“The release of a production version of the free GNU operating system (OS) has been delayed beyond the end of the year, as the current development version of the system does not support large disk partitions and high speed serial I/O (input-output), according to Richard Stallman, president of the Boston-based Free Software Foundation (FSF).” Read about Hurd at LinuxWorld.au.
Stallman: Disk, I/O Issues Delay GNU OS
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
Basically Stallman wants to get rid of Linux ASAP?
Anyway, these issues have been here for ages. It is the lack of programmers, rather a fundamental flaw of design, that prevents it from being fixed.
Clearly most people are not going to repartition their disks to be able to try out our Hurd based system
Actually, RMS, they do. Unless they have their own seperate HDD for each distro/OS, they would repartition to provide space for HURD. And due to HURD instable state, it is unlikely they would have anything more than 1-2GB for this kernel (or whatever they call it).
stop taking things out of context… that wasnt what he said.
Hurd is hurting again
I will have to agree with Robert here. RMS was speaking of Hurd’s own limitation of using larger partitions, so his line of thinking was something like this: “this limitaton should be fixed, because the kind of people who will try Hurd already have partitions waiting for alternative OSes to get installed, and Hurd should be able to utilize these existing partitions“. This does apply in my situation btw.
Just imagine having to repartition your 40Gb into 20 to 40 chunks just to be able to use it :-S How they could miss this is really beyond me.
(Although it would be actually be the same if you wanted to use Win95 because fat16 had a max size of 2Gb)
well from what i’ve read the hurd seems a very interesting concept, as for an operational OS well they being working on it since around 1990 and that’s just on the kernal the rest of the GNU os being around longer. And it’s still not anyway usable. Even so i think i’ll download it and put it on my old “Cyrix Instead” box (all 120mhz
your friendly with the Fresco developers aren’t yeah? how that going? I had alook at there snapshot directory recently seems to be good amount of activity
It’s design flaw – HURD is in development for more than 10 years. Disks were smaller at that time.
The limitation is actually quite obvious – many old Unix systems (e.g. Ultrix) had it. Limitation is because of some fields in inode structure are still 32 bits. Take off sign bit and you’ve got 2Gb restriction.
I really suspect that their (int) is actually int32. I wonder about time_t size – with the speed they develop it they hit 2036 problem before they finish.
Actually, the limitation is rather less obvious. Though the fields might still be 32-bits, the code can be written to refer to sector addresses, since disks can only be addressed in multiples of 512 bytes. In this way, Linux, for example, streches the standard 32-bit layout of ext2 to 2TB disks. It’s only recently, actually, that support for partitions (usually logical volumes of multiple physical disks) greater than 2TB has been added.
By my understanding they would rather wait for 64bit so that the code could stay clean.
Going to 64bit raises a lot of limits
{Using existing partitions to install Hurd}
This does apply in my situation btw.
Then you’re a braver man than I! *cough*
When a “feature” of Win95a toasted the partition table on my OS/2 box (which subsequently killed the filebase for my BBS!) I have taken to the theory that when testing a new OS, that it should have its own hard drive.
Why? Because I never bother putting the covers back on my machine, so I’ll just whip the poor bugger’s IDE cable out, throw it in another drive and voila! Robert is your father’s wife’s brother!
Sure, dual booting is fine when you explicitly trust the OS (such as my wife’s machine, which dual boots Win98SE & Gentoo, or my work server (still to be migrated) which is currently running Win2K Server & Win2K Prof.) However, if that trust isn’t there (such as an initial release of a new OS that has already had some form of file system issues, it gets its own disk and thus save me a heap of heartache!
(Hell, I’m sure I’ve got an old 2Gb drive lying around somewhere…)
No – Stallman does not want to get rid of Linux. The Hurd is a different type of kernel than Linux. Technically, the Hurd is a completely diffferent beast than Linux, and most of the OSes out there.
Anyone who is interested or curious should read the following docs:
http://www.gnu.org/software/hurd/hurd.html
No – Stallman does not want to get rid of Linux.
See http://www.linuxworld.com.au/news.php3?nid=1333&tid=1
It seems clear from that interview that as RMS currently sees things, people are using the GNU/Linux system (i.e. the GNU userspace with a Linux kernel) and he would rather they be using a completely GNU system (i.e. GNU userspace with a GNU kernel) It also seems clear from that article that the kernel architecture is not as important to him as having a compltely GNU computing system.
