Begone Linux, GNU/Hurd Is The Real FOSS Kernel
Вставка
- Опубліковано 18 вер 2024
- Nowadays Linux is the king of the FOSS world with BSD sitting off to the side doing it's thing, but there was another kernel that could have taken it's place, GNU/Hurd. The only problem is it arrived a little bit too late, since then it's only continued to fall further and further but it's still being developed.
==========Support The Channel==========
► $100 Linode Credit: brodierobertso...
► Patreon: brodierobertso...
► Paypal: brodierobertso...
► Liberapay: brodierobertso...
► Amazon USA: brodierobertso...
==========Resources==========
GNU Hurd Phoronix: www.phoronix.c...
Original Linux Licence: mirrors.edge.k...
Growlaw Page: www.groklaw.net...
=========Video Platforms==========
🎥 Odysee: brodierobertso...
🎥 Podcast: techovertea.xy...
🎮 Gaming: brodierobertso...
==========Social Media==========
🎤 Discord: brodierobertso...
🎤 Matrix Space: brodierobertso...
🐦 Twitter: brodierobertso...
🌐 Mastodon: brodierobertso...
🖥️ GitHub: brodierobertso...
==========Credits==========
🎨 Channel Art:
Profile Picture:
/ supercozman_draws
#Linux #OpenSource #FOSS #GNU #LinuxDesktop
🎵 Ending music
Track: Debris & Jonth - Game Time [NCS Release]
Music provided by NoCopyrightSounds.
Watch: • Debris & Jonth - Game ...
Free Download / Stream: ncs.io/GameTime
DISCLOSURE: Wherever possible I use referral links, which means if you click one of the links in this video or description and make a purchase I may receive a small commission or other compensation. - Наука та технологія
This is the year of the GNU/Hurd Desktop
Finally; a person with perspective
And ipv6.
Microkernels > everything else
@geekzombie8795
ipv4 will nat die.
With Duke Nukem Forever for Hurd installed
Hurd is being worked on by some students who want to learn about OS design and I think the leader is a professor that's having fun with this project.
Exactly my thought, and it is perfect for that purpose... students don't have bother with the linux monstrosity...
@@vaisakhkm783 Andrew Tanenbaum created Minix back then to teach his students about operating systems
@vaisakhkm783: the Linux monstrosity become apparent when you compiled a v2.6 and then some versions later again, as for me v5.15. It’s huuuuuuuuuuuuugggee.
@@vaisakhkm783 I took a class in Linux kernel programming in my last semester of university...it is indeed a monstrosity where many sub-topics could've had an entire course of their own.
It isn't Hurd it is gnu/hurd
"There are nothing more permanent than a temporary solution"
man if gnu hurd was the main thing instead of linux everyone would've had to deal with RMS, like A LOT, and like all the time. That would have been interesting to see.
RMS was shaped by what he experienced ever since. All it needs is ppl with some level of interpersonal communication skills and not just like someone who only speaks code.
and we think that Linus can be an asshole sometimes, we don’t know how good we have it in this timeline
I actually think that it's lucky that Linux beat out Hurd because Torvalds seems to be a better software manager than Stallman.
Nothing that suggest Stallman himself would be managing the Hurd
@@kreuner11 Not sure how good Stallman is at delegating either.
Maybe it was a matter of time
Considering where Hurd is in development, I'm not really sure that it had any hope of ever beating out GNU/Linux.
Stallman is an ideologist to me in the first place. I don't know how good his extreme views on things would implement into an actual functioning and up to date operating system, because even if you're FOSS you have to make trade offs in the current tech world.
Fun fact, XNU (the macOS kernel) is a combination of custom Apple parts, FreeBSD and Mach
Funner fact, keeping with Open source licensing, XNU is open source
I thought it was called Darwin…
@@xanderplayz3446Darwin was the free portion of MacOS (they no longer distribute it), basically the kernel and core userland tools.
GNU HURD is the Duke Nukem Forever of operating systems.
please elaborate
@@Hyperboid Duke Nukem Forever was infamous for being stuck in Development Hell. Took something like 15 years between its first announcement and when it actually got a release. For perspective, most games take 2-5 years to make, and getting something shorter is either a sign of a rushed job or working low-tech (2D games tend to be easier to make than 3D ones).
