Philip Charles has announced the availability of the K8 series of Debian GNU/Hurd CDs. He says: "The 2GB partition limit has gone [story]. The tests were made on a 20 GB partition. [...] The K8 disc set at least matches the quality of the K7 set which IMO was the best I had produced to date."
You can download the mini-ISO here, the other ISOs are here. In order to install Debian GNU/Hurd, read the installation instructions from the Debian website. The K8 Readme offers more general information about the release.
From: Philip Charles [email blocked]
To: debian-hurd mailing list [email blocked]
Date: Thu, 30 Dec 2004 23:48:38 +1300 (NZDT)
Subject: K8 CDs etc
I know it is a bad time of the year, but the K8 CD iso's are being
uploaded to gnuab.org at the moment. The first image 70% on its way. All
four should have been uploaded in about 40 hours. 16 KB/s upload Ahhg...
The tarball and mini-iso have already been uploaded. I hope that Guillem
Jover is not on holiday. The DVD just completed its testing an hour or so
ago, it is too big to upload and will have to be posted.
Main features. The 2GB partition limit has gone. The tests were made on
a 20 GB partition. The 2+ years old mkfs in the GNU/Linux installation
system creates the filesystem on the large partitions nicely. This means
that the Linux installer has not had to be modified and so remains at "K".
nano is broken and so nano-tiny has replaced it in baseGNU and the
tarball, so remember the "-tiny" bit when invoking the editor.
The GNU archive overflows the DVD by about 300MB. There are eight CD
images, but as usual only the first four are being made available.
The K8 disc set at least matches the quality of the K7 set which IMO was
the best I had produced to date. The quality of a disc set is largely
determined by the quality of the packages and the overall state of the
official Debian and gnuab archives.
As usual, comments and suggestions are welcome.
Phil.
--
Philip Charles; 39a Paterson Street, Abbotsford, Dunedin, New Zealand
+64 3 488 2818 Fax +64 3 488 2875 Mobile 025 267 9420
philipc@copyleft.co.nz - preferred. philipc@debian.org
I sell GNU/Linux & GNU/Hurd CDs & DVDs. See http://www.copyleft.co.nz
I still fail to understand th
I still fail to understand the big deal about HURD. It's not like the principles of open source coding are more represented than the competition (given that the Linux Kernel is GPL as well, though BK use could be a slight issue). It doesn't seem like the daemon model is at all technically superior, as even some of the main HURD developers admit that it merely makes the code really, REALLY hard to debug. It doesn't have the advantage of BSD-like stability/code-review nor Linux-like development speed, so why use it?
If this comes of as insulting or as a flame, I'm sorry, I really don't mean it that way. I'm just asking for someone to provide a reason to get excited about HURD development, or even to keep developing it at all.
HURD's architecture has some
HURD's architecture has some pretty neat features that you don't see in other OSs. For instance (IIRC) it's possible for users to write their own OS components, e.g. filesystems and use them without compromising the security of the system.
For instance (I think this is the example on the HURD site), a user can write an encrypted filesystem driver, start it up and then be able to access their own encrypted filesystem just like any other filesystem. You can't do this on traditional UNIX-like OSs without trusting the user with root privileges.
I don't know if this actually works yet, though it's more likely to work now than some of their more ambitious goals (SSI clusters, which they call "collectives").
They also talk about running multiple instances of the HURD on top of GNU Mach simultaneously, e.g. using one to debug the other. That's quite a cool feature...
Microkernels are also just an appealing concept to some (including me), in principle, despite the practical problems encountered.
it's possible for users to w
it's possible for users to write their own OS components, e.g. filesystems and use them without compromising the security of the system.
For instance (I think this is the example on the HURD site), a user can write an encrypted filesystem driver, start it up and then be able to access their own encrypted filesystem just like any other filesystem. You can't do this on traditional UNIX-like OSs without trusting the user with root privileges.
You can do that with Linux (check FUSE and LUFS at Sourceforge).
You can do that with Linux (c
You can do that with Linux (check FUSE and LUFS at Sourceforge).
I don't think the author expected you to think that was the "only" reason to try Hurd. The ends don't always justify the means :).
I've not looked at FUSE or LUFS but the fact that the whole unix personality of the Hurd is meant to be implemented in userspace means it can be safely worked around and ignored. No matter what you do on linux you are still using the linux kernel at the root of it all. Hurd will have either gnumach or L4 at the core of everything.
A better example of the kinds of flexibilities you could take advantage of might be the L4-Linux demo-cd. You can run the L4 microkernel, it's own GUI system, some L4 programs and L4-linux as a user space program.
The closest thing Linux has to achieving this is probably Xen or UML [User Mode Linux]. Again, here the way in which you achieve similar functionality is different, UML, Xen and L4 can all manage and run instances of Linux. UML does this in userspace, Xen does it by partitioning the hardware, and L4 does it by having Linux ported to it's threading and memory management system.
