Announcement

Collapse
No announcement yet.

GNU Hurd 0.5, GNU Mach 1.4 Released

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #11
    13. Joeboy: "Are you still at all optimistic about HURD, or would you agree that
    that ship has sailed?"

    RMS: "I am not very optimistic about the GNU HURD. It makes some progress,
    but to be really superior it would require solving a lot of deep
    problems. However, mainly what I think about the HURD is that
    finishing it is not crucial.

    When we started the HURD, it was for a simple reason. The GNU system
    needed a kernel, and no usable free kernel existed. We set out to
    write one.

    That problem does not exist today. Linux works ok as a kernel.

    The main shortcoming of Linux is at the level of device support. The
    obstacle there isn't a lack of ability among Linux developers, but
    rather the use of devices whose specs are secret.

    Finishing the HURD would not advance us at all in supporting these
    devices. The work that is needed is at the driver and firmware level.
    That's why our high priority task list includes items relating to free
    drivers, but not the HURD.

    That's also why fsf.org has hardware resource pages. Your help in
    updating them would strengthen us in this important battle.

    Sure, it would be nice to see a GNU kernel succeed -- but there are
    many successful GNU packages, so having one more is not crucial."

    Source: http://blog.reddit.com/2010/07/rms-ama.html


    Even Stallman sounded ready to move on, and that interview was given 3 years ago.

    Comment


    • #12
      Originally posted by doom_Oo7 View Post
      Did you try Debian GNU/Hurd ?
      Yes, and doesn't start on my apu hehehe

      Comment


      • #13
        Originally posted by blackiwid View Post
        the biggest problem I think is driver support, but maybe I am wrong here.
        But there are some features, that are not there in linux, really nice features. So the question is is that a design thing or why is linux not doing that.

        So you could say after 30 years linux cant do stuff that hurd can do, so its a failure
        Mostly, I think it comes down to one killer feature - Linux can actually boot on modern hardware, not just in a VM. Unlike Hurd, which considers SATA support a novelty, doesn't support large disks, multi-core processors, or (being 32-bit only) more than a couple of GB of memory.

        Seriously, some of those features might sound nice, but well... what's the value of all that wonderful theory and purity of design, if can't actually do anything useful?

        Comment


        • #14
          Originally posted by mrugiero View Post
          Optimization and microkernels aren't really hand in hand, you know? A microkernel, by its own architecture, implies a lot of extra calls and inter-process communication. It's its nature. It's focus is simply not speed, but reliability and security. Your system will probably be a lot more stable using a microkernel, but an optimized microkernel + optimized servers will always be slower than an optimized monolithic kernel providing the exact same features (and you could limit to that exact same features by compiling it with only those you need). For uptime critical applications, go with the micro kernel. For time critical, a monolithic kernel might be wiser.
          if thats true that performance and reliability is the only thing that matters, why do we talk about the very slow btrfs file system and not use the faster ext4 filesystems forever, why will every mainstream distro switch in the next 3 years to btrfs?

          Because pure speed doesnt matter that much anymore. Developer Time and bug-freeness matters, and a micro-kernel is better for that, and features matters again, I seen things a few years ago in hurd that today dont exist in linux really great stuff, not something nobody would want. something like "ln -s ssh://server//bla.txt local-bla.txt" (replace that with ftp or something else).

          Why is something so usefull and obvious that u want to have such stuff not implemented in linux? Maybe you can say me different reason than the microkernel structure why its not there, but still its not there.


          and that something like that works has to do with modularity and a microkernel is more mudular, so u could make maybe a monolyth kernel also more micro-ish or make something hybrid out of linux, but the only thing why linux is good is because it is the projekt with most developer years invested ever. (I am talking about the kernel)

          So lets think X-org would have the same amount of developers the last 10 years than linux had, would we also now talk that the X11 design is perfekt or something different makes no sense because we worked good enough around the problems with the design?
          Last edited by blackiwid; 29 September 2013, 07:54 AM.

          Comment


          • #15
            Originally posted by blackiwid View Post
            the biggest problem I think is driver support, but maybe I am wrong here.
            No, is simply not done..... and mostly likely like never be.

            Comment


            • #16
              Originally posted by blackiwid View Post
              if thats true that performance and reliability is the only thing that matters, why do we talk about the very slow btrfs file system and not use the faster ext4 filesystems forever, why will every mainstream distro switch in the next 3 years to btrfs?
              Performance and reliability aren't the only things that matter. There are also features and hardware and software support. Hurd lacks more than Linux in the latter three, and by design will lack more than Linux on the performance side. So, again, the point where it could win would be reliability. I never put either performance or reliability over each other, but stated the uses of each. Also, you are mixing OS stability with data integrity. Most people can have a hang once in a while, but most people can't afford losing data. That's why every distro will eventually switch to btrfs. There's another thing, which is if the reliability benefits outweighs the performance drawback. btrfs is at least half as fast as ext4, IIRC. That's probably not the case for microkernels against monolithic kernels, because almost everything will require a switch between user space and kernel space. The performance drawback is likely to be big.

