From comp.os.minix, January 29, 1992
Andrew Tanenbaum argues that Linux’s monolithic kernel design is outdated. Torvalds responds, and an extraordinary debate about operating-system architecture follows.
Other sources:
I mean, technically it is. Made for terminal, but emulates them. Mostly to run desktops, which are simulated. The initial tty is made for modems. The underlying gcc even has things like a dns resolver built-in, separation of concerns my ass.
“many companies that are household names today–Netscape, Yahoo, Excite” lol
The down votes are because of the stupid clickbait title
The downvotes are because of the stupid people who downvote things they didnt bother reading even the first paragraph of
Yup. But it’s also the title of the Usenet thread so… ;)
I agree of not doing editorial changes to titles. But if a title is not good enough, I usually add my own context or information after it.
Good call. Done. Added the year it was posted
Linus “my first, and hopefully last flamefest” Torvalds
This signature is even funnier with the benefit of hindsight.
The debate was interesting, and AFAIK the Linux kernel of today is kind of a mix between the 2 philosophies.
Sad to see this downvoted, it is an interesting aspect of Linux history, but I think maybe the headline should have reflected that, instead of looking like a statement of fact, which it definitely is not.
A parallel attempt to Linux at designing a micro kernel by the GNU team, has completely failed to gain traction.What made Linux succeed however, is the adoption of the GNU license, this is what made the huge difference back in the day IMO. A community project that would always remain a community project.
And the Linux desktop today is what it is because the GNU license protected the integrity of Linux. AFAIK 80+% of the Linux desktop is based on GNU licensed code.I agree, the title should’ve been a question with a brief description.
Or like in Hacker News, they put the date when it’s old stuff (something like “Linux is obsolete (1992)”)
That would’ve been a better idea on my end
In case you didn’t know, you can edit your post title!
Done! Thank you
The debate back then was already Tanenbaums fight against his own Minix’ obsolescence.
As I understand it Minix was quite popular among the scholars it was designed for.
But Tannenbaum wouldn’t make it “free”, although it did come for free with the book and you could get the source code for free. It was not licensed to enable further developments by users.I stand by my claim that the problem was 100% the license. If Minix had used the GNU license, chances are pretty good that Minix could have become what Linux is today.
But of courser there are other factors, like how Linux worked with GNU, which was already a nearly finished OS that just lacked a kernel. So adding the 2 together and voila you had a complete OS.
I stand by my claim that the problem was 100% the license. If Minix had used the GNU license, chances are pretty good that Minix could have become what Linux is today.
Reading through the posts, I have to agree. It seemed like a lot of the folks wanted to turn MINIX into a full fledged OS, with all of the features which were emerging at the time. But Tanenbaum’s need to stay “the head surgeon” locked it into a paradigm that stopped progress. He couldn’t accept the idea that there would be 100 different flavors of MINIX, all distributed by different groups.
I also think that David Miller’s post near the end of the thread was spot on. MINIX wasn’t a fully fledged OS and was never meant to be one. Tanenbaum created an OS and book for his students to buy when they arrived in his class (as is tradition in academia). And he did such a good job of it that those students then turned around and tried to turn it into the OS Tanenbaum never intended it to be. And Tanenbaum resisted that. Had he been willing to let go of the OS and let it grow outside his control, we’d probably all be using derivatives of MINIX today. But he didn’t, so Torvalds was inspired to write Linux and threw it out into the ether. It was intended to be a full OS and was also available for people to make a 1000 different distributions of. Tanenbaum seemed convinced that this would cause chaos and would lead to an unsustainable mess. And he was kinda right, when someone says they use Linux today, you have to ask them which flavor before you can offer them any help. And we get all the flame wars and stereotypes which come with the various flavors (I use Arch, btw ;-) ). What I don’t think Tanenbaum, or any one else for that matter, foresaw was that this would all be OK. I also don’t think it would have worked before Al Gore “invented” the internet. By opening the door to the world and people everywhere, we now have a digital space where a thousand butchers with axes can come along and hack at a shared pig. But instead of complete chaos, the end result is a fantastic OS, with a lot of chaos around it. Like the chaos of life itself, there is a lot of beauty in that chaos.
Indeed. Tanenbaum treated it more as an academic thing, not like a viable product. It had potential, but was totally wasted.
But by making that failure, he also failed the academic thing by not allowing users to make and share changes.
He saw it as a teaching tool, as context and to promote his book, which is a huge limitation on a teaching tool in a quickly changing environment.
Argued. Decades ago. And now? Who does even remember Tanenbaums’ “Minix”?
There is an opinion that nowadays Minix is more widespread than Linux. Minix is used in microcontrollers, CPU firmwares, in many other places. It is just hidden in blobs that end-users don’t tinker with.
Even considering the billions of android devices using the linux kernel? Did someone do the math?
Anyway, that’s very interesting, kinda like 2 animals who competed in the same environment and natural selection made each of them specialize into a niche
I just looked into this and it’s in the Intel Management Engine.
Inhowfar it can be actually called MINIX in there. If you have a minimalistic 8-bit processor working as a boot processor for an Intel or AMD CPU, they don’t implement features like virtual memory down there.
I believe it is more like a MINIX-ish system down there.
Inhowfar
Daemon summoning isn’t fair! I’ll tell mama!
Statistics schmatistics. Like sure, one can say that Minix is used in microcontrollers and all around the world and thus the usage ratio between Linux and Minix is about 3% vs 98%; but one could also argue that these microcontrollers and firmwares are not actually “used” by someone, they are just middleware and accessory just in the same way we don’t count count LED headlights separately for tallies on driving (and pedestrian) culture, and then we can firmly say that ratio is now 3% to ~0%.
Minix being “used” widely is correct. It is installed in x86 computers to allow NSA a proper backdoor to monitor EVERYONE.
Thus, complaining about Chinese backdoors is a red herring. It is simply choosing whose spyware you prefer, USA vs China.
Me. I set up a minix system a couple years ago.
I did, too. But for me, that “couple of years” is more than 30. And even back then, it was already “legacy”.
Well, you might just be old then. It was 2017 or so for me.
MINIX? In this century? My goodness, what ancient machinery was that?
A Packard Bell 286 that someone pulled out of god knows where. It was running DOS and based on what I saw it must have been from a car dealership, until I accidentally plugged a drive reader’s power plug into its drive upside down and killed it.
A CF card makes a great IDE SSD, so when I put minix on it, that thing flew. I had an ISA Ethernet card, but I never got as far as getting the drivers compiled.
Why can’t they just duel or something. Keyboard fencing. Mouse throwing. Vim beeping…
For people downvoting, if you actually read it, you’ll understand it’s not what it looks like. It’s a piece of history, not an opinion.
yeah the micro vs monolithic debate has been going on for a long long time. so far monolithic has won out usage wise.
Monolithic is just way easier to implement and maintain while also providing better performance





