Re: Hurd Advocacy?
Hurd Advocate <email@example.com> writes:
> What can the Hurd community do to promote their favorite OS?
> It seems like the Hurd doesn't really have a critical mass of
> users to spur the growth we'd all like to see. So I was wondering what
> everyone thought was the best way to attract more developers/users to
> the Hurd. The reason I initially looked into the hurd is the BitKeeper
> fiasco. I found it unpalatable for free software to be beholden to a
> proprietary master. I thought it would't hurt to look at the
> alternatives. And I came across the Hurd. From what I could initially
> find out, it seemed like it had interesting and modern architecture,
It is scary to find out bitkeeper is good for something ;).
> which could solve the "Linus doesn't scale" problem more cleanly than
> the BitKeeper solution. (And here I'm assuming that things like
> userspace device drivers and the fact that any part of the system can
> theoretically be replaced on the fly really does solve that problem).
There are not much userspace drivers yet, but this is the
feature. Everything in the Hurd can be replaced on the fly
AFAIK. Think about the TCP/IP stack, filesystems, etc.
> was, of course, also attracted just because it was something new and
> different. So I wonder what attracts everyone else to the Hurd. To
> that end here are a couple of questions I have.
1) Political reasons. AFAIK all Hurd hackers care more about freedom
than about technology. (Like bitkeeper vs. CVS).
2) Technological reasons. Did you read the papers on
http://hurd.gnu.org. I can also advise you to install GNU/Hurd and
have a look what is possible with GNU/Hurd and what wasn't with
3) New technology, new possibilities. Users can do more, but the
system is more secure.
4) The community.
5) It's fun. :)
> - Is there any killer-app for the Hurd (available now or in the future)
> that we think will bring the masses in? Or phrased a different way,
> is there any one feature that people would be willing to think about
> converting over for.
This works perfectly for me.
> distributed OS?
This is not possible yet. Because the Hurd runs on top of a
microkernel and uses RPC for almost everything this is really easy to
add as a feature for the Hurd. (Easy compared to monolithic kernels).
> better security model?
One thing I can think of is no, one or more UIDs.
> more customizable? (is that a word?)
For my feeling it is. Can you give me an example of what you want to
> - Are hardware compatibility problems more of a problem for newbies, or
> is it the lack of software which stifles adoption. (And for the
> record, I think the killer-app would be Linux and the Hurd running
> side by side on top of the same micro-kernel. That would make
> migration easier, since you could still have access to your important
> hardware and software that hadn't been ported over yet)
Running GNU/Linux and GNU/Hurd simultaneously is really easy to add as
a feature on some architectures. For example the mklinux distribution
uses a hacked linux kernel that runs on top of Mach.
It is already possible to run multiple GNU/Hurd installations
simultaneously (this is called a neighbourhurd). For mklinux such
support is easy to add. IIRC someone already wrote some code for this.
There are many packages for Debian GNU/Hurd available. Of course much
more programs should be ported to the Hurd. (porting = fixing bugs in
the program or in the Hurd most of the time).
The driver support is ok I think. It isn't really great, but most
commonly used ethernet cards are supported by GNUMach.
> - Is it hard to attract developers because the project is too complex.
> Instead of just learning one system, you have to learn about two: the
> hurd and mach. And who would want to learn about mach when it's
> scheduled for removal whenever the L4 kernel gets traction (3-5 years
> out?)? Or is it the "multi- threaded servers are hard to debug"
> problem still.
I think it is hard to learn about the Hurd (that's what others say).
I think it is better to learn not everything, but just start hacking
and you will learn. Not everyone will agree with me about this, but it
works for me.
The multi threaded servers is not a problem for me while
debugging. Actually it is easy to develop stuff like translators
because everything is in userspace.
> - Is a lack of documentation the real hard thing for new developers to
There is very little documentation. Luckily the comments are really good.
> - Are we nice enough to newbs? (I tend to think so, but there was a
> little hissy-fit about change-log colon-placement for hello.c on
> bug-hurd last month)
It's just like any project. A lot of different people, a lot of
different attitudes. :)
> - Do we suffer from a lack of charasmatic leadership and direction?
I don't think so.
> - Is there any one thing which could be fixed to attact a lot more
I doubt if this is a requirement for an OS still in development.
I think this won't ever be added to Mach. I'm sure it will be done for L4-Hurd.
Does it work? Doesn't it?
> journaled file system?
I agree. IIRC ext3 is on Ognyans todo list. :)
IIRC there was a problem with pthreads...
> - Is advertising our problem? Do we not get enough exposure to
> potential developers? (And here I'm thinking CS undergrads) I'm
> thinking that a new developer could have a lot more influence on the
> design of the Hurd (since it's still in flux) than say a more mature
> project like Linux or FreeBSD.
I agree we need more developers. I think the project looks a bit dead
to the outside world. Perhaps a release can help us, and releasing
more frequently. Also news sites like kerneltraffic and kerneltrap and
even slashdot can help.
> - Does anyone think that companies like RedHat or IBM might think about
> funding some summer college internships to work on something like the
I'm afraid they are only interested in things that works in production
> - Is there any future development that might drive people to the Hurd?
> Like the SCO garbage or DRM binaries/signatures in the Linux kernel?
I don't hope such things happen. But I would like it that people will
have a look at GNU/Hurd. It is a good thing that people care about
free software and switch to (or keep using) free software because of
that. I'm sure about one thing, GNU/Hurd will always be completely
> - Is our installation proceedure/Debian system overly obtuse?
I don't really understand your question. Debian GNU/Hurd cannot be
easily installed like Debian GNU/Linux is. (I hope that answers your question).
> - Are we always destined to play catch up with Linux? (eventhough we
> had a headstart?)
the Hurd should be capable of doing everything Linux can I think.
For me the following things need to be done before I will use it on my
desktop all the time:
- The port to L4 must be done.
- The 2GB limit should be fixed.
- PPP should work