The Hurd is a different type of kernel than Linux. Technically, the Hurd is a completely diffferent beast than Linux, and most of the OSes out there.
Well, this is true, the Hurd is currently Mach with plans of eventually transitioning to L4 down the road. However that doesn’t really play any role in RMS’s desire to eventually replace the GNU/Linux system with the GNU system.
I don’t really understand your response. Do you agree with me or do you disagree? RMS has never said he wants to get rid of Linux.
The Mach (and soon L4) is only part of the Hurd architecture. I have a feeling you know that, though.
I disagree, and think it’s clear RMS wants to replace the GNU/Linux system with a fully GNU system. Here’s some snippets from the interview:
“We actually have the GNU kernel working and we can now produce the GNU system, as opposed to the GNU/Linux system that people have been using so far,” said Richard Stallman, who is in India this week to attend a GNU/Linux Day in Pune.
“Linux is a kernel, and now we have our kernel which is an alternative to Linux, and they both work in the context of the overall GNU system, as the kernel alone won’t run without the rest of the system,” he said.
Distributions of GNU/Linux include non-free software and that diverts the user and developer community from the goal of freedom, according to Stallman.
“One of the reasons we are looking forward to having the GNU system finally available from the GNU Project is that it will be only free software,” Stallman added.
The true nature of the Hurd can be revealed through the power of shell utilities. Just perform this handy conversion!
echo “Hurd” | tr H T
I can see that a mile a way. TURD. Makes sense to me!
How is this anything that can be called a “GNU OS” if the GNU people didn’t develop the kernel? According to the article they are using someone else’s implementation of the Mach kernel…right? Am I misunderstanding this?
Shouldn’t this new OS be called GNU/OSKit since the logic behind GNU/Linux was that Linux was the kernel??
Good to see Stallman will also be working to ensure that businesses won’t be able to profit from providing web services. One day he’ll realize that not everyone can live on donations like he does…
…in 2002 we would have GNU 1.0. I’ve been following the discussions on the debian-hurd list for quite some time. And I was absolutely surprised when RMS spoke about having GNU 1.0 in 2002. This absolutely contradicted the state the Hurd is in. It is an interesting thing, of course, and I wish it the best of success!
But it’s still way far from being suitable for productive use. So I consider it foolish to promise something that anyway you will not be able to deliver (the main reason being the limited number of Hurd hackers). The more it surprises me since, as far as I see, RMS is not at all involved in GNU/Hurd development
Shouldn’t this new OS be called GNU/OSKit since the logic behind GNU/Linux was that Linux was the kernel??
No, because the real kernel is still HURD. OSKit just provides a base for it. It’s no different from Darwin which is the kernel of Mac OS X and is based on Mach.
Isn’t it funny how some people like to attack things they don’t understand? It’s especially funny when the attack is totally inane.
People should remember that GNU stands for “GNU’s not UNIX”. As good as Linux is, it is still essentially a clone of UNIX. HURD is very different. It’s probably more similar to an OS like Plan 9 than it is to UNIX. I think that is one major reason why it has taken so long to develop. The ideas and concepts used in HURD are so new and different that people couldn’t relate to it. With Linux having been established around the same time, people decided to stick with what they knew: a UNIX-like system.
Isn’t it funny how some people like to attack things they don’t understand? It’s especially funny when the attack is totally inane.
People should remember that GNU stands for “GNU’s not UNIX”. As good as Linux is, it is still essentially a clone of UNIX. HURD is very different. It’s probably more similar to an OS like Plan 9 than it is to UNIX. I think that is one major reason why it has taken so long to develop. The ideas and concepts used in HURD are so new and different that people couldn’t relate to it. With Linux having been established around the same time, people decided to stick with what they knew: a UNIX-like system.
Yeah, but it was for sarcastic reasons… I forgot the “:-)”
On Fresco… it still doesn’t work on Hurd, some pthreads problems. Most of the development activity you see is bug fixes and repaired make files for GGI. You see, 0.3, or Milestone 1, is coming out in the next 6 months. So, please, for the love of your mailboxes, don’t subscribe to fresco-changes and leave for 6 months Internet-less vacation.
RMS GNU dreams are his right. I don’t see why that bothers anyone really. It’s not like he’s plotting to destroy linux or beat up Linus, or anything like that. He created a system to further freedoms, sent it off into the world to fend for it self, other people have taken it and used it to do great things, but he himself has his own great ideas he wants to see.
