Re: Hurd Advocacy?
On Fri, Aug 15, 2003 at 06:06:25PM -0400, Shawn Boyette wrote:
> This is exactly the conclusion I reached a few days ago when I read
> about L4 through Debian Weekly News. I keep wanting to play with the
> HURD, but it keeps on not existing.
Well, if you only want to play, there are enough toys out there that are
more existing than the Hurd. The current existing Debian GNU/Hurd should
nevertheless be enough to play with the Hurd. As a proof of concept, it
works quite nicely.
> This is almost farsical,
> announcing a switch to a new kernel architecture (which, I might add,
> is already deprecated by its developers -- Pistachio is the current
> branch of L4Ka, not Hazelnut) before the previous new kernel
> architecture (OSKit) migration is even completed.
We never considered Hazelnut to be a platform for the Hurd, it is way too
limited for that. Not sure where you got it from that Hazelnut is a target.
Our is Pistachio.
The OSKit integration into GNU Mach is not related to the Hurd at all. It
is just a different set of drivers in GNU Mach. Not many people are
interested in hacking on that, and this is for good reason. Although it
would make the device glue code in GNU Mach more sane, it wouldn't change
much from the user experience in the Hurd (in other words: it would still be
slow and unstable). The main features you would get with GNU Mach v2 are a
couple of drivers (like, a random device, and better serial port support for
PPP).
> Is this a project to produce a kernel for a GNU OS, or is it some sort
> of never-ending, ivory tower, trans-academic wankfest for kernel
> geeks?
The Hurd is not interesting at all from the perspective of producing yet
another unix core. There is BSD, there is Linux, and between the two of
them there is absolutely no niche for yet another monolithical kernel
intended to be used in production systems.
So what the Hurd is about is indeed its user-space design. However, just as
the Hurd on GNU Mach as a proof of concept is a good success, it is, as a
production system, a fantastic failure. There are several fundamental flaws
that absolutely must be addressed to hopefully make the Hurd competitive in
performance, stability, security and features. When looking at the options,
L4 seemed to be the logical conclusion, because it furthers the fundamental
design concepts of the Hurd and implements them more consequently than Mach
did.
Going to L4 will, unlike OSKit, require a complete redesign of the Hurd.
Much of this redesign has already been done on the conceptual level, and we
are slowly working towards the actual implementation. There are a couple of
interesting issues to solve on the way. All this is going on without a
guarantee that the final result will work better than the Hurd works now.
Although we have the feature-component of it under control, nobody can
predict if the performance and stability will improve and become
competitive, or if it turns out that it won't. Nobdoy can tell because
nobody attempted to implement a system like the Hurd before at this scale.
The Hurd is one of the few innovative free software projects. While most
other projects are "more of the same" in that they reimplement well known
and proven concepts, the Hurd is walking on new grounds, with all the
problems that this implies. And it is lacking some very experienced
developers with lots of spare time and motivation. We won't attract such
developers with GNU Mach, though, I can tell you that.
Now, about the ivory tower comment. Actually, most people on the earth have
stopped believing in a microkernel approach a long time ago. I think that
the history of operating systems has as much to contribute to the history of
the Hurd as other factors. However, if history is any judge, there is a
good chance that the hype curve will bend the other way some time in the
future. Hopefully we will then be able to profit from that. But that
requires us to find answers to questions so we can offer them to interested
parties at that time.
Thanks,
Marcus
--
`Rhubarb is no Egyptian god.' GNU http://www.gnu.org marcus@gnu.org
Marcus Brinkmann The Hurd http://www.gnu.org/software/hurd/
Marcus.Brinkmann@ruhr-uni-bochum.de
http://www.marcus-brinkmann.de/
Reply to: