Slashdot Mirror
Prof. Andy Tanenbaum Retires From Vrije University
When Linus Torvalds first announced his new operating system project ("just a hobby, won't be big and
professional like gnu"), he aimed the announcement at users of Minix for a good reason: Minix (you can download the latest from the Minix home page) was the kind of OS that tinkerers could afford to look at, and it was intended as an educational tool. Minix's creator, Professor Andrew Stuart "Andy" Tanenbaum, described his academic-oriented microkernel OS as a hobby, too, in the now-famous online discussion with Linus and others. New submitter Thijssss (655388) writes with word that Tanenbaum, whose educational endeavors led indirectly to the birth of Linux, is finally retiring. "He has been at the Vrije Universiteit for 43 years, but everything must eventually end."
"Microkernels are still better, you little punk!" With an engraving of a shaking fist.
"Vrije" is a Dutch adjective, meaning "free". So either you write "vrije Universiteit", or you write "Free University". "Vrije University" is just silly.
"A multithreaded file system is only a performance hack. When there is only one job active, the normal case on a small PC, it buys you nothing and adds complexity to the code. On machines fast enough to support multiple users, you probably have enough buffer cache to insure a hit cache hit rate, in which case multithreading also buys you nothing." - Andy Tanenbaum on the "LINUX is obsolete" Thread from 30 Jan '92
Nice to see a so called "expert" so far off. Seriously, not the first CS Professor to be completely backwards. I've met a few of those too. :-)
We suffer more in our imagination than in reality. - Seneca
I don't need no stinkin' OS.
Get free satoshi (Bitcoin) and Dogecoins
A lot of people have the wrong impression about the good professor after the infamous exchange, but they miss that this is what academics do, and despite the flameyness of the exchange, Linus and Tanenbaum had a great deal of respect for each other. After all Linus was, for all purposes, Tanenbaums greatest student. I remember borrowing his book from UWA and getting the disks from the UWA computer club, following the instructions to get a functional minix up, then following his book to write a driver for my highly bugshit WANG (yes that was the brand name lol) hard drive controller. I learned more from that about how computers *really* work, than almost any thing I've ever learned. The difficulty of his book was notorious, probably the only books I found harder was Walter Pistons music theory book "Harmony", and Deleuzes philosophy text "Capitalism and Schizophrenia". And like those books, in its field Tanenbaums work shook the foundations of academia.
Enjoy your retirement old man, you deserved it.
Excuse the Unicode crap in my posts. That's an apostrophe, and slashdot is busted.
More than Minix, I remember Tanenbaum for his "Computer Networks" textbook. Especially this:
"The nice thing about standards is that you have so many to choose from; furthermore, if you do not like any of them, you can just wait for next year's model."
"Vrije University" in the title sounds realy strange to me, as a native Dutch speaker. Vrije isn't a city, "Vrije Universiteit" means "Free University," which indicates it's not linked to e.g. the Catholic church. Just FYI.
Does this mean the death of Minix3? That would be a shame I'd like to have seen a good open-source microkernel OS - a sort of "open source OSX".
So...Dr. Tannenbaum's other project is Electoral-vote.com (2), an election prediction site (and one of the first). Any clue what's going to happen to that?
Finding God in a Dog
"You can download the latest from the Minix home page"
Is he still charging $169 for Minix?
I own both "Operating Systems: Design and Implementation" and "Distributed Operating Systems". When I saw the retirement announcement, I cracked them open for the first time in many years to recall how much I learned from them.
But my favorite of Tanenbaum's works is "Structured Computer Organization". I suppose it may be a bit dated, but I still recommend it to anyone who wants to know how computers work.
I'm sorry, but the best quote from that book is actually this one:
Never underestimate the bandwidth of a station wagon full of tapes hurtling down the highway.
In my networks class, we extended the calculation to a 747 full of DVDs (the best we could do at the time). Maybe one of these days, if I have a minute, I'll go back and do an A380 full of flash drives.
I ran Minix for a year or more on my Atari ST - having a UNIX-like operating system on a machine I could have at home was a truly awesome thing. Tanenbaum's work is fascinating, useful and will be around for a good while...which is more or less the definition of "successful" in academic circles.
The debates with Linus were interesting - but I always felt that they were arguing at cross-purposes. Linus wanted a quick implementation of something indistinguishable from "real UNIX" - Tanenbaum wanted something beautiful and elegant. Both got what they wanted - there was (and continues to be) no reason why they can't both continue to exist and be useful.
Tanenbaum's statement that the computer would mostly be running one program at a time was clearly unreasonable for a PC - but think about phones or embedded controllers like BeagleBone and Raspberry Pi? Perhaps Minix is a better solution in those kinds of applications?
www.sjbaker.org
I remember when Microsoft paid Ken Brown throug the Alexis de Tocqueville Institution to do a hatchet job on Linus claiming that Linux was stolen from MINIX. Now Tanenbaum, who has criticized the Linux kernel design and had some spirited exchanges with Linus, could have just said nothing and let Linus fend the FUD off by himself, but instead he stepped up and did the honorable thing by decimating Brown's arguments that Linus could have come up with the Linux kernel in just a year and his competency as a researcher/writer.
http://www.cs.vu.nl/~ast/brown/
http://www.cs.vu.nl/~ast/brown/rebuttal/
They have to use "Vrije" because it was discovered that not only is the name of "Free University Compiler Kit" obscene, but it's also misleading: the software is non-free.
Even today there is usually one process grabbing most of the CPU time
Yeah, the antivirus.
in most scenarios Andy's single-threaded file system access would still serve single-user desktop machines quite well.
Is a single-threaded file system still practical on multi-spindle PCs? These include machines with a boot SSD and a data HDD, or a boot HDD and an optical drive, or a boot HDD and an external USB SSD used for sneakernetting files too big for the available Internet connection. And by "desktop" did you mean to exclude laptops?
This is truly one of the very few profs who can talk about software design.
In the early 80s, I did a Unix systems startup in the UK: we were an early licensee of Unix from AT&T and sold VAXen with BSD installed and supported. DEC UK hated us. DEC US happily sold us CPUs.
In April 1983, the European Unix User's Group (EUUG), held a conference in Bonn, Germany. The speakers included Bill Joy, Sam Leffler, Steve Bourne and Andy Tanenbaum.
It was a hugely memorable event, including Prof. Tanenbaum's presentation. We were paying AT&T $200 or so for each Unix license. Not a huge deal for a $100,000 VAX system. But, even then, many of us could see a future where Unix or something like it would run on countless devices, including cars and washing machines. In fact, when I worked for AT&T in 1984 (yes, I know, it was "a learning experience"), I was pitching exactly that to OEMs. It was clear that something cheap or free would be required. So, back in 1983, when dinosaurs ruled the Earth, Prof. Tanenbaum gave us all the seed of a thought that free (as in beer) software could change the world.
As an aside, his presentation was a little hard to follow, but worth the effort, because his English wasn't that great. A Dutch guy sitting next to me said that his Dutch was pretty sketchy, too. I have no means to verify this but, if true, he would join a small group of my friends and acquaintances who don't speak any (human) language well. They're all engineers :-).
I also learned that, despite Bonn being largely flooded because of heavy rains, nothing stops a Unix conference, and that the "Geoffnet" signs I saw all over the place weren't a promotion for a new network stack, but meant "Open" in German.
D'oh. Accidentally posted as a Coward and misspelled Prof. Tanenbaum's name. Carry on....
In the early 80s, I did a Unix systems startup in the UK: we were an early licensee of Unix from AT&T and sold VAXen with BSD installed and supported. DEC UK hated us. DEC US happily sold us CPUs.