But Duke Nukem Forever did released
Hurd and Half life 3 confirmed to release simultaneously
@@emretasdemir8028not the one that was originally announced. it took many forms in development and only the last one got completed and released. earlier builds of DNF have leaked and are completely different games
The fact I can sit here and comment on this video while using a free OS with a game running on the side is truly a wonderous thing.
Thank you for mentioning Debian GNU/Hurd. I just downloaded it and now playing with it in KVM, just for the hell of it.
Have fun
It had some truly advanced ideas, it was complex to get working back then but they got it mostly going now.
From what I understand, Torvalds’ original license would be binding, but it would allow a significant amount of debate if he ever needed to litigate it. That would make enforcement a lot more expensive, which was probably prohibitively expensive for Torvalds already even if the license was tightly-worded. Therefore, I think the license was actually a perfect fit because it clearly articulated without hard-to-understand legalese what Torvalds wanted at the time. Basically, it was what it was, and there was/is nothing wrong with that.
Microkernel has obvious advantages for teaching OS development because with the bulk of the functionally being in user space individual students can write their parts without the risk of crashing the system. Personally I first encountered this way back in the 1970s on OS/370. Communications code could run either in user mode or supervisor mode, but you got about 10 percent better performance if your code ran in supervisor mode. The core of both CICS and IMS/DC in particular was called directly from the kernel, and I created a product which similarly ran the most performance critical pieces in supervisor mode. This product supported 20,000 simultaneous users on a 4-core 5 Mips mainframe. Processors were REALLY slow back then. But everything was thoroughly tested and debugged in user mode before it was moved to supervisor mode.
If those are still concerns today and there aren't much better tools to simplify and automate kernel development, I'm actually worried for the state of software engineering in general..
Who cares about crashing a virtual machine?
@@smorrow Well, with user space you are sure to catch the exception with the stacktrace in the VM itself. Not sure if a kernel crash can be debugged that easily.
@@seeibe The problem with monolith kernels and this idea that microkernel is bad because they're slow is that we have to accept sub-par engineering over concerns that really are not related to the software, but to the way hardware runs.
Why are we making hardware that doesn't optimize for doing IPC between different addressing virtualized spaces ?
And yes, I know the cost about the transition from kernel mode to user mode, but maybe passing messages between the kernel/user space shouldn't need to have the entire stack/cache stripped off before the quantum of time of the scheduler, we have hyper-threading after all, its possible to engineer the hardware such that two threads can synchronize a queue .
Sort of the most critical drivers that control virtual memory, everything should really just run in user mode.
Operating system developers work really hard for Intel fixing their stupid architecture, not only intel, but most CPU designers. Which is ironically the philosophy of RISC and having a micro-kernel to begin with, we are going to do the hard things on the software, not hardware. Which makes sense when the hardware is expensive, but we're now in 2020, hardware is cheap, and software is the expensive part.
To be fair, all of those concerns might as well be just historical and don't apply much to modern computers, its just operating systems keeping monolith because they're all legacy. One could totally create a better micro-kernel operating system that would actually be faster on modern hardware, not only faster, but more secure and stable. Specially because it is easier to understand and change.
I once asked RMS about Hurd, and he told me that the main problem was it's Microkernal design. Specifically, he said that debugging errors in a message passing environment was difficult due to a lack of traceability. With a monolith, you get a stack trace to the beginning of time. With a microkernal, you might be able to see the most recent message you were passed.
(pls note that I'm not an OS dev and I asked this like 15 years ago, so exact details may not be 100% precise)
Nowadays you could probably debug that using VMs. Not sure if all the necessary tooling is already there tho
@@mskiptr to be honest I feel like the tracing model for micro services being extended into a microkernels service space makes the most sense to me
Being a microkernel is a major limitation, not just on debugging and driver support, but on performance with good security. Remember the takeaway from Spectre and Meltdown was kernels leaking data to userspace is bad, and preventing it is expensive.
It is probably at least partially solvable by creating a separate class of usermode programs that _can_ leak data, after all, your filesystem driver is probably _not_ trying to spy on your database program, but especially pre SVM, microkernels have a major performance hurdle on x86 based machines.
@@lpprogrammingllc That's why safe-language operating systems are the future™
And that kids, is why you really don't want anything to micro, be it kernel or service, or soft! :D
Based Torvalds wanted even the poorest of persons to have access to Linux.
TLDW: Torvalds is a developer, Stallman is a project manager. Spending nearly a decade trying to pick a kernel and license instead of actually writing any code made a project like Linux inevitable.
Stallman does have plenty of developers under him
@@BrodieRobertson For sure. And not without reason - He's a juggernaut of FOSS advocacy. But in this case, Stallman and Tanenbaum spent more time on the sidelines. A bit like how Linux is taking on PC thanks to Windows being so slow to improve itself and spending more time marketing hardware and locking down access instead
Torvalds hasn't written a line of code in years. In truth both of them are project managers nowadays, but both of them started as developers
you must be very ignorant to call the man (Stallman), who developed the GCC "a project manager"
Stallman is a idealistic phylosopher before anything else, this is his strenght and greatest weakness.
Sometimes it is better to stfu, concede and solve a problem with a compromise instead of wasting time.
It is not always about how you want something done, but about what you want it to achieve.
I bet that Hurd would not be complete today even if it had all the Linux developers on it from the beginning.
There would be no hardware support for it, because you cannot force an ideology on the world, it needs to grow, like in the movie Inception. Linux incepted open source into people and hardware manufacturers and even today it has not grown enough. With stallman that would still be in a halt. Cult vs Movement.
Wow, a well-researched video about the HURD! Well done.
I don't agree with your assessment that we'd be using Hurd simply if Linux came out later. GNU is a lot stricter so there'd be no firmware and thus no support, I think a BSD probably would be what we use
That's also entirely possible
I think same.
The thing about Hurd is, it's so modular that extending the kernel is equivalent to installing application software. People would probably write servers for proprietary firmware blobs against the FSF's objections. Due to the Hurd's design, the FSF has no real way to stop them.
@@Gooberpatrol66 interesting
Plz do a walk-around of the Debian Hurd on a virtual machine, I don't think there's any on youtube.
Systems With JT did one for last years release, but there could definitely be more!
Microkernels were ahead of their time, the hadware just wasn't there. Todays systems have many cores, virtualization instructions, IOMMUs etc. so some of those μkernel ideas are getting implemented. DPDK shows that processing packets received directly into userspace performs very well. Since type-1 hypervisors are often μk, any hardware with OS exclusive drivers is automatically supported
I'm not sure how much of a problem was in the actual hardware. There is a subtle parallel between a microkernel and microservices, and the added complexity that all that message passing introduces.
@@nowave7 didn't we get the complexity anyways with SMP systems? Those systems have hardware coherency protocols and atomic instructions that hide some of the operational complexity
yes and you will write microcode to patch spectre for a service based microkernel where the leak can happen after un unknown number of hops between ring modes and can leak in any of them since processor internal state needs to be kept for a chain of events to execute correctly... and can differ in behavior after any of those user mode services change version. God bless the poor guy in charge of that.
microkernel mfs cutting their limbs and putting it back with tape(this way, if the limbs brake, they can simply remove it and the core will still run)
I liked this wonderfully informative video. May the sentiment analysis algorithm bless it with many views.
Well, what can we learn about this story? The importance of the Minimum Viable Product, or MVP!
The discussion between monolith and microservices happens for any software, really. And it still goes on today. If one is better than the other, clearly it's not by much. The GNU people, used to an academic environment, could not even start the project without settling the best architecture, licensing, etc., because that's how academic funding works.
Torvalds, on the other hand, focused on the most immediate problem. No one could run any of this free software, even for studying. So he released the most basic thing that worked and tried to figure out along the way.
Years of experience developing software now tell us that's almost always the right way to go. Well done, Linus.
“GNU is the system, and Linux is one of its kernels.” -RMS
@@zorkman777 do i care about stallman? stallman has his ego on mars
@@zorkman777Gentoo?
RMS needs to shut his mouth and worry about his personal hygiene. One shower a week would be a great start.
@@TheTransporter007 Maybe start doing some exercise too 🤣
I'd like to interject for a moment...
Surprised no one has thought of porting UML to Hurd as a way of providing a shim for supporting additional hardware yet. After all, a different version of the GPL can definitely run in user mode without violating it at the kernel level - you’d get all the benefits of GPLv3 plus all the hardware support of a user-mode device compatibility layer copied verbatim from Linux. Win-win.
@@09f9 64 bit compatibility is nice because then you can use more than 4GB of RAM. (The limit of 32 bit operating systems.)
@@09f9 and this is why I think that pushing for 64 bit development is the correct thing to do since microkernels already juggle so much I don’t think juggling memory is also another thing that needs to handle if it could be simpler with more bits to access. (Plus it could offer speed benefits over bank switching.)
I am definitely not an expert in anything I just talked about, I’m just speaking with the minimal amount of knowledge I have on computer hardware and software.
I want a micro kernel so I can feasibly compile my os in a reasonable time. Learning os dev is hard when everything is a black box, and time constraints cause a pseudo black box for me
That makes sense, but as Linus has said a few times, microkernels are awful for performance and memory efficiency.
@@rightwingsafetysquad9872 hence why everyone who tried microkernels (Windows NT and Mac OS X were originally microkernels) moved to a hybrid design.
I failed to compile a v5.15 Linux kernel recently at least half a dozen times. And the error output was nothing I could work with.
Back in Linux v2.6 times it compiled and worked.
@@rightwingsafetysquad9872 hmm. I didn't know about that. Even still I think I'd rather tackle performance issues than hours of compile time and monster sized error logs
@@jonathancrowder3424there comes a point when design constraints become the limiting factor in optimization. Sure, throwing compute at any problem fixes it "eventually", but that's a band aid that relies on compute power scaling faster than complexity bloat. Monoliths are more performant, in part because they don't rely on message passing. The tradeoff is initial complexity and compile times.
I do wonder if gnu/herd was the winner if windows, macos and free bsd might be bigger since fsf’s focus on freedom goes to a level with hardware incompatibility
That hardware incompatibility is only there because hardware manufacturers choose to do so. The world would indeed be different if we worked together instead of against each other. Second best example from my pov is the amount of “Linux Distros”
@@zizlog_sound I agree, but to be fair, nearly all distros are just toy projects where someone wanted to make their own "default" configuration, selection of installed software, and branding. There are realistically only 4-5 relevant distros, which all others stem from, and they all exist without any consequence whatsoever to the ecosystem.
Unless you are writing some uber-specific software (i.e. package management), there is no such thing as developing for a specific distro that other distros will not also benefit from.
@@ForeverZer0 ,
you are pretty much agreeing to what I say. When I mention the amount of Distros as another example, it's to criticise the endless amount of them. It's insane to have so many distros.
@@zizlog_sound I was not very clear, I am saying that your point of thousands of distros as "not working together" has an incorrect premise. I personally don't see any use for all the distros either, but there is very little work going into one that can't be shared with all others. With the exception of package managers, nearly everything is universal and can be "shared". It may not be accurate to call it "working together", but it surely isn't "working against each other".
A perfect example is Cosmic Desktop. It is being developed for PopOS, but the entire Linux ecosystem can use it and will benefit from it. This is the same with nearly all software. The differences between distros is remarkably small from a development point-of-view. No one develops "for Ubuntu" or "for Arch", they simply target Linux, and it works for all of them. Again, package management is the only notable exception to this.
I very much agree that the huge number of distros isn't very useful, no argument here, but who am I to tell someone what hobby project they choose to develop?
Linux was in the right place at the right time.
It appeared when the tooling to use it was already made and it had no real, ready-to-use competitors.
For the license torvalds wrote up probably doesn’t help its not his first language
Absolutely plays a factor as well
Looks like it stalled, man. 🕶
Yeah, well, you know, that’s just, like, your opinion, man.
Can't wait for news of exokernels (if they are even used these days)
No joke, I really like the idea of exokernels of simply exporting hardware
it will make things harder for devs creating application for that platform, by adding some more stuff to take care of
That sounds exactly like what I have in mind
@@xade8381 Can you elaborate differently? I couldn't understand what you said (maybe rewriting it in a different manner)
Maybe you tried to say that this would be just another system to support, thus making devs life harder
@@tablettablete186 just check how exo differs from both mono and micro (for eg. resource management section)
u'll get what i said.
can't share link here
@@xade8381 Yeah, YT didn't notify me... anyways
Now I get what you said, but libOSes can be developed by the kernel devs, so applications devs would only need to dynamically link them at startup. Similar to WINE in Linux
I hope they get 64 bit support before they start needing more than 32 bits for timestamps 😅
I was a fan of GNU in the late 80s and early nineties. I wrote the "This week in the Hurd and Debian GNU/Hurd" for a couple of years. Then I started writing my own software and didn't have time to continue with it sadly. (On the upside, my software became the most highly cited scientific Free Software.)
name of your software?
