When is GNU Kernel going to be released?
- Inicie sesión ou rexístrese para enviar comentarios
Just wondering
Mitja
I am a member!
You mean HURD?
https://www.gnu.org/software/hurd/#index5h1
I am a member!
If you mean HURD there is an Arch and Debian release I believe. The hardware support isn't the best though. iirc SATA and MP isn't supported.
The plan is to release with Debian Wheezy by the end of this year or the beginning of 2013.
I am a translator!
I wonder what technical advances Hurd has in comparison to Linux? I know that microkernels are supposed to be better than monolitic ones, but what exactly does it mean?
A microkernel is easier to maintain. A true microkernel has about 6.000 lines of code, compared with 30m of Linux
A microkernel has more stability because it transfers some processes to the user space. If a Linux or Windows driver crashes the whole system crashes because drivers run in kernel mode. A driver crash does not crash the microkernel.
In a microkernel design you can port the same system between different microkernels, e.g Hurd/L4. This allows more flexibility.
In a monolithic kernel if you change anything you have to recompile the whole kernel again and again. In a microkernel you compile kernel and servers separately (my guess)
It can be ported to other architectures with much less effort.
You can read this 6 page introduction to the differentes between the two http://www.vmars.tuwien.ac.at/courses/akti12/journal/04ss/article_04ss_Roch.pdf
Or watch this presentation by Mr. Tanenbaum in youtube http://www.youtube.com/watch?v=bx3KuE7UjGA
I am a translator!
A fascinating presentation, thank you for the link. After watching this it does seem like a good idea to switch to microkernels from monolithic in the long run, just to keep the drivers out of the kernel. But I can't help but think he is overemphasizing the unreliability of existing OSes. Even Windows isn't what it used to be, and GNU/Linux just seems to never crash. Since I've figured it out and can surely say that I don't unintentionally break something, I've had to reboot a computer by reseting maybe 2 times in one year (and I use my machines a lot!).
I am a member!
Honestly, while Hurd might look interesting from a hacker's point of view, it's not worth waiting for, and IMHO there's no real reason to continue development of it. Linux Libre is more than sufficient.
I am a translator!
It might be sufficient for now, that's true, but as far I understand the issue, it gets harder and harder to maintain with each version. Drivers are inside the kernel right now and having one huge program is always more difficult, than having many smaller ones, that interact through some sort of interface.
I know very well, that at this point Linux-libre is very stable and performs very well and what not. But imagine what will happen in, say, five years, when we will have much more drivers in the kernel. It will be like a fragile house of playing cards, if one small thing goes wrong somewhere, you're in very big trouble.
I believe Linus Torvalds made a huge contribution by starting Linux, but he wasn't a professional programmer at that point in time! He was a very talented student, with lots of time at his hands. No wonder he made some design choices that aren't too perfect.
Until now, the advantages of a microkernel are more theoretical than real. Anyway it is true that with processors that may, in the near future, have hundreds of cores, Linux may go through scalability troubles. I doubt the number of drivers is an issue though (do you have any reference?): Linux is monolithic but highly modular too.
Debugging a microkernel is very hard and explains why the HURD is still far from being usable in production. The GNU project did not know that when the technical architecture was chosen.
- Inicie sesión ou rexístrese para enviar comentarios