View previous topic :: View next topic |
Author |
Message |
meekjt n00b
Joined: 22 Sep 2002 Posts: 27
|
Posted: Thu Dec 19, 2002 7:13 am Post subject: HURD or Mach kernel for Gentoo? |
|
|
Have any developers thought about porting a micro-kernel such as Mach or HURD to Gentoo?
Its just a wild idea and I wouldn't know the first thing to do to start, as I'm just a user. From what I've read, it seams a micro-kernel could have some benefits for a desktop system. I recently tried out Apples Darwin for x86 and really liked it for what its worth.
What do you think? |
|
Back to top |
|
|
Evangelion Veteran
Joined: 31 May 2002 Posts: 1087 Location: Helsinki, Finland
|
Posted: Thu Dec 19, 2002 8:11 am Post subject: Re: HURD or Mach kernel for Gentoo? |
|
|
meekjt wrote: | Have any developers thought about porting a micro-kernel such as Mach or HURD to Gentoo? |
Well, looking how "fast" HURD is progressing, the decision whether to use HURD-kernel or not is still about 8-10 years away . |
|
Back to top |
|
|
qnx l33t
Joined: 25 Jun 2002 Posts: 638 Location: Göteborg, Sweden
|
Posted: Thu Dec 19, 2002 8:49 am Post subject: |
|
|
I don't even understand why GNU is developing their own kernel.... It will be _very_ hard for them to establish on the market, even server market, where Linux has grown strong and big now. Why are they trying to develop it??? _________________ Registred Linux user #191143!
Abit NF7-S rev. 2.00 (BIOS v. 2.7)
AMD AthlonXP 2500+ (Barton)
PATA Seagate ST3120022A
SATA Seagate ST3200822AS & Silicon Image 3112 chipset
Gentoo Linux |
|
Back to top |
|
|
charlieg Advocate
Joined: 30 Jul 2002 Posts: 2149 Location: Manchester UK
|
Posted: Thu Dec 19, 2002 10:52 am Post subject: Don't you know? |
|
|
GNU is about ideals... and GNU can't rely on a kernel that, although GPL, isn't under their control.
I hear it's pretty usable but awfully slow when compared to Linux. _________________ Want Free games?
Free Gamer - open source games list & commentary
Open source web-enabled rich UI platform: Vexi |
|
Back to top |
|
|
plate Bodhisattva
Joined: 25 Jul 2002 Posts: 1663 Location: Berlin
|
Posted: Thu Dec 19, 2002 11:11 am Post subject: |
|
|
Behold the sudden surge in interest for the HURD! No doubt that's because we've all been waiting for it since 20 years. Seriously, the HURD isn't all that bad, but of course you'd have to go and build a different system around it. You can't just 'emerge system' and compile a HURD kernel to run with it. Debian has accepted the full workload of building a distribution around the HURD since many years, but it's probably not the kind of thing you'd recommend a fledgling major like Gentoo at this point in time... I don't see anybody overflowing with enthusiasm for this project yet, but who knows, maybe around Gentoo 3.x? It's not as if the HURD would be in any particular hurry. Here's a thread with a very good discussion of the underlying issues, by the way. I completely fail to see why a microkernel architecture like the HURD or OS X would be "awfully slow", though. |
|
Back to top |
|
|
duff Guru
Joined: 19 Jun 2002 Posts: 466 Location: Clemson, SC
|
Posted: Thu Dec 19, 2002 6:04 pm Post subject: |
|
|
qnx wrote: | I don't even understand why GNU is developing their own kernel.... It will be _very_ hard for them to establish on the market, even server market, where Linux has grown strong and big now. Why are they trying to develop it??? |
And I can't believe some people are wasting their time developing new Linux distrobutions when Redhat and Debian are all you'll ever need...oh wait. |
|
Back to top |
|
|
absinthe Retired Dev
Joined: 06 Oct 2002 Posts: 111 Location: San Francisco, CA, USA
|
Posted: Fri Dec 20, 2002 4:12 am Post subject: |
|
|
microkernels are a Good Idea. unfortunately, we're just not there yet... as much as I'd love to be.