              Because pure speed doesnt matter that much anymore.
              So you know better than the specific users about what matters to them?

              Developer Time and bug-freeness matters, and a micro-kernel is better for that,
              Never said they don't matter. I said they aren't the priority in certain use cases and that they are in another. Read before answering, please.

              and features matters again, I seen things a few years ago in hurd that today dont exist in linux really great stuff, not something nobody would want. something like "ln -s ssh://server//bla.txt local-bla.txt" (replace that with ftp or something else).
              Features matters, like supporting the hardware where you need the OS to run. And I don't mean drivers, but 64-bit or multi-core support, for example. Features matters, and you need to have priorities for them. Being able to make such symbolic link to a server via SSH is far less important than being able to boot, specially when you can overcome this by formatting and mounting the server's filesystem as NFS.
              Also, micro kernels don't lead to bug freeness, and lead to more work in developer time, because you can't just ignore the fact you need to write the servers, that would be shortsighted, don't you think? You only isolate the bugs from the kernel, that's why the system becomes more reliable, but at the same time you get more complex drivers because of the extra calls you need to handle, you get more complex modules because of the same reason, and all of that means the developer will have a harder time writing and will be more prone to bugs. The only benefit is the code will be cleaner, better isolated, and that those bugs will not bring your system down, because they are all outside the kernel and can be restarted without a reboot.

              Why is something so usefull and obvious that u want to have such stuff not implemented in linux?

              Maybe you can say me different reason than the microkernel structure why its not there, but still its not there.
              I don't see how that feature is something that I want to have. Also, I will say it again: how file systems are handled is not inherent to micro kernels or monolithic kernels. You could treat everything as a file on Linux, but evidently nobody considered going further in that approach a killing feature worth coding it.

              and that something like that works has to do with modularity and a microkernel is more mudular, so u could make maybe a monolyth kernel also more micro-ish or make something hybrid out of linux, but the only thing why linux is good is because it is the projekt with most developer years invested ever. (I am talking about the kernel)
              IIRC, Hurd started way before than Linux. So, by your criteria, it shouldn't have even had any attention, because Hurd already had more development on top of it than Linux. Maybe Linux is just not that bad and unusable, or reliability is not the only thing that matters, or maybe just the micro kernel thing is really beautiful in theory but not as much IRL. I don't know of a single commercial OS using a pure micro kernel approach. Also, I already told you a way to go to a micro kernel approach but starting with a feature full kernel, and you simply ignored it.
              And more, I don't see the relation you state between using a micro kernel and the support for the approach "everything is *actually* a file", nor why don't you consider monolithic kernels (which aren't really pure monoliths nowadays) not modular. Support wise, the modularity you need is in the organization of the code, mostly, and you can keep that in monolithic kernels. Runtime modularity is more related to reliability (being able to stop core OS functionality, if behaving wrongly, without rebooting).

              So lets think X-org would have the same amount of developers the last 10 years than linux had, would we also now talk that the X11 design is perfekt or something different makes no sense because we worked good enough around the problems with the design?
              First, I never said Linux design is perfect, I know it is not. You are reading it all in an utterly moronic way, and it's starting to annoy me. Second, X.org had a lot of developers working on it through the last 10 years. Third, making no sense and having no point are different things, and I only stated the latter. Fourth, there are problems with micro kernel design, too. Fifth, most of X.org design flaws weren't flaws when it was designed, so I'd rather call it obsolete than calling it badly designed.
              And again, if you want a micro kernel approach you don't need to start over from zero, you could as well write the micro kernel, turn the Linux kernel into a server, so you have a usable kernel, and gradually turn the subsystems and drivers into servers, and gradually crop them from the huge Linux server. Even while it sounds more complex, I think that's a saner approach, considering how you won't get any feedback from users that needs things done otherwise.
              Last edited by mrugiero; 29 September 2013, 12:33 PM.