In April 1983, the European Unix User's Group (EUUG), held a conference in Bonn, Germany. The speakers included Bill Joy, Sam Leffler, Steve Bourne and Andy Tanenbaum.
It was a hugely memorable event, including Prof. Tanenbaum's presentation. We were paying AT&T $200 or so for each Unix license. Not a huge deal for a $100,000 VAX system. But, even then, many of us could see a future where Unix or something like it would run on countless devices, including cars and washing machines. In fact, when I worked for AT&T in 1984 (yes, I know, it was "a learning experience"), I was pitching exactly that to OEMs. It was clear that something cheap or free would be required. So, back in 1983, when dinosaurs ruled the Earth, Prof. Tanenbaum gave us all the seed of a thought that free (as in beer) software could change the world.
As an aside, his presentation was a little hard to follow, but worth the effort, because his English wasn't that great. A Dutch guy sitting next to me said that his Dutch was pretty sketchy, too. I have no means to verify this but, if true, he would join a small group of my friends and acquaintances who don't speak any (human) language well. They're all engineers :-).
I also learned that, despite Bonn being largely flooded because of heavy rains, nothing stops a Unix conference, and that the "Geoffnet" signs I saw all over the place weren't a promotion for a new network stack, but meant "Open" in German.
Anyone else laugh themselves stupid at some of the predictions of the future in those posts? The idea that x86 would go away and GNU/Hurd would supplant Linux...
Predicting the future is REALLY hard.
Non impediti ratione cogitationus.
Really, his books are quite good, I used his the operating systems book in my undergraduate classes. I honestly found reading his book more productive than going to the classes.
Actually, the "official" English translation is "VU University Amsterdam".
So...Dr. Tannenbaum's other project is Electoral-vote.com [electoral-vote.com] (2 [wikipedia.org] ), an election prediction site (and one of the first). Any clue what's going to happen to that?
Andy Tanenbaum may be retiring. But there are two things he will never abandon as long as he is physically able: computer systems and politics.
It's interesting noting that QNX handles messages much better than Microkernels. I had access to the source code at 6.3 and now it's closed.
Unless you go and learn through that path, you won't be able to win the argument.
I did a CS Under-grad in India (late 80s, early 90s) and Prof. Tanenbaum's book was the prescribed textbook for the OS course. Since most of us students couldn't afford the book (or many other books, for that matter), the local photocopy shop had brisk business selling us boot-legged copies. I remember the bound version of the photo-copied book was in landscape mode, since they photocopied two pages at time on to one A4 sheet.!
Not only does he use polling data to do a good job of predicting the races and the control of the US Senate/House (his track record here and a comparison of his model to Nate Silver), but he has, IMHO, excellent explanations of how the campaign managers are thinking and the likely impact of political news.
It is surprising to me that being located in Europe that he 1) cares and 2) is so wired into the US political scene. I hope he continues.
In reality, both were correct, in their own way. You have to remember the hardware they were using in 1992. Its NOT what you have today. Not by a long shot.
---- Booth was a patriot ----
Tannenbaum wanted his thing to be a learning tool, simple, intentionally small and easy for students to grasp, yet enough to show how things worked. Efficiency wasn't part of the equation, functionality was not a priority. Education was. So the theory that a microkernel based system is theoretically the best way to go was what Tannenbaum professed. Torvalds wanted his system to use a macrokernel. Its a design decision. To mitigate the very large kernel size, Torvalds designed and developed loadable (and dynamically unloadable/reloadable) kernel modules. You get a small kernel size, and none of the message passing overhead of a microkernel. So small memory footprint, and high performance. Tannenbaum didn't want to see innovation in what he considered to be 'old technology' and even though Torvalds got an IEEE Turing Award for his innovation, the theory war flared, and that was more than 15 years ago. Certainly Tannenbaums mailing list shrunk after Torvalds started Linux, and to suggest that Tannenbaum was crestfallen may not have been a far stretch.