Debian has a Hurd port going. As well as one built around FreeBSD...
I would like to see Gentoo take a stab at using the FreeBSD kernel. The VM in particular on FreeBSD is highly evolved and well regarded. As the new SMP code in FreeBSD 5.0 and consequent GCC support for it matures, I think it could be a real nice alternative to the Linux kernel, particularly for systems that would run as servers... FreeBSD has proven itself over the years in that area running Yahoo and Hotmail...
Just a thought. |
|
Back to top |
|
|
plate Bodhisattva
Joined: 25 Jul 2002 Posts: 1663 Location: Berlin
|
Posted: Fri Dec 20, 2002 10:01 am Post subject: |
|
|
Hey, thanks, I didn't know about the Debian *BSD ports. Charming explanation as to why they did it, too:
Ladies and Gentoomen, I believe we have got ourselves a mission here: Gentoo BSD, with portage instead of ports... |
|
Back to top |
|
|
absinthe Retired Dev
Joined: 06 Oct 2002 Posts: 111 Location: San Francisco, CA, USA
|
Posted: Fri Dec 20, 2002 5:34 pm Post subject: |
|
|
plate wrote: | Hey, thanks, I didn't know about the Debian *BSD ports. |
Yeah, it's not something you hear about much -- since it's still under development, but I always thought it'd be a great idea. NetBSD has already proven itself the most portable *nix. FreeBSD still has the best VM and scheduler among free *nix, or at least anecdotally it seems that way.
plate wrote: | Ladies and Gentoomen, I believe we have got ourselves a mission here: Gentoo BSD, with portage instead of ports... |
I'd help out with that.
Open questions: bsd libc or glibc or newlib?
freebsd or netbsd kernel?
fork the kernel or maintain a patchset? (i'd suggest the latter) |
|
Back to top |
|
|
absinthe Retired Dev
Joined: 06 Oct 2002 Posts: 111 Location: San Francisco, CA, USA
|
Posted: Fri Dec 20, 2002 5:48 pm Post subject: |
|
|
From a high-level it's not hard to do. The implementations of stuff like procps and network will be different, so some things you will naturally need to stick with BSD userland stuff on.
But if we could get an ISO built that has the base system on it, at least enough to create the filesystems, basic binutils, and Portage -- so that people could install a system on x86 and sync up with the Portage tree (or perhaps a separate branch) -- that would be a start. Then it would just be a matter of hacking away at ebuilds for stuff until things worked.
That's the bottom line, you just need to get the system up with Portage stable first and then everything else should follow... |
|
Back to top |
|
|
Woland Apprentice
Joined: 02 Aug 2002 Posts: 248 Location: Russian Jack, Alaska
|
Posted: Wed Dec 25, 2002 4:10 am Post subject: |
|
|
Microarchetexture kernels are a Bad Thing. I know because the The Good Book told me so. So far be it for me to exhibit any sort of Hurd mentality. Besides, one really will be forced to call it the Gentoo/GNU/Hurd and in the end we might just all slash ourselves silly.
I can say one thing that I wouldn't want to see, and that is a Gentoo BSD with BSD style init. Now BSD style inits may just be fabuolus for some people, but I find them a cumbersome pain, and besides, Gentoo is justifiably famous for its fantastic init system.
As to Net vs Free, that is a difficult question. On the one hand Free seems to be stable as a rock and very scalable, on the other hand, with NetBSD I can finally run Gentoo on my toaster.
Last edited by Woland on Mon Mar 17, 2003 7:09 pm; edited 1 time in total |
|
Back to top |
|
|
absinthe Retired Dev
Joined: 06 Oct 2002 Posts: 111 Location: San Francisco, CA, USA
|
Posted: Thu Dec 26, 2002 1:23 pm Post subject: |
|
|
Woland wrote: | Microarchetexture kernels are a Bad Thing. I know because the The Good Book told me so. So far be it for me to exhibit any sort of Hurd mentality. Besides, one really will be forced to call it the Gentoo/GNU/Hurd and in the end we might just all slash ourselves silly.