GNU Guix + Hurd is the OS of the future!
literally
@@formbi 🤣
@@vaisakhkm783 The independence of kernel daemons allows for some very nice capabilities (some of them would improve Guix which is the most smartly-made distro even more) which Linux can't really have, but sadly the lack of hardware support kinda makes those advantages inaccessible. One can only hope that some talented hackers do join the Hurd team…
I'm running Guix, but I doubt I'll ever use the Hurd part. Maybe virtualized, on some VPS one day
I want to believe…
Everyone talks about how amazing "microkernels" are, but they're not so amazing that anyone actually wants to use them.
I've heard from so many microkernel ethusiasts that they're better but I find that hard to believe with the current state of computing
The period of time in which everybody thought it was going to be the free OS everybody used was very brief by my reckoning. I was around for that time. I became aware of the GNU project in 1988 or 1989. I was in my late teens at the time, but taking classes at the U of MN.
I was initially not completely sure Linux would succeed because it didn't have a marketing department. And so I tried to use SCO Unix (which was absolutely awful because SCO wanted anybody who did development work for it to pay through the nose, getting the GNU tools up and running on it was a huge pain due to the lack of header files). And when it became available, UnixWare. I gave up and switched to Linux round about the time RH 6.0 came out. It was obvious that Linux was taking off commercially by then.
I never even tried to use Hurd.
I still think GNU/Linux deserved to be the famous combo, GNU/Hurd sounds like a messy hell (both by the name and by their story that you just told us)... Plus Linus Torvalds is just *the* guy for this system!
Could Hurd cannibal drivers from Linux? Add some warper API that pretend to be Linux and driver would still think is run inside Linux, similar how Wine allow you to run Windows apps under Linux.
They did that for a while, it was called DDE. It took too much maintenance so they abandoned it and are using NetBSD drivers now because the NetBSD kernel is inherently modular.
I was born in the year Stallman created GNU! And I like it.
Thanks for an interesting tech history video! More of this please!
gotta love Brodie's history lessons!
ACTUALLY what you’re referring to is…..
... is Manjaro.
The first Linux distro was released before GNU Hurd was, assuming Hurd ever was released.
Think was very informative, thank you!
Brodie is absolutely correct when it came to in kernel tree Linux device driver support, however
I don't think it was GNU Hurd being released after Linux was that major issue but the performance cost of the Hurd & Minix Micro-Kernels vs Linux & BSDs Mono-Modular-Kernels.
That was very likely a factor as well
Alpine really hurds GNU 😂
I love the thumbnail
This is more well researched than the majority of things written about GNU Hurd, congrats
GNU/Hurd is necessary for humanity
It's surprising to me that Hurd is still kicking
I mean it is.. but it's hurd.
Now cover GNU Shepherd!
(and maybe OpenRC, runit, Upstart, and so on)
I have said this before and will repeat it here. In 1991 everything was vapourware except for Linux that was able enough to run and be compiled from the word go. At the time that Windows 95 was released Apache 1.0 was released and the standard Linux distribution could be used to serve sites on the web. Yes everything else caught up with Linux though it took up to a decade and even longer to become stable. Linux was the right product, at the right time and at the right price while the remarkable thing was that opposed to the vapourware, Linux was stable and ran (at that time) on everything in the stable of 386/486 and above .
How about this poorly thought through analogy: Linux was a puny but stable catamaran while everyone else was trying to build Titanic.
@@RedSntDK That was exactly it. Once the C compiler worked and BASH became available they had a functional system that was stable and could serve as a platform to deliver services. Though I will point it out again, without Apache 1.0 appearing in 1995, ISP's right across the world would not have adopted Linux and it would have remained an obscure hobby project.
That stable catamaran was sufficient as a platform for Amazon, Google and later Facebook to build their empires upon while doing the same on the competing products by Microsoft was impossible back then and to a large extent still is impossible.
Great video Brodie.
The Linux kernel is a very big mess afaic. Try going through “make menuconfig” and find what you need. Even after several times doing so, I had to go line by line to find what I was looking for.
Maybe, someone knows why the config cannot be in alphabetical order and it’s of higher priority than actually going by the alphabet.
