27 years in development and still no stable release. What went wrong with GNU/HURD?

27 years in development and still no stable release. What went wrong with GNU/HURD?

Other urls found in this thread:

gnu.org/software/hurd/hurd/subhurd.html
gnu.org/software/hurd/hurd/running/debian.html
160592857366.free.fr/joe/ebooks/ShareData/Monolithic kernel vs. Microkernel.pdf
blog.darknedgy.net/technology/2016/01/01/0/
twitter.com/AnonBabble

There's been pickup in GNU/Hurd interest lately since microkernels are popping up everywhere like QNX and L4e. Hurd definitely works on top of Debian and active development is still going on there.

History claims Stallman at the time went with the Mach microkernel because the BSD kernel was too early a release and needed a lot of work, so they focused on Mach instead. They also wanted a new kernel that wasn't monolithic, meaning you can restart misbehaving drivers like you would any app, and it would be impossible for any userland program to crash the OS.

It was too early for bringing kernels into the modern era with hardware-enforced process separation, and userland sandboxes and at the time monolithic kernels were 'good enough' so people went with them instead though lately all that's been changing, esp with so-called advanced operating systems courses being the norm in schools these days and students churning out all kinds of experimental OSs, many of them like GNU/Hurd with radically different internals.

If anybody wants to screw around with Hurd it runs no problem in a Debian VM and on GuixSD

Linux, mach, etc.

If they had adapted it off bsd the world would be massively different.

Richard Stallman objectively hurt the world with GNU, since now microkernels are only looked at as pet projects, not as the godsend and future of tech that they are. Monolithic kernels have just about 0 advantages over microkernels.

Goddamn, I had no idea about the shit stallman did...

The monolithic kernel design actually helped the free software movement a ton because the GPL license means hardware support has to be implemented in free software. I'm not sure the linux kernel could have grown as popular as it did without the GPL or the monolithic kernel design desu.

Nah, Stallman couldn't use the BSD 4.4-Lite release at the time because it was locked up in AT&T/Unix System Laboratories lawsuits. It also required a ton of work to get it working as a standalone kernel and they admitted at the time it was an error to go with Mach instead as they assumed that would be less work to get a working kernel.

Then Linux came around and RMS abandoned Hurd as they finally had a kernel without copyrights to work on.

Regardless work on Hurd persists both with archhurd and debian hurd

They still have his papers as required reading at MIT, specifically the work him and Sussman did on backtracking with circuit SAT solvers using symbolic programming. RMS was the superhacker of his day churning out a crazy amount of projects then he decided to abandon it all for politics when Linux kernel was released and bolted onto GNU userland.

They completed the GNU/Emacs operating system first.

Microkernels fucking suck, that's why.

i just realised hurd is a play on "herd".

gnu-herd.

wow.

I GOTTA HURD, THE NIGGER CATTLE

I thought Hurd had "federated" modules and no core, meaning the system would keep running as long as there is at least one module up.

>According to Thomas Bushnell, the initial Hurd architect, their early plan was to adapt the 4.4BSD-Lite kernel and, in hindsight, "It is now perfectly obvious to me that this would have succeeded splendidly and the world would be a very different place today".

Also it doesn't matter if it works now, the game is ruined. We already have L4, Hurd is irrelevant, but the point is that it tainted the microkernel water, which fucked os development for 30+ years.

Nah, because manufacturers could have released drivers that were OS-independent (to an extent).

You're absolutely wrong, there is not one single area monolithics beat them in. Not one. Maybe around the time hurd was being developed, yeah, but even now resource wise the ring change performance cost isn't even a thing anymore.

Fuck HURD we have MINIX

i don't understand how minix is out and hurd isn't

Parallel is now a thing, because of multicore CPUs and the Hurd, which is similar to a microkernel (but not one in theory) can be run in parallel on multiple different servers using even separate programming languages.

Plus some schools still use Hurd for their OS design grad classes

For anybody who doesn't know, Hurd is a bunch of servers on top of Mach.

The main deal is the concept of translators, which are similar to 9P file servers and namespaces, and how all the Hurd servers are implemented. This design also enables easy persistence/running in parallel across networks if you wanted.

The Hurd actually has its own system for running multiple concurrent instances of itself, subhurds: gnu.org/software/hurd/hurd/subhurd.html

If you want to hack on an OS for research purposes Hurd is the best going these days because it allows for flexibility and persistence so if you fuck something up it doesn't crash. There's a Qemu image here if interested gnu.org/software/hurd/hurd/running/debian.html

you know how the BSDs have a source tree that contains everything

why doesn't GNU have this? it would be nice to have the kernel and userland in one spot and build it all automatically

commie pipe dreams never work

I don't believe you.

Overengineered, designed by committee like the rest of the GNU stuff

Read about why Mach failed as a microkernel and how L4 made microkernels great again

Hurd could become great if it stopped being "big and professional" and instead of trying to offer everything in UNIX focused on offering a minimal design. Deliver a fast pure microkernel design, without caring about security (aka let the servers handle security and don't validate messages in the kernel) or features or user request. After you achieve this goal, freeze the kernel and try to integrate more servers without compromising the purity of the original design.

But GNU can't do this. Look at the GNU utils compared to the BSD utils or at the BASH design.
They may be fast and cover all the bases but when you implementing a new kernel concept that mentality puts the project in development hell for a quarter of a century.

>context switches are nowadays free
Yeah, no

160592857366.free.fr/joe/ebooks/ShareData/Monolithic kernel vs. Microkernel.pdf

>L4 and QNX, have proven that speed is not an argument
against ยต-kernels anymore.

My sources are cited, where are yours? This argument is old and not relevant to the conversation anymore, because we've progressed in design since failures of hurd.

So I repeat: If Richard Stallman had not tried Hurd, or adopted BSD, then the world would be a better place computationally.

They should have written it in forth and they would be done by now and faster than linux too

so i hurd you liek gnu in your kernels

...

>backtracking with circuit SAT solvers using symbolic programming

I just looked into this and I didn't understand a word of it.

>tfw I'll never be that smart

Not him, but it did beat them at being easier to develop.

Anyway, what is the current state of HURD?

Wasted its chance: Failed to migrate to L4.

After that, most people left the project for better pastures.

Better pastures include projects such as seL4, genode, minix3, helenos, fuchsia, harveyos.

blog.darknedgy.net/technology/2016/01/01/0/

>What went wrong with GNU/HURD
The delusional toenail eating homeless man

So-so performance, free software developers mostly not caring about it because Linux always looked more promising and was more interesting to work on.

Stuck in pre-liedtke world.

haha mor like GNU/Turd what a piece of shit

Freetards are autistic losers who never create anything of value, who knew.

>it would be nice to have the kernel and userland in one spot and build it all automatically

Guess what systemd is doing :^)

Linux happened and everyone went to that.

HURD devs today openly admit that they are doing it for fun, research and because it is an interesting project. They have no corporate backing or anything.

>freetard shit
>stable
>working
>reliable
>functional
>on-time
>etc

Only worthless people work for free, only people who's time is worthless use the products of a worthless developer. Its one big ecosystem of losers on both ends. So it kinda works, but only if you're some kind of loser.

>hurd
>interesting
-_-

what i didn't suggest?