              Comment


              • #17
                Originally posted by mrugiero View Post
                First, I never said Linux design is perfect, I know it is not. You are reading it all in an utterly moronic way, and it's starting to annoy me. Second, X.org had a lot of developers working on it through the last 10 years. Third, making no sense and having no point are different things, and I only stated the latter. Fourth, there are problems with micro kernel design, too. Fifth, most of X.org design flaws weren't flaws when it was designed, so I'd rather call it obsolete than calling it badly designed.
                And again, if you want a micro kernel approach you don't need to start over from zero, you could as well write the micro kernel, turn the Linux kernel into a server, so you have a usable kernel, and gradually turn the subsystems and drivers into servers, and gradually crop them from the huge Linux server. Even while it sounds more complex, I think that's a saner approach, considering how you won't get any feedback from users that needs things done otherwise.
                First of all, I dont know why you get so pissed, I am talking about about a technology I did not attack you or something like that. The Fork Linux aproach has some problems, first I dont think its easy to rewrite a monolitic kernel to a micro-kernel, often in such big refactorings its easier to rewrite stuff, but yes maybe it would be better. 2nd I dont want to develop such stuff so it does not really matter how you come to a good micro-kernel, another problem is that the linux kernel is to unfree, so if you start from a freebsd kernel you get patent law suits like hell, with linux u cant fork your code to normal gpl, just to a modified version (v2-only).

                So to go the gnu way would be the best, to get 2 things done, its not very effishent to rewrite a complete kernel to only fix 1 of 2 problems. If you do that work fix both.

                I mean if you would say 3 and later, somebody could fork it if gpl 4 would suck and comply only to version 3 rules so I dont see any disadvantages in that, on the opposite end you cant upload changes in gplv3 if you like your code protected by that. Even GPL2 is bypassed by DEFINE_KERNEL_BLABLA_DO_WHAT_YOU_WANT_ITS_NOT_GPL symbols.

                Android is a piece of shit, and if we talk about speed android is the best example that for most people (most people arent linux network admins, but most have a smartphone) speed does not matter the java-api makes android so slow and especialy ram-hungry that you need basicly double the memory and double the cpu speed to have the same speed than a apple or microsoft phone needs. That leeds to the need of bigger akkus...


                BTW I never said that NOBODY cares about speed, of course some geeks do, but most people dont (android). btw I dont know about apples smartphone os, but its a hybrit? kernel or something and liek a said 1000x faster than garbage java + linux in android.

                I could even live with a gpl2 + later kernel... If I would make changes they would be released in gpl3 and if the new linus would not pull that, ok I would have no big problems with that, if you comply to gplv3 u get all stuff from gplv2 the companies that dont want to comply to gplv3 just get the hole deal their problem.

                AND one nother thing if lisenses would not matter everybody would switch to bsd because its cleaner and better than linux.
                Last edited by blackiwid; 29 September 2013, 02:36 PM.

                Comment


                • #18
                  Originally posted by blackiwid View Post
                  First of all, I dont know why you get so pissed, I am talking about about a technology I did not attack you or something like that. The Fork Linux aproach has some problems, first I dont think its easy to rewrite a monolitic kernel to a micro-kernel, often in such big refactorings its easier to rewrite stuff, but yes maybe it would be better. 2nd I dont want to develop such stuff so it does not really matter how you come to a good micro-kernel, another problem is that the linux kernel is to unfree, so if you start from a freebsd kernel you get patent law suits like hell, with linux u cant fork your code to normal gpl, just to a modified version (v2-only).

                  So to go the gnu way would be the best, to get 2 things done, its not very effishent to rewrite a complete kernel to only fix 1 of 2 problems. If you do that work fix both.

                  I mean if you would say 3 and later, somebody could fork it if gpl 4 would suck and comply only to version 3 rules so I dont see any disadvantages in that, on the opposite end you cant upload changes in gplv3 if you like your code protected by that. Even GPL2 is bypassed by DEFINE_KERNEL_BLABLA_DO_WHAT_YOU_WANT_ITS_NOT_GPL symbols.

                  Android is a piece of shit, and if we talk about speed android is the best example that for most people (most people arent linux network admins, but most have a smartphone) speed does not matter the java-api makes android so slow and especialy ram-hungry that you need basicly double the memory and double the cpu speed to have the same speed than a apple or microsoft phone needs. That leeds to the need of bigger akkus...


                  BTW I never said that NOBODY cares about speed, of course some geeks do, but most people dont (android). btw I dont know about apples smartphone os, but its a hybrit? kernel or something and liek a said 1000x faster than garbage java + linux in android.

                  I could even live with a gpl2 + later kernel... If I would make changes they would be released in gpl3 and if the new linus would not pull that, ok I would have no big problems with that, if you comply to gplv3 u get all stuff from gplv2 the companies that dont want to comply to gplv3 just get the hole deal their problem.

                  AND one nother thing if lisenses would not matter everybody would switch to bsd because its cleaner and better than linux.
                  I'm going to file this one under the "YOU DON'T KNOW WHAT YOU ARE TALKING ABOUT" category.