Yeah have been following GNU/Hurd since the 90s and, well, it's been almost ready for 30 years.
4:03 And so did Stallman lol
I didn't realize Hurd was still around, they've been working on it for 25 years.
I initially read "GNU/Hurd Is The Real FOSS Karen" and I have to say it kinda makes sense, the FSF is indeed the FOSS Karen 😂😂
I was always interested in the GNU HURD and also microkernels in general. It would be nice to be in a world where it's not all (Linux) or nothing (HURD); where different projects can attract enough interest to all progress forward naturally. Diversity is good everyone. Unfortunately, that's not the kind of world we live in, and as it is going now, I don't expect the HURD to ever really pick up and be usable on a production system.
We still have the various BSDs and illumOS at least as alternatives to Linux, so that's good. But it would still be nice to have a true microkernel.
I just hear "mock", not "mach"...
GNU Mok
Gen Z finds out accents exist
Right, I said mach
@@elimgarak3597 English-language native speakers find out that words from other languages exist
@@BrodieRobertson you didn't
More than six minutes in before I realized "Mock" is "Mach".
GNU/Hurd is a living corpse...
... we require a resurrection spell.
If I understood Stallman correctly he very much insisted on the word play and meant GNU to be pronounced like 'new' (just like the animal by the way).
I want Redox to succeed
Stallman is great and we are highly indebted to him. Tannenbaum is a legend and Mach or Minix are great as well.
But... They don't play well with others
RMS eats his own toe jam .. in very public situations
I was not only present for several of the "GNU/Hurd is going to matter next year" waves, I've booted and run Hurd. I remember being very excited, and booting Debian GNU/Hurd, and finding it to be a huge nothingburger, because it did less than the Linux kernel of the time, the free/libre aspects weren't amazing, and when you put Debian into strict libre mode, the amount of working hardware dropped sharply again.
Tbh I'm a bit happy that Linux took hurd spot. The gnu devs and FDC have shown they go pass pro FOSS and into anti-proprietary in some cases and you could never make a desktop work without jumping through a ton of hoops and making comprises with that mindset imo
I recall reading about the Hurd micro kernal back in 1999 shortly before i switched from Windows 98 to Suse Linux 6.3. I also read up on FreeBSD, but I it looked like Linux with it's better hardware support (which was still atrocious back then) would be most likely to actually work. I recall I still had to learn to optimise the kernel settings and recompile the kernel in order to squeeze better performance out in order to be able to play DVDs without stuttering on Xine.
Back then, I was fairly certain that FreeBSD was going to be what Linux is today. It was certainly more complete and stable than Linux - and in the beginning, it even had slightly better hardware support. Of course, once the big corporations started backing the Linux kernel, that quickly changed. The rest is history.
x64 Amd GNU
Brodie please look up how Mach's name is pronounced (I don't think I can put a link here)
ua-cam.com/video/VnQbh85sYCw/v-deo.html
The fun thing about Australian English is it cobbles together a bunch of pronunciations
@@BrodieRobertson Mach isn't an English name anyway
Mach = mock
@@BrodieRobertson I can't think of a single time I heard someone pronounce Mach as mock in Australia (I've heard "mack" and "mark" but never "mock"); what part of Aus are you from?
To be really fair and accurate, if linux hadn't existed we wouldnt be using Hurd, we'd be using BSD. BSD was ready, but was in litigation. A frw years later and it would have been a far far superior choice to untested software hacked together by the internet.
Just a thought.
I was curious about Hurd again, but it still doesn't support 64 bit.
Hmm, given Debian/LINUX does not support i386 or even i586 something to run on older hardware with 65% of the repo available! Wonder if LibreOffice ported!!!!!
Hurd and Half life 3 confirmed to release simultaneously
GNU/Hurd was never the new hotness, it was strictly overshadowed by Linux, FreeBSD and OpenBSD.
I use Arch Hurd btw
GNU/Hurd could be a great fallback/alternative to GNU/Linux, now that I'm just learning about this.
Yes, I did know about GNU/HURD, but no, I wasn't around when everyone was excited about.
I knew Linux and played with it some even in late 90's. I knew BSD and tried to install it because friend at school had unhealthy fascination with FreeBSD. GNU Hurd... imagery of rolling tumbleweed comes to mind ...
Wait Minix used to cost money back in the day? Tannenbaum literally has the source code at the back of the book... all of it. Wow first I've heard of it.