The board of regents prefers that you write "VU University Amsterdam" now, just FYI.
During the course of my PhD I had many interactions with AST, including taking some classes he taught, giving presentations he attended, attending presentations he gave, and attending a reception at his house.
He is truly an amazing guy!
He was never afraid to tell you he thought your idea was stupid, as is demonstrated with the Linux / MINIX debates. Andy is very opinionated and an excellent debater. I did manage to win him over in some debates, which stands as a very high achievement indeed and impressed my colleagues. It helps that I was right! ;)
He is very deep into US politics, so don't start him talking about it unless you have some time on your hands.
I wish him all the best in enjoying his photography, politics and grand children! He is passionate about all three.
Interesting that he doesn't list Amoeba among his achievements. I find it far more impressive than Minix.
Ah "Structured Computer Organization" - I remeber it well. 1990 college grad, still doing embedded OS work to this day.
iAPX432->i860->i960
The 960 lasted as a 'developed' product until the mid-late 90s, and had production AFAIK well into the '00s.
The i860 was a 64 bit part that has similiar shortcomings to the later Itanium (assumption of compilers being able to handle optimization of the instructions/data rather than the cpu), but mostly died due to internal politics (It was actually like 2x as fast as the 386 when it was released. And it was used in a *LOT* of equipment at the time, including early 3d rendering modules from Sun and SGI and Co-Processing boards from Hauppauge.
You can wikipedia the rest, but it was a pretty impressive design for it's era that got clobbered by office politics rather than any direct 'inferiority' of the processor. Imagine if we'd all been running 64 bit processors back in the 80s/early 90s and where we could be today.
Much like git and a number of other design decisions along the process of Linux's development, loadable modules was a reactionary rather than proactionary development.
You're probably too young to remember, but loadable kernel modules didn't make it in until either 1.3, or 2.1 (I'm pretty sure it was the latter since I remember having to build kernels by hand well into the 2.x days) Additionally, while not documented as such, you had to include 'stub modules' for every subsystem you wanted to be able to later compile modules for without reinstalling/rebooting the kernel. Additionally, many modules forced static values during build (didn't support plug 'n play) and thus required not only to be built on a per-system basis, but also configured for the specific hardware you had in your system. Need two of those cards in your system? Too bad.
Many of the drivers didn't get fixed until 2.3/2.4 and it wasn't until 2.5/2.6 that most of them became truly 'plug and play' in the manner most people associate with them today.
At the time, win9x was, on many systems, far more modular and robust in regards to handling changes in the system configuration, assuming you already had the proper drivers installed. (Linux was often better if you had supported/generic hardware for switching around, but only if lilo didn't break due to your previous bios being chs or lba and the new system being different, and didn't mind wasting a large portion of your ram at the time for the kernel and ALL drivers, which at the time could not be unloaded. I don't remember specifics but it was 8-22 megs of ram, which took forever to load and cut your usable memory size anywhere from a third to less than half.)
Andy Tannenbaum is an American, born an raised in the USA. Given that he should know some English, or a local variant thereof
The haters and trolls notwithstanding, Minix was a worthy accomplishment; and may yet prove more important in the future than first thought, given Red Hat's ongoing destruction of Linux.
Professor Tanenbaum is a great man; and truthfully, I have always wished that Linus Torvalds had been kinder to him. Not all of us are necessarily meant to stand fully in the spotlight, and although perhaps both history and the debates proved Linus right, it would not have cost anything to allow the Professor to keep his dignity.
May he have as much time with his grandchildren as he wishes; and when the time comes, an easy and joyous passing.
Virtualization OSs are not microkernels. Microkernels require the bulk of everything to go into userland, such as device drivers. In Hypervisors, OTOH, device drivers and a whole lot of other common resources that OSs use are packed in the kernel.
Plan 9 is an academic project? I thought it was an internal research project of Lucent/Bell Labs
Oh the adverbs, they hurtses, they hurtses