Sometimes the point is to find a new way to build the system you want so you can leverage the atomic pieces in a new way later.
UML does this in userspace,
Xen also requires Linux to be ported to it's own architecture, with special memory management, processor context management, etc.
I'm not sure what you mean by partitioning the hardware... To clarify: it doesn't just statically partition the hardware (although that can be configured if preferred) but multiplexes virtual machines on it dynamically.
The difference is
Yes, you can have user space filesystems in Linux now, but its still a module someone hacked up as an afterthought in a monolithic kernel. Imagine if just about everything other than the most basic kernel functions ran in user space and could be swapped out or chained etc at will, within limits set by the admin. Of course this implies a performance penalty (though CPU designers are working on lower cost context switching), but the real advantage is architectural elegance and flexibility.
A lot has been made of the performance penalty of microkernels, but I think of it in the same way as scripting languages: sometimes flexibility trumps brute speed.
You can implement filesystems
You can implement filesystems in userspace with FUSE and LUFS but can *untrusted* users implement their own filesystems for private usage? I suspect the answer is no, whereas HURD is designed to support this sort of thing.
Filesystems for private usage?
Pardon my ignorance, but why on earth would users want to create filesystems for private usage? All users that I know of with exotic interests like that have full access to at least one computer.
And, if they should feel like creating a filesystem for private usage, they all use Bochs, even if they have full access.
The utility of filesystems
It's been said that in UNIX, everything is a file (e.g. /dev/sda, /dev/null). In the Hurd, everything is a directory, and this concept is pushed even further than crazy files in Linux like /dev/random. The result is a whole different way of approaching problems. In Linux, one frequently uses cat to pipe a file through stdout->stdin of another utility. In a Hurd system, one might instead cp a file into a directory that represents the utility's input. "creating a filesystem for private usage" will be something that occurs almost every time you run a program.
This is perhaps a misunderstanding: "create" here does not mean "design, develop, and implement". It basically means "mount". The cool Hurd thing is to let users and their programs mount filesystems (called "translators") that do arbitrarily cool things, without having any effect at all on other users. At a minimum, it allows every user to be root of their own little domain.
Much of what Hurd tries to accomplish can be approximated by emulators and virtualizers, but these approaches are less powerful, less elegant, and way slower.
because...
Why do kids still play in the mud ... besides all this "great" technological achivements?
Because its a part of their growing, part of their learning, part of their expressing and developing their creativeness...
I think that, while Linux will remain mainly a monolitical kernel (an it should stay that way), it will (should) implement some of the techniques developed in HURD... user-private file-systems, out of kernel drivers, etc...
And who knows, one day we might find that the evolutionary path that Linux has taken ends in a dead-end and doesn't fit the computer systems and ser needs in the future.. if that happens ew better have an alternative.
I think the thing is not that
I think the thing is not that one should use it. The thing is that there is an alternative - a GPL-Microkernel. Maybe at some point (maybe in 5-10 years from now) we will be happy that this alternative is existing. Till then it is a nice thing for developers. I do not want to use anything but GPL-kernel. I have never tested it, though.
From my perspective, the HURD
From my perspective, the HURD represents an attempt to put together a hypermodern microkernel OS that is 'pure' OSS and which has the advantages in stability and adaptability that are at least theoretically possible in such a system. Recognize that this represents an 'experiment' in every sense, and it is easy to see why the FSF folks are pursuing this. Will this be valuable for the rest of us? I believe so, if only because the experience with the only current production microkernel based OS (Macintosh System X) in general use has been very favorable, and the availability of an alternative to 'Gnu-Linux' in the OSS arena would be a favorable development, expecially for developing nations and organizations needing a more lawyer-proof alternative to the Windows mess.
PowerPC
The question is when will we be seeing hurd on alternative architectures such as the powerpc architecture.
Re: PowerPC
Peter Bruin is working on the PowerPC port. He has the base stuff working mostly, but there is no distribution to install. I heard he now moved on to porting the Hurd to L4 on PowerPC. Search the bug-hurd (and perhaps debian-hurd) mailing list archives for his name for details.
Michael
Unknown host
Sadly it appears that gnuab.org (and the various download sites listed here & elsewhere) is non-existent, which make it rather hard to download the iso to try it out.
gnuab cd mirror
You can find the CD images here as well:
http://freebsd.czest.pl/dunstan/Hurd/
Read about it here:
http://lists.debian.org/debian-hurd/2005/01/msg00003.html
Re: Unknown host
There is a mirror for the K8 series here, which has the first ISO and the mini-ISO.
I verified the contents of the mini-ISO via its MD5 sum.
Michael
What about Bochs?
Has anyone gotten the new ISO to install and run with Bochs?