Microkernels are irrelevant for my day to day computing work

Story: Tanenbaum-Torvalds Debate: Part IITotal Replies: 4
Author Content

May 18, 2006
7:47 AM EDT
I don't really care what kind of kernel is running under my desktop. If it is monolithic and it performs well and supports my hardware, it's dandy. If it is a muK and it performs well and supports my hardware, it's dandy too.

The gist of the matter is that Andy Tanenbaum may have written Minix and maybe muK's are the non-plus-ultra, but as long as there is not a major viable and easy to use muK project on the scale of the monolithic Linux kernel, I don't care how nice muK's are on paper.

Advice to Tanenbaum: Get crackin' and provide a muK that can act as a complete drop-in replacement for Linux. Only then will this debate leave the realm of ifs, buts and "My muK is better than your Linux".

May 18, 2006
12:21 PM EDT
If you read what he has to tell, he doesn't care about muK's, he only cares about an OS which has a longer crash-free period than the life of the hardware.

As far as the Linux replacement is concerned: He's working on that. I plan testing Minix 3.0 for LXer, but it may take a while before I find the time to do so.

May 18, 2006
1:45 PM EDT
Well, so he says hkwint, but he seems to think that will require a muK. I readily accede to his expertise in muKs, but i question his understanding of how these things become mainstream. Many, many superior solutions never become mainstream. To become the major direction of development, there needs to be much more recognition and support. I just don't see that now.

On the other hand I could see this having a major role in Linux, or whatever comes after that, 5 or 10 years down the line. I'll certainly be waiting for that.

May 18, 2006
2:50 PM EDT
It's only when people become really annoyed with the software they currently have, and they know there is a viable alternative, that some other OS may become mainstream.

Since most people don't seem to be annoyed with the current state of Linux, apart from the people behind CoyotOS, Theo de Raadt, Green Hills software and Andy, and since people don't know about the alternative, this won't happen soon.

Just like people weren't very annoyed about Windows 3.11 and didn't know about Linux in '93-'94 maybe, or maybe not. Some superior software never becomes mainstream. Inferno (Vita Nuova's enhanced P9 product) is around for a while and is able to serve commercial purposes at the moment, and you don't hear that much from it. BTW, most people don't grasp the idea of 'formally proven security' (like in CoyotOS), and almost all people consider Linux save, mainly because all Linux people lobby by telling Linux IS save. Linux maybe doesn't suffer from virusus and malware at the moment, but someone wouldn't want to look at the remote holes it had.

When Theo de Raadt first told about Linux being buggy and 'quality of code not being a priority of Linus' I found it very harsh and wish he should kept his mouth shut. Thinking about it more and more, I can't disagree anymore.

I think Andy has a point when speaking about buggy drivers in Linux, you wouldn't want to know how video drivers from a user-friendly mainstream distro like Ubuntu (LiveCD) caused my computer to freeze, and even the "three finger salute" didn't work; I had to reside to the power button. The biggest problem of Linux may be the sheer size of the kernel getting out of hand with all these new drivers, and nobody being able to oversee the kernel. If you look at the growth rate of the size, things may get out of hand sooner than most of us may think.

So, only when people will see there's an alternative, and only if Minix can play DRM'ed content from Hollywood (which won't probably happen) and is able to synchronize with Exchange (silly examples, but you get the idea I'd say) people could make muK mainstream.

It's also the BSD license which kinda 'disables' both *BSD and Minix from becoming mainstream in my opinion, but people could enhance Minix and put it under GPL2/3 as far as I know.

Sadly, a lot of Linux 'fanboys' think Linux is perfect, and would say: "Who is not with Linux is against it". They become the rusted Windows users of the future.

Anyway, this whole discussion aside, I think we should follow Tanenbaum's request, and just test Minix3 to see what it can do, and what it cannot do.

As a Gentoo user, I'd say 6-8 seconds for the whole kernel (and more stuff as I understand) to compile is incredibly fast. To compile some small program like Lynx takes two minutes on my 2.0 ghz AMD 64 for example.

May 18, 2006
2:54 PM EDT
"To become the major direction of development, there needs to be much more recognition and support."

I disagree. To become a major direction it should far surpass Linux. 99% of the people don't even come near the limits of the Linux kernel. Unless a muK gives them a very compelling advantage over what they have, switching is just too much work. It's the same reason why some people stay with Windows. It's the same reason why yEnc is still popular in newsgroups. Having even longer average uptimes is not a compelling reason (since Linux is already capable of very long uptimes). Personally I don't see in what area's a muK could gain such an improvement over Linux that it would tip the scales.

Posting in this forum is limited to members of the group: [ForumMods, SITEADMINS, MEMBERS.]

Becoming a member of LXer is easy and free. Join Us!