I can say one thing that I wouldn't want to see, and that is a Gentoo BSD with BSD style init. Now BSD style inits may just be fabuolus for some people, but I find them a cumbersome pain, and besides, Gentoo is justifiably famous for its fantastic init system.
As to Net vs Free, that is a difficult question. On the one hand Free seems to be stable as a rock and very scalable, on the other hand, with NetBSD I can finally run Gentoo on my toaster. |
I'm not about to go spending my time explaining the benefits of a microkernel to someone who it is apparently wasted on.
Re: BSD init. In theory, if portage evolves as it is possible, people can use whichever style of init they prefer at some juncture.
Last edited by absinthe on Thu Dec 26, 2002 4:18 pm; edited 1 time in total |
|
Back to top |
|
|
aja l33t
Joined: 26 Aug 2002 Posts: 705 Location: Edmonton, Canada
|
Posted: Thu Dec 26, 2002 1:48 pm Post subject: |
|
|
Wow! Microkernel vs. Monolithic. That sure takes me back. Remember Linus' famous 1992 flame?
Now, time for the important unanswerable questions - emacs or vi? |
|
Back to top |
|
|
absinthe Retired Dev
Joined: 06 Oct 2002 Posts: 111 Location: San Francisco, CA, USA
|
Posted: Thu Dec 26, 2002 4:32 pm Post subject: |
|
|
I think you will note that Linus agreed that microkernels are a better design for almost all reasons, but he said the difference was, Linux is available now. Which of course, he was right at the time because the GNU HURD is still vaporware. He's still right today. Linux wins on the bullet points of what it provides today, but that doesn't mean it's a sane architecture.
But HURD aside, I don't think you need to look any further than Mac OS X to see that microkernels as a technology aren't vaporware. That's based on Mach, which is a microkernel. Interestingly, that was in turn derived from 4.2/4.3BSD at CMU.
Whether HURD is finished or not, doesn't mean microkernels are any less viable. Compared to working on a microkernel, improving a monolithic kernel is a bit like stacking band-aids.
Eventually, any prolonged use of the current Linux kernel design is going to be like polishing a turd. You could keep on doing that, but why? Ideally everyone would move on with something that makes a bit more sense long-term. |
|
Back to top |
|
|
gsfgf Veteran
Joined: 08 May 2002 Posts: 1266
|
Posted: Thu Dec 26, 2002 6:42 pm Post subject: |
|
|
is windoze a microkernel? _________________ Aim:gsfgf0 |
|
Back to top |
|
|
Peripatetic n00b
Joined: 17 Dec 2002 Posts: 10 Location: Florida
|
Posted: Thu Dec 26, 2002 10:04 pm Post subject: Monolithic v. Microkernel, CISC v. RISC, and C v. Everything |
|
|
absinthe wrote: | Eventually, any prolonged use of the current Linux kernel design is going to be like polishing a turd. You could keep on doing that, but why? Ideally everyone would move on with something that makes a bit more sense long-term. |
I hope Linus will never decide to dump the kernel and start from scratch. Fortunately, if that ever happens, people can just fork it and keep it going. I would have thought that we would have realized by now that elegance of design (which is really an illusion anyway) and other technical advantages are but a small part of what determines whether a technology is embraced by millions or destined for the margins.
I think Linux will evolve not by radical changes in design (btw, microkernels are old technology by now, too. Exokernels, among other things, are now the new new thing. And who knows what else will be coming up the OS horizon. The only sure thing is that 30 years from now IBM will still be making good profits on their mainframe contracts.) but rather by absorbing what is good about microkernels into its current design, much as x86 has absorbed most of the design advantages of RISC over the years.
