Slashdot Mirror
Europe Funds Secure Operating System Research
narramissic writes "A Dutch university has received a $3.3 million grant from the European Research Council to fund 5 more years of work on a Unix-type operating system, called Minix, that aims to be more reliable and secure than either Linux or Windows. The latest grant will enable the three researchers and two programmers on the project to further their research into a making Minix capable of fixing itself when a bug is detected, said Andrew S. Tanenbaum, a computer science professor at Vrije Universiteit. 'It irritates me to no end when software doesn't work,' Tanenbaum said. 'Having to reboot your computer is just a pain. The question is, can you make a system that actually works very well?'"
The question is, can you make a system that actually works very well?
I'm glad someone finally got to asking this question.
I thought Windows was secure. Why not use that? *cough* *cough*
I thought OpenBSD was secure. Why not use that?
You just got troll'd!
.. they want their funding back.
Seriously , I thought minix had been put out to pasture years ago.
Also what are 5 people going to manage that entire corporations and thousands of OSS developers failed to do in the last few decades? Ok , one of them might be the next Alan Turing and surprise us all but I won't hold my breath.
Yes, imagine that: A professor trying to teach students how to implement something new and potentially useful rather than clicking ok in the 'solve my problem' wizard.
How many times is this old chestnut going to be tossed around?
MS tried a microkernel with NT and its HAL. It didn't really work very well. Most Unix varients don't even bother to try.
I think you are right at the moment. I am not sure that you will still be right when processors are 256-core or greater. I think that at some point the overhead of microkernals will be made up for by utilisation of greater parallelisation.
Has anyone noticed how more and more stuff gets moved from the Linux kernel into user space these days; FUSE is a good example. History may record that, broadly speaking, Tanenbaum was corrent and Torvalds was not. Anyway, I assume you are saying that since Linux has been so much more successful than Minix, we must listen to Torvalds and ignore Tanenbaum? On that basis, we should listen to Gates and ignore Torvalds!
No, but dividing things into smaller pieces makes it easier to fix those pieces in isolation. It's easier for a microkernel system to be self-healing because of that isolation.
This is not an amazing revelation. We've known about the idea of isolating changes since the invention of the sub-routine. The reason microkernels have always been relegated to second-best is that they require more context switching than a regular monolithic kernel. The tradeoff between "fast enough" and "reliable enough" has for some time now favoured "fast enough".
But that's changing -- people's computers are getting plenty fast. The 10-15% slowdown Tanenbaum claims for Minix3 is less of a drag than, say, an anti-virus program and could serve to more effectively prevent viruses into the bargain.
People who say microkernels are passe forget our industry is not set in stone. Priorities change and technologies change with them. In the last 10 years performance has become progressively less important than reducing bugs or speed of development. Microkernels have lots to offer in such a world.
Classical Liberalism: All your base are belong to you.
That's a rather ignorant viewpoint.
Tanenbaum argued for greater modularity and really that's no bad thing, his arguments were pretty solid theoretically. But as we all know, just as the most beautiful, maintainable, stable software designs are sacrificed in business for something that works now even if it has it's flaws, Linux was available, easy to use and just worked the way people wanted, that didn't mean it was inherently better in theory or that Tanenbaum is wrong anymore than it means Windows is a vastly superior OS to Linux and MacOS X simply because it has such a massively larger user base.
Basing your view on Tanenbaum's one comment towards Torvalds is also rather ignorant, throughout the discussion you're referring to, Tanenbaum was well composed and formed coherent arguments, whilst Torvalds at times acted like your average troll.
You see, the very fact Windows is far and away the most popular OS followed by MacOS X followed by Linux is evidence enough that popularity means nothing in terms of the actual quality of an OS, it merely shows which played the business game better.
Tanenbaum is worth listening to, his ideas and justifications included in that 17 year old discussion you mention aren't wrong even if his predictions on the future of computing were. This is a man who understands the theory of how to make a better OS more so than most people do, and yes possibly even more so than Torvalds. The problem is that he's a theoretical guy, so whilst his proposals may be better, they may not be practical at the time they're announced or he simply may not have the time to dedicate to proving their practicality. If they're not practical at the time he proposes them though that doesn't mean they'll never be practical as changes in computing architecture or even raw computing power may make them practical.