This is similar to the Linus kernel patch topic on Slashdot several days ago. Linus created Linux and sent it into the world for everyone to share. He himself has his own ideas, and that’s his right to do his own thing.
Freedom to “do” is the whole point. Don’t talk about it. DO IT!
I have out-grown bashing RMS, but really, for the sake of freedom, why create something totally new when it doesn’t per se fixes a lot of problems with current ideas and architectures, only adds on to them?
I mean, the Linux kernel. Under what license? GPL. It is so simple for RMS to use it in his GNU OS. It would be a totally free (and useless) distribution. If he likes (if he is too cheap to buy BitKeeper to reimplement it), he can even fork it, get some Linux developers, and viola!
Rajan, see my above comment: ” People should remember that GNU stands for “GNU’s not UNIX”. As good as Linux is, it is still essentially a clone of UNIX. HURD is very different. It’s probably more similar to an OS like Plan 9 than it is to UNIX. ”
RMS doesn’t just want to clone UNIX, he wants to replace it with something better. Hence GNU’s Not UNIX. Replacing all the usermode UNIX tools with free software equivalents is only the first part in this plan. Once that is done, it would be considerably easier to transition people over to the new paradigm that HURD offers.
As for your question, “why create something totally new when it doesn’t per se fixes a lot of problems with current ideas and architectures, only adds on to them?”, what makes you think this? Why is HURD so bad? If you’re just looking at the license then you’ve missed a major point of HURD.
I don’t think there is anything stopping people from using proprietary software with GNU/HURD. It is a microkernel-based design, so you should be able to install new kernel modules without having to compile anything.
I hate to say this. But is HURD an OS or a political statement. If the motivation for the OS is just to have it under RMS’s GNU license, is it any wonder it hasn’t reached release after 19 years. Frankly Stallman’s tactic of license first and code second is the opposite of how it is supposed to be done. I also get the feeling that he is a terrible project leader. Most users don’t care if they can source weather it is under the GNU, GPL, or gasp the BSD license. RMS’s over zealousness will probly prevent any company from even considering building an OS distribution once the kernel is finally released anyway.
When will stallman get it…HURD is developing slowly because the world has lost interest in it (if the world even ever had much interest in it)…and yes it does reveal it’s 1980’s era design in several ways…micro kernels are a glorious academic failure, nearly every system that started off as a microkernel has morphed into a modular monolithic or faded into obscurity (Windows NT – morphed, NextSTEP/OS X – morphed, mkLinux – faded, lites – faded)…
Also, Stallman is definitely out to get Linux because his control-freak personality clashes with the refusal of Linus to be Stallman’s bitch…
HURD is interesting from a concepts and research OS standpoint, but does anyone honestly ever expect to the HURD to be a truly useful OS? by the time they even acheive the functionality of Minix they’ll be decades behind the rest of the world’s OSes…and how much longer before this is server grade? last I knew, HURD could barely run XFree, let alone a desktop environment of any sort
I have no problem with HURD itself, just Stallman’s ridiculous claims about the HURD…that it is somehow realy advanced (80’s design)…and that it is useable today (doesn’t even run half the software Linux and the BSDs do)…these statements are just absolutely ludicrous
-bytes256
I’d split the difference of your comments. I think RMS would like to see things like SuSe’s YAST go away. Also, I’m sure he would like to have plenty of people using the GNU system, or else it wouldn’t be in development.
But he actively promotes Debian GNU/Linux. On the GNU site he tells people that the easiest way to get started with free software is GNU/Linux. (http://www.gnu.org/help/help.html#helpgnu)
If I had to wrap it in one sentence, I’d say that RMS wants GNU, not GNU/Linux to be the poster-boy of FSF.
I think you may have oversimplified the situation somewhat. While I think the monolithic-but-modular design taken by Linux is a good choice, the design taken by HURD is an interesting one. It has a great deal of flexibility, and unlike *NIX it looks well-suited to distributed computing. This will become increasingly important in the future as the number of computerised devices that we carry around and have in our homes grows in number. Grid computing (currently being pushed hard by IBM et al.) would also be a good application for HURD. For more info, see this interview with a HURD developer: http://www.kerneltrap.org/node.php?id=5
Is BEOS a micro-kernel design? If so in my opinion it was successful. For me this was the coolest operating system.
Responsiveness was just amazingly fast and everything seemed easy to find and just worked when supported.
I must admit though that I cannot really see any advantages a micro kernel has over the modular design Linux uses.