If you look around, we are all still using technology that was supposedly obsolete 20+ years ago. When Andy Tenenbaum (I do recommend his books if you are interested in OSes) said that Linux was obsolete in 1991, he was absolutely right. But guess what, it doesn't matter. What makes or breaks a technology (or any product) is how many people use it. Linux has users and for that it's worth polishing. Anyway, you'd be surprised just how much shinier a polished turd is than an unpolished gem (or whatever else). |
|
Back to top |
|
|
moosesocks n00b
Joined: 09 Dec 2002 Posts: 36
|
Posted: Fri Dec 27, 2002 12:21 am Post subject: |
|
|
gsfgf wrote: | is windoze a microkernel? |
Short answer: Yes
Long answer: All the NT based OSes are microkernels, and win9x is in a league of it's own (I don't think DOS can even be considered a kernel...) |
|
Back to top |
|
|
aja l33t
Joined: 26 Aug 2002 Posts: 705 Location: Edmonton, Canada
|
Posted: Fri Dec 27, 2002 12:52 am Post subject: |
|
|
absinthe wrote: |
Eventually, any prolonged use of the current Linux kernel design is going to be like polishing a turd. |
Warning: The following is a response to an obvious troll. Those with better things to do should ignore...
Absinthe, as I tried to imply in my faint attempt at a humourous post above, you can argue cogently for or against monolithic kernels, because each has advantages and disadvantages (as do exokernels, as one of our colleagues has helpfully shared). If you don't believe that, you're not even wrong.
Even Tanenbaum, the great champion of microkernel architecture, has backed away from an absolutist position on it. (Don't believe me - check out his "Modern Operating Systems", 2nd ed, pp 62f). Tanenbaum has good things to say about the Exokernel idea, as well as owning to the fact that a pure microkernel architecture is "not completely realistic". Although he does still assault pure Monolithic architecture, Linux hasn't met that description since the addition of modules. In addition, among the things that Linus would (I am sure) like to take back about what he said in his (now a decade old) Minix flame is the statements that microkernels were necessarily better.
Now that I'm done with the argumentum ad authoritatum, we should address your 'polishing a turd' motif. That might be vaguely appropriate if we were still patching Linux 1.0. We're not. There have been numerous stable releases since the early days, each one involving significant (and often major) rewrites of the kernel code. And here's the important bit - those rewrites were based on real-world testing and application of the theoretical concepts by thousands (or hundreds-of-thousands) of actual users on actual kernels, not on abstract theorizing about vapourware kernels.
As we look forward to the 2.6 stable release (dear God, please let it be soon) we're going to to see Linux move onto the stage as a real, tested, cutting-edge platform. There's a VM rewrite (esp. the support for large memory), new scheduler code (tasklets!), multithreading improvements, modifications to the devfs, incorporation of new filesystems, overhaul of some of the device support, etc. Linux isn't an old idea getting legs, its the cutting edge of REAL kernel development - not theory.
Polishing a turd? Geez - that's one shiny turd. |
|
Back to top |
|
|
pilla Bodhisattva
Joined: 07 Aug 2002 Posts: 7729 Location: Underworld
|
Posted: Fri Dec 27, 2002 2:02 am Post subject: |
|
|
Warning: behave yourselves or I will call your mothers (and lock the thread).
aja wrote: | absinthe wrote: |
Eventually, any prolonged use of the current Linux kernel design is going to be like polishing a turd. |
Warning: The following is a response to an obvious troll. Those with better things to do should ignore...
|
_________________ "I'm just very selective about the reality I choose to accept." -- Calvin |
|
Back to top |
|
|
absinthe Retired Dev
Joined: 06 Oct 2002 Posts: 111 Location: San Francisco, CA, USA
|
Posted: Fri Dec 27, 2002 9:37 pm Post subject: |
|
|
aja:
Those of us interested in microkernels were having a discussion here (you will note that it's the subject of this thread) -- and a reasonable one at that ... until you decided to chime in and opine on what you feel about microkernels. If you hate microkernels, feel free to open a new thread and flame away.
We were also talking about strapping on other monolithic kernels as well in addition to the HURD or something like it.
