userland, therefore you get pkgsrc which has over 17,000 packages and is actively maintained by several OS projects.

Genode Framework

It allows you to build a custom microkernel OS which allow you to mix and match various components of other operating systems.

It should with enough development allow the following OS conjunction:
Illumos ZFS
Linux video drivers
FreeBSD network stack

Currently it is mostly L4 and Linux.

That would be extremy retarded.

You’re a smart guy.

For you.

you are very very ignorant

Read it as genocide.

What's wrong with SeL4 or MINIX that the Redox guy or Google can't just use them?

nih

Thanks, user. Didn't know about this, but it's very interesting. Unfortunately, I don't see it ever making it into any niche.

Fair play redox merchant.

lol this dumb cunt got btfo twitter too.
twitter.com/ticki_

i'm still not touching redox because rust but atleast the dev seems alright. it's too bad he chose rust.

somehow i missed this the first time around

actually it's this idiot's repo, xir is the only one in it.

...

Has anyone tried breaking linux out into a microkernel yet?

I think that’s the idea behind Hurd Linux, but development is very slow.

user-developer with full ring -3 access is the way to go tbh

Genode is pretty great, and is actually the most usable of the three OSes so far. It's also architecturally immune to Meltdown and most of Spectre due to being capability-based. They've got ports of VirtualBox 5 (with guest accel and per-device USB passthrough), Qt5, SDL, libretro, and even working 3D accel for some Intel GPUs. In a release or two it will be usable as a secure place to run your Linux and Windows VMs, at least on amd64 with the NOVA microkernel/microhypervisor.

These are doomed for extreme faggotry. Redirect to seL4 at once!

Genode can run on seL4, or other microkernels.

You need to lurk more

Whomever you are, wherever you go, please post your gpg or OMEMO fingerprint so I can read your opinions. I haven't been on 8ch for almost a year because of how inane posters have become, and it's rare post like yours that make me monthly review.
I can't stand Jim, or centralized imageboards, but I refuse to ignore valuable insight from external source, lest I become an echo-chamber.

Check out NOVA. hypervisor.org/ The philosophy is that "hey, a good microkernel and a good hypervisor are very close to the same thing, so let's combine them!"

Way to misread the post

Conventional microkernels are performant because you can efficiently map virtual memory and avoid stale caches.

With this new class of exploits is it possible to design a microkernel that can not leak information without additional cache flushes?

Ignorance of history unappreciated
Mitigation is now security
If hardware fails, should software do what hardware cannot?

What's the benefit of a microkernel over a monolithic kernel?

Microkernels let you choose the opinions you want in your operating system.

It is feasible to prove the correctness of the microkernel due to the volume of code we can study. This means the microkernel can be proven that its algorithms are correct. Microkernel systems can be more reliable because a faulty service can easily be restarted. If that service was inside a monolithic kernel and it wasn't part of a driver module, you'd have to restart the whole computer.