                  Comment


                  • #19
                    If you want to flame around go ahead...

                    I wanted to discuss why I think a alternative to linux and bsd is needed and most likely it should be a micro kernel, because linux design is old, yes its mainstream now, its the dominant system now, windows is on its die-bed, if there would be no hardware-windows bundle it would be even worse... but I count smartphones and tablets as pcs and there windows is dead.

                    And that will not change for the next 5-10 years maybe even 20 years. But maybe Hurd is really dead (even Hurd was never = Hurd it was different oses as far as I read it, so which hurd in the future is now dead would be the question) but I just say, linux is about rewriting everything why should the kernel be the only thing where this will never happen?


                    Replace Linux with GPL FOSS Operation systems. or "GNU"
                    Last edited by blackiwid; 29 September 2013, 03:19 PM.

                    Comment


                    • #20
                      Originally posted by blackiwid View Post
                      If you want to flame around go ahead...
                      It's not flaming. You are arguing based on a number of misconceptions. Let me highlight a few for you.

                      2nd I dont want to develop such stuff so it does not really matter how you come to a good micro-kernel, another problem is that the linux kernel is to unfree, so if you start from a freebsd kernel you get patent law suits like hell, with linux u cant fork your code to normal gpl, just to a modified version (v2-only).
                      GPL v2 or v3 or any subsequent versions do not absolve you of the liability of patent violations. I assume that you know the difference between patents and copyright laws in various regions?

                      The assertion that a BSD license makes you more susceptible to patent violation is ludicrous at best.

                      Android is a piece of shit, and if we talk about speed android is the best example that for most people (most people arent linux network admins, but most have a smartphone) speed does not matter the java-api makes android so slow and especialy ram-hungry that you need basicly double the memory and double the cpu speed to have the same speed than a apple or microsoft phone needs. That leeds to the need of bigger akkus...
                      This is all just hot air. The Java apis do not make android slow. The implementation of the Java apis may. What are you basing this on anyway? Do you have any hard truths to back this up?

                      BTW I never said that NOBODY cares about speed, of course some geeks do, but most people dont (android). btw I dont know about apples smartphone os, but its a hybrit? kernel or something and liek a said 1000x faster than garbage java + linux in android.
                      Again, just hot air. Can you provide proof that iOS performs 1000x faster than android for any given work load?

                      AND one nother thing if lisenses would not matter everybody would switch to bsd because its cleaner and better than linux.
                      This is subjective and mostly false.

                      Note: I prefer FreeBSD to any Linux distro.

                      I wanted to discuss why I think a alternative to linux and bsd is needed and most likely it should be a micro kernel, because linux design is old, yes its mainstream now, its the dominant system now, windows is on its die-bed, if there would be no hardware-windows bundle it would be even worse... but I count smartphones and tablets as pcs and there windows is dead.
                      Windows is far from dead even if you counted smartphones and tablets as PCs. You are throwing a lot of hot air around.

                      You claim that an alternative to Linux and BSD is needed and that it should be a microkernel but you have not said why this is so.
                      Why is this alternative needed and who needs it?

                      And that will not change for the next 5-10 years maybe even 20 years. But maybe Hurd is really dead (even Hurd was never = Hurd it was different oses as far as I read it, so which hurd in the future is now dead would be the question) but I just say, linux is about rewriting everything why should the kernel be the only thing where this will never happen?
                      Yes, for all intents and purposes, the HURD is dead. The HURD suffers from quite a number of deficiencies and it may be easier to re-implement it from scratch than fixing it.

                      Linux is not about rewriting everything.

                      No one needs to rewrite the Linux kernel.

                      Anyone is free to implement a microkernel and a microkernel based operating system. This does not need to have anything to do with Linux.


                      I'm quite interested in microkernel OS development and I follow the development of a few of them. Some of the better ones:

                      NOVA Microhypervisor. Contribute to udosteinberg/NOVA development by creating an account on GitHub.


                      HelenOS
                      http://genode.org/ You should really check this one out.

                      Richard Braun (HURD guy) is working on a replacement for MACH to create a HURD-like system. He thinks the HURD is dead also.

                      You should note that all microkernels are not created equal. The difference between any two can be even greater than the different between Windows NT and Linux. So just saying "microkernel" means nothing at all. A lot of people even claim that some kernels are microkernels when they are in fact monolithic kernels(Haiku' kernel, XNU. Windows NT and a few others). Even calling them hybrids does not make sense from the perspective of someone who understands what a microkernel is and how it differs from a monolithic kernel.

                      Comment

                      Working...
                      X