Yes, I do believe that the continued use of the present linux kernel design is harmful in the long run. There's a lot of good ideas and code inside of that kernel, but that architecture is less than ideal.
Does that mean that people won't use it for another 10 years? Absolutely not. It may be with us for a very long time. And it may be very successful that way.
However, Law of Diminishing Returns... I'd rather see the technology evolve. Just because we can continue with the current Linux kernel architecture doesn't mean we should. |
|
Back to top |
|
|
plate Bodhisattva
Joined: 25 Jul 2002 Posts: 1663 Location: Berlin
|
Posted: Sat Dec 28, 2002 3:00 am Post subject: |
|
|
Oh, but finding a new toy doesn't necessarily mean we should dump the old one. This is what I think a "distribution" is all about (and what I found charming in Debian further up this thread), i.e. alternatives being made available to people who don't hack gcc for a living, and distributed with ease of use and maintenance in mind, no matter how the underlying kernel (or lack thereof) decides to treat the hardware (or abstraction thereof)...
It's all a matter of necessities, there's no panacea that I can see. All those operating systems from DOS to Unununium are just tools, each with a purpose, all with limitations as to what they're good for. Streaming media wasn't a problem until recently, but since it is very much boggling people's minds now, a renewed interest in real-time operating systems is quite understandable. Same with exokernels, totally fascinating stuff. I don't understand a word of most explanations I've read, but if it's a viable alternative to traditional microkernels, why not make it applicable to real-life-scenarios? While we're at it, how about, say, Gentoo XOK?
Yes, I think I may have had too much Glühwein last night. Lads, this is so much more fun than the business I'm in (marketing... ), I should never have left Academia... |
|
Back to top |
|
|
avenj Retired Dev
Joined: 11 Oct 2002 Posts: 495 Location: New Hampshire
|
Posted: Mon Mar 17, 2003 2:10 pm Post subject: |
|
|
I'm actually not very pleased with plate's decision on https://forums.gentoo.org/viewtopic.php?p=245259#245259 (this thread is stale and bogged down with totally unrelated things). Regardless of that, I'm doing real work on Gentoo GNU/Hurd. If anyone else has an actual interest in this, e-mail me (avenj@gentoo.org) or contact me on IRC (avenj on freenode) and I can give you more information about it.
I would prefer not to discuss it too in-depth on the forums because I disagree with the arbitrary shifting of posts just for the sake of shifting posts. Accordingly, I will restrict my future forums posting and will try to stick to significantly more unmoderated forums that provide for user freedom. |
|
Back to top |
|
|
gilesjuk Guru
Joined: 11 Feb 2003 Posts: 412 Location: Staffordshire, UK
|
Posted: Mon Mar 17, 2003 2:52 pm Post subject: |
|
|
moosesocks wrote: | gsfgf wrote: | is windoze a microkernel? |
Short answer: Yes
Long answer: All the NT based OSes are microkernels, and win9x is in a league of it's own (I don't think DOS can even be considered a kernel...) |
AmigaOS was a pre-emptive microkernel long before Windows (1984-1985 approx). _________________ ^<^>^<^>^
G. Jones
-=-=-=-=-=- |
|
Back to top |
|
|
avenj Retired Dev
Joined: 11 Oct 2002 Posts: 495 Location: New Hampshire
|
Posted: Mon Mar 17, 2003 3:05 pm Post subject: |
|
|
For anyone who saw my earlier post about Gentoo GNU/Hurd, I decided to start a new topic in Gentoo Chat for it (this one isn't very relevant anymore).
I apologize to plate for being a little unnecessarily harsh. It was an early-morning overreaction. |
|
Back to top |
|
|
plate Bodhisattva
Joined: 25 Jul 2002 Posts: 1663 Location: Berlin
|
Posted: Mon Mar 17, 2003 3:20 pm Post subject: |
|
|
No harm done. avenj's new thread is here. |
|
Back to top |
|
|
|
|
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
|
|