> nearly every system that started off as a microkernel has
> morphed into a modular monolithic or faded into obscurity
Except QNX which doesn’t even have drivers in the kernel,
and still offers the lowest latency around and hard real-
time assertions.
AFAIK QNX is doing well enough and is used a lot in critical
applications. They’re just not into the PC market.
PS: Is that the L4 nanokernel developed in Dresden that is
being considered for HURD?
RMS has publically bashed Linus and the Linux kernel on innumerable occasions over everything from bit keeper to proper OS naming, while still profiteering off of numerous Linux conventions. His zealotry has driven him insane (a mach based microkernel should be sufficient proof!!), and I think the final name for the HURD kernel should be CMU/UU/HURD if you want to follow the logic of RMS. Afterall, mach originated at Carnegie Mellon University and was later further developed at the University of Utah.
I have no problem with him developing his own kernel, I just do not appreciate the way he treats the Open Source (he doesn’t like that name either) community.
ps
I hope He doesn’t make it on to the GNOME board, because his extremist political beliefs could prove harmful for GNOME develpment.
Everyone seems to bash RMS — are you doing it because everyone seems to think it’s cool to do it or do you have a real reason to.
RMS was building the FSF and writing free software while most of you/us were still in diapers. The fact that so much of linux software is actually built or inspired by RMS and the fact that he gets bashed so much is sheer ingratitude.
And if I had built an almost complete unix clone, without a good kernel and some other dude built a kernel and everyone else was going to forget my contributions, I’d be damned pissed to. It should be called GNU/Linux as MOST OF THE SYSTESM IS GNU stuff. Calling it GNU/Linux makes sense when other kernels are dropped in such a GNU/Hurd, etc.
Does RMS deserve a seat on GNOME, — absolutely. Before you forget the man that single handedly kick started the open source movement, look in the mirror and ask yourself, “what have I done in my life time to change the world?”
Besides that, GPL is the MOST EMPOWERING END USER LICENSE ever devised.
Back in the early 1970’s the idea for having programs compile into a universal assembly language (called P-code) that could be run on any architecture was quite popular in academia. A lot of experiments were done with P Code. However the PC came out and a whole generation of programs became highly focused on performance issues and very quickly higher level languages were out of fashion for production uses. Performance won out over manpower savings.
In the mid 1990’s the P-code ideas started to heat up again with Java. Today .net, java, parrot all basically are P-Code systems. While performance issues are still a major criticism of P-Code systems it appears that the manpower savings outweigh the performance issues; to the point that by say 2010 maybe 2015 its likely that most code will compile to some 2nd wave P-code of one sort or another.
Micro kernels are in many ways similar. There is no question that monlithic kernels are faster. The is no question that micro kernels are faster to program, update, debug, configure… The only real question is how much is the enhancement is programmer productivity worth in terms of the performance hit (and remember performance hits in the kernel effect everything so they are very expensive in terms of additional hardware requirements).
The Linux kernel has serious long term maintainability issues. That’s the reason that Linus has to be strict about kernel modules. The level of skill required to work on the kernel is quite high and as a result work progresses more slowly than many would like. Would it be worth losing 30% performance to get updates 3x as fast? Right now probably not. In 10 years?
The reason for Hurd is that eventually people are going to want extremely easy to configure / modify / manipulate kernels and are going to be willing to pay say 30% of their CPU for it.
From what I know RMS has done a lot. However that does not mean that he cannot be criticised.
Making statements like Before you forget the man that single handedly kick started the open source movement, look in the mirror and ask yourself, “what have I done in my life time to change the world?” is pure fallacy ad hominum.
I agree people should give him his due, but they are quite entitled to say why or why not his ideas are flawed without having to have changed the world. It is up to anyone reading those views to assess there relative merits.
Yama, what I meant is that RMS, in the PR, is advertising HURD on the basis on “freedom” (freedom to freeload of other’s work). Besides, to me, taking Linux and making it a microkernel and extending it and slowly making it different is entirely possible. In fact, MkLinux did just that.
Making something different for the sake of being different doesn’t go to all to high for me. No HURD developer ever explained in detail why their idea in theory would be much better than UNIX and Linux in specific.
BE OS isn’t a microkernel, Be OS was. Be OS was slow in kernel related stuff, especially with BFS which was slow with heavy files. The responsiveness you see is the fact that the UI was multithreaded, not because the underlying OS is fast.