Hopefully he'll put this funding to good use and it'll help provide him the time and resources he needs to take his ideas beyond mere theory and he'll be able to backup his theories with actual working demonstrations rather than just arguments now. You can be a Torvalds fanboy all you want but Tanenbaum and Torvalds are two different people - Tanenbaum is someone who comes up with theoretical new concepts, Torvalds is someone who takes existing concepts and implements them well. Both have their strengths, but writing one or the other off is foolish when both have a lot to offer.
Minix did get an reputation of being unstable some 20 years ago, but of course - much have happened since then.
The one thing that hasn't changed though is that Minix is still just a toy system that's meant to be poked at in schools and that nobody actually uses (yes I know about the 3 rabid Minix users, they probably run AmigaOS too).
Oh, wait, now it finally supports X11 (woohoo !). Wait, has it got a mouse driver too ?
However Minix3 *does* feature support for "Over 650 UNIX programs" (such as man, mkdir and ps). *650* ! It's like 130 × 5 ! Think about it !
Granted, starting from a small scale system such as Minix is certainly simpler than with a much more mainstream OS such as one of the BSDs or Linux but even if anything comes out of the project, it won't ever gain even "niche" status. More people must be running Plan9 or Inferno.
The whole idea is utterly futile, except possibly if the code or the concepts can be reused with another system later on.
May contain traces of nut.
Made from the freshest electrons.
Dropping C... for what exactly? We're not talking application level security. We're talking kernel level. That means talking to the bare metal. Even if you implement a microkernel with userspace modules for everything, and with those modules written in something more robust than C, that last crucial bit of code that is the microkernel itself is probably going to end up being written in C with ASM snippets, simply because at some point you need to explicitly state what the hardware is doing.
Yes, most developers moved to Linux and stopped writing that pesky, unstable software that anyone actually uses.
Keeping a kernel that is 10 years behind the leading edge in file systems or communications, especially by kicking it all out of the kernel and saying "Naah-naah-naah! Not my problem!!!!" is like having a very secure car that doesn't have a reverse gear, seats, or door handles. It certainly helps contribute to stability. But the associated software to handle USB, firewire, packet filtering, or network file systems just isn't up to speed.
And with almost everything going to interpreter environments today (Python, Ruby, Java, .Net), there's a better argument that building a JIT as a kernel component and that the message passing overhead is less of an issue.
Let me get this right, after stating that the advantage of a microkernel lies in the much smaller size in LOCs, you just argued that adding JIT compiler to the microkernel itself is a good idea?
Part of the idea behind a microkernel is that you only need to prove correctness for a small amount of code. The other part is that, when you want to add features, you only need to prove the features you want work correctly. So, instead of proving that each driver works correctly (which, for most environments where this stuff really matters, only needs to be done for a "handful" of drivers), you just upped the ante and said "prove the whole JIT compiler works correctly". And the "message passing overhead" pales in comparison with a poorly-optimized JITC, which is what you get if you want to keep TLOC count low.
[citation needed]
All these years after the Tenenbaum-Torvalds debate Linus admitted his prof was right? You'd think that would have been in the news somewhere.
That kind of car is actually build regularly by most car manufacturers. The amount of money spend on those cars is often in the same ballpark, or even more. They call it concept cars, and they generally also only explore certain aspects of cars while happily ignoring others.
Thats is not going to be your car for daily use. Minix probably isn't going to be you daily OS anytime soon either, but that no reason not to spend research money on it. The IT industry could do with some more proper research instead of just reinventing the same weels (but this type using XML and HTTP!) all the time.
"Some researcher"!? The guy (prof. Tanenbaum) wrote the original Minix, which was the OS that inspired Linux and hosted the early stage of its development. Also see: http://en.wikipedia.org/wiki/Tanenbaum-Torvalds_debate