That's par for the course for the (notoriously greedy) textbook publishing industry.
I wish HURD was on par with Linux, it’s always nice to have some form of competition even when both are for the benefit of their users. (Plus it would be really cool to have a completely Free Software operating system.)
there are several Linux-based distros out there which are completely free, and you can theoretically use the Linux-libre kernel in any distro
@@shallex5744 I know although more competition is better for all of us, it makes us double down on making our services better for the user/customer.
*shrugs*
Bsd Unix says hi
You should never write 'GNU/Hurd'. If you want to refer to the kernel, it's just 'Hurd'. If you want to refer to the whole OS, it's just 'GNU'. If you want to emphasize that it's a GNU project, it's 'GNU Hurd'. If you're running it with Guix, you can also call it 'GuixSD' or 'GNU GuixSD'. But the designation 'GNU/Hurd' is pointless, because unlike in the case of GNU/Linux, the name does not denote a hybrid system.
I think stallman hates when the Linux Kernel has to make use of binary blobs
Yes
Yes, but note that technically the Linux kernel doesn't use any binary blobs. It merely (usually) ships with binary blobs to load into peripheral hardware memory to make the hardware function. The binary blobs could be on ROM chips in the hardware instead. The reason they're not always on ROM chips within the hardware (some of the time they are) is that it is easier to update firmware on peripheral hardware if it's loaded from the system drive than if you have to flash it to a ROM chip.
Richard Stallman tries to use hardware that has open source firmware as well as drivers rather than just using hardware that has its closed source firmware contained on ROM chips (which would also not require binary blobs to be loaded from disk). He goes to great lengths to avoid any closed source code, be it software or firmware. This is very difficult to do with current hardware.
The problem with Hurd was *not* that it came late. The problems were the following: 1. it never became really stable, 2. it was designed for a fast full process task switching, and the x86 family of processors are notoriously slow at that full process task switching. A monolithic kernel is the best design choice on such a flawed processor family. The microkernel reliant Hurd kernel would perhaps be the best design in an alternative universe where technology isn't as inherently lousy as human technology.
Part of it's issues were because it came late and developers were not interested after Linux showed so much promise
Gosh i waited yereears for hurd thinking it was going to be amazing....
I just wonder if Brodie would actually be able to talk in handcuffs.
Herd is not Unix. It's similar but different and not taught in any university. Linux is a real sysV API and syscall compatible with Unix
At last good news!
The Hurd ideal is what NextStep was doing before Apple bought it and made the MacOS system 10 work as well as it did in the old PowerPC days. Unfortunately, reality set in and going from the old (MacOS/Next) Mach kernel into the modern MacOS kernel had to take place for architectural reasons - kernel space is where it's at for hardware access. Until hardware allows faster access outside of ring 0, that's just pragmatism.
That's what we have now, with virtualization extensions and IOMMU. I wonder if any current HURD research takes advantage of them?
@@shanent5793 there were some attempts, but they really didn't go anywhere
I knew about Hurd for a while now, but it was always the clear loser of the race. I still wish it would eventually be usable, as I prefer the minimalist approach more, but that may only be ways into the future, or never.
The reason that HURD is not what ended up being used is a lot more about performance problems of microkernels than it is about when it was "ready."
Notice that none of the major operating systems for x86/x86_64 use a real microkernel (they use hybrid kernels because of microkernel performance issues).
I highly doubt that HURD would be heavily used if its development were two or three years further along at the time Linux came on the scene.
Oh, seemingly i was Born because the FSF was foundet in 1985 ? :D not? nor correlation because of my Birthday? :D
there is also arch hurd version
Linux as a name sounds way cooler than GNU, HURD and even BSD
arch hurd btw
i tried debian hurd hack in 2000 i belive, but almost nothing was working, only std GNU tools worked, not a good memory
So linux is not a microkernal, but is not monlithic, which of the definition is large and unchanging. Linux is a large kernal, but it does have the sbiloty to change. It is called loading or inserting modules.
As micro-kernels go, I find Redox to be more interesting.
I'm glad it got supplanted by the Linux kernel. I dislike recursive acronyms.
maybe one day i will install nvidia driver in Arch/Hurd
Are you saying we could have BSD's kernel from the beginning?
Speaking of Hurd you could talk about Redox OS
Which uses a RUST Micro-Kernel