[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Hurd state

From: Marcus Brinkmann
Subject: Re: Hurd state
Date: Thu, 06 Jan 2005 21:44:03 +0100
User-agent: Wanderlust/2.10.1 (Watching The Wheels) SEMI/1.14.6 (Maruoka) FLIM/1.14.6 (Marutamachi) APEL/10.6 Emacs/21.3 (i386-pc-linux-gnu) MULE/5.0 (SAKAKI)

At Thu, 6 Jan 2005 07:31:30 +0100,
jemarch@gnu.org wrote:
> I have to admit i am lost about the actual Hurd state.

That's arguably good news.  When I joined the Hurd in 1997, there was
little doubt about the Hurd state at that time: It was just about at
the end of dieing a slow death.

So, if today the state of the Hurd looks vaguely alive, then this
means that it is not totally dead.  And in fact, I think it is
positively alife, at a slowly growing rate.

What is the reason that the Hurd is crawling, lingering at times, and
not skyrocketing like other free software projects?  There are many
reasons.  But two seem to be dominant:

First, GNU/Linux is good enough for most people.  Second, it is not
clear that our approach is better.

I think the first is evident, and doesn't need much explanation.  The
second point is more subtle.  From what I heard on conferences etc,
almost everybody buys into our vision.  The goals we have are good,
and useful.  However, there is substantial suspicion if it is actually
technically possible to deliver these goals with the design we have
chosen.  OTOH, it is not clear what alternatives there are.  It seems
people have resignated and accept the current state of the art of
operating system design to be the best achievable, given the demands
of the market.

In other words, we are in a position where we have to provide proof
against overwhelming evidence.  A vision is not quite good enough.
This directly translates into a lack of resources, and for the
resources we have, a lack of technical guidance: We can't tell those
people who want to help _how_ they can help, because we don't know.
We are pioneers.  We try to do something that nobody has ever done
before, and we don't know if or when we can achieve what we want to
do.  We should not give up before trying at least.  It is an ambitious

> I know mach has
> been deprecated in order to take the advantages of a (supposely)
> cleaner microkernel design: L4.

Some of us believe that Mach has fundamental flaws which prevent the
Hurd from playing out its advantages.  For some of these problems, L4
offers a solution, but L4 comes with its own set of new problems.
This is why you won't find an official statement somewhere, only
personal opinion pieces.

> - There are some official (commonly accepted) strategy for the Hurd
>   development?

There is an official vision.  There are some proven techniques,
especially at the level which is most relevant for Hurd design, which
is the server interfaces.  Thomas, Roland, and the others have shown
that you can implement a POSIX compatible operating system with a
Hurdish design.  This is a powerful demonstration.  Neal and me are
attacking completely different problems at a completely different
level.  The necessary changes are deep, but they will not not be very
big at the surface level, at least not conceptionally.

The Hurd is a huge project, and there is substantial room for parallel
development, without wasting effort.

> - If so... where is such strategy documented? (Not in the Hurd
>   webpages, certainly).

The vision is documented, and you can find out more by lurking and
asking.  I am sure you are very much aware of our goals.  Is there a
technical strategy?  Not a single one, and how could that be?  Nobody
knows.  There is substantial documentation on microkernel design, etc,
but it is not readily accessible.  You won't find a short paper on
"truths about microkernel based multi-server design" because nobody
knows with enough certainty to compile such truths.  We know better
what doesn't work than what does.  We have ideas about what could work
and what problems there are to overcome.  We don't know much more than

> But RMS has been starting to talk about the Hurd as a
> "not essential" GNU package, since there is another Free kernel
> (namely Linux).

And he is right.  The goal of the Free Software Foundation is to
provide a free operating system, not to excel in operating system
design.  Only because we are not essential we are free to delve into
unproven concepts of this scale.

Insofar, I would agree that it was a mistake by the FSF to buy into
unproven, cutting edge operating system design in the early 90s when
the Hurd was started.  If they had chosen the simpler route of writing
a monolithical kernel, things would probably have worked out
differently.  But this is history.  The GNU project did succeed
anyway, independently of the Hurd, and this is a huge benefit.  Now,
legally and politically, the Linux kernel is not without problems, but
nevertheless things have worked out remarkably well for the free
software community.

> The Hurd is starging being considered by the free
> software community (and even the FSF) as a "curiosity" rather than a
> serious project. How sadly.

It's up and down.  We are not worse off than a couple of years ago,
quite the opposite, I believe.  My impression from talking to people
on conferences etc is different, by the way: A lot of interest, and
respect, but also scepsis about the possibility to overcome technical
challenges, which are always acknowledged.  The conversation usually
starts with the other person saying: "I would love to see the Hurd
succeed, but [a thousand reasons why it doesn't seem to be technically


reply via email to

[Prev in Thread] Current Thread [Next in Thread]