Signs of life from GNU Hurd

https://lwn.net/Articles/452296/

by sciurus • 15 years ago

76 62 15 years ago

lwn.net

leif2• 15 years ago

GNU Emacs and GNU Screen do not compile. What?

rbanffy7• 15 years ago

No Emacs? It's useless ;-)

That, OTOH, may explain low productivity by its developers.

lelf• 15 years ago

Emacs worked for me. But that was long time ago, in “by hand”-cross-compiled system, not any Debian.

(BTW, cross-compiling the entire system is tons of fun — for cross-gcc you want libc, but to compile libc you need cross-gcc)

leif2• 15 years ago

That sounds about as fun as rock climbing with your teeth.

Also, our usernames are alarmingly similar in this font.

hollerith• 15 years ago

This time for sure :)

pnathan• 15 years ago

Interesting! If they can really get it ticking this time, it would be really interesting to see how a microkernel plays out in 'popular' use.

It's sort of funny how the major kernels are primarily monolithic (please correct me if I'm wrong wrt recent versions of Windows), but academic research says microkernels are better. Worse is better? First-mover advantage?

acabal• 15 years ago

I'm not an expert in this but I think the issue is one of tradeoffs. I think I remember reading (someone correct me if I'm wrong) that broadly generalized, microkernels have better security at the expense of performance, vice versa for macrokernels. See the "Tanenbaum–Torvalds debate" on Wikipedia.

Either way I'm interested to see if HURD will ever take off in any real sense. For example, could we see an Ubuntu/HURD mix in 5 or 10 years? Will it even matter with that kind of timeframe? Would there ever be any practical advantage to use HURD vs Linux besides more "freedoms"?

plq• 15 years ago

> I think I remember reading (someone correct me if I'm wrong) that broadly generalized, microkernels have better security at the expense of performance, vice versa for macrokernels.

that's true, but the "security" is not really what we perceive in the post-google, massively distributed era.

microkernels made sense back when key to uptime was hot-swappable devices. e.g. if your nic goes awry due to a hardware problem, it can crash its driver. in case of a monolithic kernel, this would in turn crash the os, whereas with a microkernel, the rest of the system should continue humming just fine. this makes replacing the failed component possible without causing downtime by shutting down the single node the service runs on.

but nowadays, we know how to set up systems so that taking down an entire node (gasp!) won't harm the operation of the service as a whole. so i don't think microkernels are that relevant anymore.

rnemo• 15 years ago

You're partially right with Windows. Windows NT, since the beginning IIRC, uses a hybrid kernel in which things like drivers and IPC are and such are still in the kernel, but the application subsystems and fileservers run in user space. Plan 9 has a very similar design.

eru• 15 years ago

Hypervisors like Xen can be viewed as microkernels.

macrael• 15 years ago

Does OS X (and iOS, I suppose) being built on Mach not count?

glhaynes3• 15 years ago

The BSD-style kernel "subsystem" of XNU, Darwin's kernel, is statically linked with Mach and executes in privileged mode. So, while it "has" Mach and some parts of the kernel interact with other parts via Mach port interfaces, XNU isn't really a microkernel as classically defined anymore since it provides all the facilities you'd expect of a non-micro kernel.

wmf• 15 years ago

OS X was built on Mach 2.x, which was pre-microkernel. And the modern XNU isn't microkernel-like at all.

gaius• 15 years ago

IIRC Tru64 or OSF/1 (nee Ultrix) was a mainstream Unix based on a microkernel.

The problem with microkernels is their advantages (in terms of, a component of the kernel crashing won't take down the box) has rarely been worth it in light of the disadvantages (all that message passing hurts performance). So "better" depends on your POV.

glhaynes3• 15 years ago

Would've been awesome in 1980 on a timesharing system. Now I've got as much control of all my personal devices as I want and I don't want any more control of devices at work than what I need to do my job (with control comes responsibility).

Someone• 15 years ago

You still run stuff at home? :-)

I think we would see different comments if the HURD were performing reasonably and AWS ran it.

Having said that, the speed at which the project moves does not help its cause.

epsilondelta• 15 years ago

Hurd is still using Mach? I don't see the point if it's just going to cheat around the IPC problem like OS X did.

And in the past decade, there have been more promising microkernels, like L4. Unless I missed something and there's still a good reason for Hurd to still be using Mach.

ch0wn• 15 years ago

There were attempts to use L4 and several other micro kernels like Coyotos and Viengoos, but as far as I know all of them stalled and the Mach variant has been the only one to be continued to present day.

russellallen• 15 years ago

These articles always compare Hurd to Linux. I'd love to see a comparison with Minix3 or Genode/L4 instead.

bane• 15 years ago

It's probably because RMS always compares (indirectly) Hurd and Linux.

kls4• 15 years ago

RMS has never really liked the fact that Linux was not under direct GNU control. I am sure the rift is greater after the rejection of GPL 3 by the Linux camp. RMS has a vision for how things should be and tries to align the GNU offerings to that vision. Some agree with it, some kind of agree with it, and some think he is a fruitcake. Understanding his vision puts the push for Hurd into perspective. He cannot achieve it with Linux because Linux is steered by stewards that don't exactly line up with his world view.

rbanffy7• 15 years ago

They have to compare to something someone knows about. If you see a comparison of Minix3 and Genode/L4 with Linux, you can determine what the comparison between Hurd and them would be.

onan_barbarian6• 15 years ago

Grits teeth...

It's _still_ pointless. There's no chance of displacing Linux or the BSDs in practical terms, and there's no particularly interesting ideas in it. We already know you can build a Unix on top of a microkernel and it will sorta work.

Enough, already. Kill this late-1980s design and build something interesting already. Surely at least one of the things that have happened between 1990-2011 in hardware, software and networking warrants a rethink of OS assumptions rather than this stale rehashing?

p9idf3• 15 years ago

What do you care what other programmers do with their time? The ideas in Hurd are particularly interesting enough to those who work on it. That alone justifies its continued development.

onan_barbarian6• 15 years ago

This is disingenuous. Under this logic there's no such thing as a project that's more or less objectively interesting.

If 999 programmers decide to work on 999 basically identical Soduku solvers, and 1 programmer decides to hack up a novel object recognition system with OpenCV, and all 1000 guys are equally keen on their own projects, we'll just have to all agree that all thousand projects are equally interesting and call it a day.

I care because it's wasted effort that would be better spent elsewhere.

glassx• 15 years ago

there's no particularly interesting ideas in it

I sorta agree.

I particularly would like to see operating systems implemented using JIT compilation and software isolation. Managed code running safely on the ring0 of the processor.

Microsoft Singularity started this but I'd like to see the open source people trying something like that! I really think we need processor/architecture independency if we really want to achieve freedom.

p9idf3• 15 years ago

Here you go: http://en.wikipedia.org/wiki/Inferno_operating_system

prodigal_erik• 15 years ago

Sadly, neither JNode (Java) nor House (Haskell) look very active. I keep thinking I should start something though I'm sure there are hundreds of embryonic kernels out there already.

wslh• 15 years ago

SqueakNOS? http://wiki.squeak.org/squeak/1762

derleth2• 15 years ago

> I particularly would like to see operating systems implemented using JIT compilation and software isolation. Managed code running safely on the ring0 of the processor.

Lisp Machines were arguably like this, except with microcode instead of JIT. AS/400 (IBM midrange systems, now called System i) does something like this as well, except the whole program is compiled from bytecode to machine code on first run, or whenever the bytecode on disk is newer than the machine code on disk.

The biggest change is that you have either a trusted compiler, a garbage collector in the kernel, or both.

rbanffy7• 15 years ago

Don't you think it's a bit tragic that the two most sophisticated OSs widely available are either direct descendants of or largely inspired by AT&T 60's technology, while the third is the bastard child of VMS?

Try new ideas, prove them with a large audience and see what you can get from them.

One of these days someone will come up with the idea that replaces Unix. Didn't happen until now, but the only way to be sure it never happens is to stop trying new ideas.

yuhong• 15 years ago

Yea, for one thing they all use null-terminated strings thanks to C (even Plan 9 unfortunately) which resulted in many problems.

dredmorbius3• 15 years ago

Don't you think it's a bit insightful that the two most sophisticated OSs widely available are either direct descendents of or largely inspired by AT&T 60's technology?

That these technologies were highly modular, focused on a simple core, emerged as an independent (an unauthorized) side project at AT&T, and were immediately adapted and extended by the academic and research communities?

The principle change that's come about in the descendant / inspired instances present today is that there's a strong assurance that the license of these projects is and will remain free, allowing, say, independent entities, individuals, and organizations to try new ideas, prove them with a large audience, and see what you can get from them.

There will be an idea that replaces Unix. I don't know what it will be. However, it will be called Unix.

praptak• 15 years ago

"One of these days someone will come up with the idea that replaces Unix."

I believe it will rather build atop of it, maybe changing it slightly. Sort of like OOP & functional didn't replace structural/procedural programming - both paradigms still use procedures and both changed them slightly to suit their ways.

beej712• 15 years ago

Perhaps you can pay them to not work on it. Kickstarter FTW!

onan_barbarian6• 15 years ago

Kickstopper? :-)

We used to joke about governmental 'Stop Grants' at our startup to match the 'Start Grants'. The idea would be that they would pay you to "please, just go away".

99992• 15 years ago

I'm surprised by the amount of negativity in this thread. I was initially going to say that writing a working, solid, free kernel was akin to climbing Everest in terms of human achievement. After the first guy does it, does it really diminish the others' contributions? If they do it using a micro-kernel, isn't that kind of like climbing the mountain without oxygen? But given the amount of community effort that is required it's really a lot bigger than the Everest analogy. Maybe it's more like space travel.

Regardless, I feel like these devs are working towards something worthwhile, even if only they benefit from the experience (as they imply in the article, they've already directly benefited from their free work, good for them!). Is there some sort of opportunity cost that I'm not aware of here? Could we say that they should be working on developing a driver for the latest Nvidia graphics card for the Linux kernel instead? Would more people benefit from that? In the next sixty months, probably, in the next sixty years, it's a lot harder to say...

beej712• 15 years ago

Arch Linux^H^H^H^H^H has had a Hurd variant for some time.

http://www.archhurd.org/

I installed it on VirtualBox. Smelled like Unix.

fununclebob• 15 years ago

Something relevant: Symbian has had a microkernel architecture for most of it's life: http://en.wikipedia.org/wiki/Symbian

For a brief shining moment, it was open source. It's legacy is very interesting, worth a read to see just how far the little OS that could has come along since it first ran on PDAs. In fact, it's atypical architecture gives Symbian some advantages over contemporary OSes in terms of absurdly low power and CPU requirements. It is also quite secure, offering granular security settings for almost anything - the only Symbian malware that ever made the news were those that the user installed him/herself. Sadly this also gives it a very complex programming model (for example, it uses an obscure variant of C; also many basic services like audio need to be accessed through a server running on device) and a difficult threading model. Qt provides a beautiful API that wraps over the native layer these days, but it seems it's simply not fashionable among the majority of developers.

Signs of life from GNU Hurd

Related Stories