Index Home About Blog
Newsgroups: fa.linux.kernel
From: Larry McVoy <lm@bitmover.com>
Subject: Re: [OT] HURD vs Linux/HURD
Original-Message-ID: <20030719172311.GA23246@work.bitmover.com>
Date: Sat, 19 Jul 2003 17:24:28 GMT
Message-ID: <fa.gh5l2jt.1u48uhl@ifi.uio.no>

> - GNU/Hurd, the  whole systems, is  actually GNU tools  (libc, linker,
>   ...)  on top  of the  GNU Hurd  (set of  servers) and  the  GNU Mach
>   microkernel.

Mach wasn't written by GNU, it's a BSD based kernel pried apart into chunks
by people at CMU.

> - GNU Mach 1.x uses drivers from Linux 2.0.36 (IIRC)
>
> - GNU  Mach 2.0  (actually 1.9,  as a  beta version),  uses  the OSKit
>   framework, and such drivers from Linux 2.2.12
>
> - pfinet (our  TCP/IP server)  comes from Linux  2.0 IP stack
>
> I'm not aware  of other use of Linux code inside  the Hurd project, or
> even inside the GNU project, but there may be.

Drivers and networking account for about 50% of the total lines of code.
The bulk of the work in any operating system is typically drivers.  The
generic part of Linux (non-driver, non-file system) is tiny compared to
the rest.

If the Hurd gets its drivers from Linux then it should rightfully be called
Linux/HURD (or Linux/HURD/GNU).

work /tmp/linux-2.5$ bk -rnet cat | wc -l
326411
work /tmp/linux-2.5$ bk -rdrivers cat | wc -l
2605850
work /tmp/linux-2.5$ bk -r cat | wc -l
6618524
--
---
Larry McVoy              lm at bitmover.com          http://www.bitmover.com/lm


Newsgroups: fa.linux.kernel
From: Larry McVoy <lm@bitmover.com>
Subject: Re: [OT] HURD vs Linux/HURD
Original-Message-ID: <20030719181249.GA24197@work.bitmover.com>
Date: Sat, 19 Jul 2003 18:14:25 GMT
Message-ID: <fa.gl692jr.1q44vhn@ifi.uio.no>

On Sat, Jul 19, 2003 at 07:46:54PM +0200, Ga?l Le Mignot wrote:
> Hello Larry!
>
> Sat, 19 Jul 2003 10:23:11 -0700, you wrote:
>
>  >> - GNU/Hurd, the  whole systems, is  actually GNU tools  (libc, linker,
>  >> ...)  on top  of the  GNU Hurd  (set of  servers) and  the  GNU Mach
>  >> microkernel.
>
>  > Mach wasn't written by GNU, it's a BSD based kernel
>
> Totally wrong. You're confusing the  Mach operating system (with UX, a
> BSD-server on top of the  Mach micro-kernel) and the Mach micro-kernel
> itself.

The microkernel part of any reasonable microkernel is tiny.  The QNX
microkernel used to fit in a 4K instruction cache.  To say that the
microkernel is the operating system is ludicrous, that's like say
this series of 5 instructions which happen to get run a lot are the
whole program.

Without the BSD part, you had no operating system, no devices, no nothing.
What you had was a mechanism which can context switch, something that
every first year OS student has written (or should have).

I stand behind the statement and I've read all the Mach papers, all
the Mach code, and lectured on it at little places like Stanford
University.  I'm pretty sure I know what I'm talking about.

> > pried apart into chunks by people at CMU.
>
> GNU Mach is  a modified version of OSF Mach  which is modified version
> of CMU Mach.

Whatever.  That's your label.  Personally, I despise this business of
taking someone else's code and renaming it.  It's not GNU code, the
GNU people didn't write it.

>  > Drivers and networking account for about 50% of the total lines of code.
>  > The bulk of the work in any operating system is typically drivers.  The
>  > generic part of Linux (non-driver, non-file system) is tiny compared to
>  > the rest.
>
> Maybe  for  you,  an  OS  is  drivers.  For  me,  it's  a  design,  an
> architecture, a  philosophy, and a way  to defend a value  that is not
> important for you: Freedom.

I've got a shelf of OS texts, probably close to 90% of all the OS texts
written and I don't recall any of them teaching that you should take other
people's code, rename it, and claim it as your own in the name of freedom.

>  > If the Hurd gets its drivers from Linux then it should rightfully be called
>  > Linux/HURD (or Linux/HURD/GNU).
>
> Stop trolling, thank you.

Hey, you want to spout nonsense then be prepared to be challenged.  I'm
just responding to something that is obviously incorrect, that's not
trolling, that's setting the record straight.  I think it was Dave Miller
who told me the other night that an unchallenged incorrect statement
becomes true by default and I agree.
--
---
Larry McVoy              lm at bitmover.com          http://www.bitmover.com/lm


Newsgroups: fa.linux.kernel
From: Larry McVoy <lm@bitmover.com>
Subject: Re: [OT] HURD vs Linux/HURD
Original-Message-ID: <20030719184519.GB24197@work.bitmover.com>
Date: Sat, 19 Jul 2003 18:47:22 GMT
Message-ID: <fa.gk693bu.1p44upg@ifi.uio.no>

On Sat, Jul 19, 2003 at 07:41:23PM +0100, Christoph Hellwig wrote:
> On Sat, Jul 19, 2003 at 11:12:49AM -0700, Larry McVoy wrote:
> > The microkernel part of any reasonable microkernel is tiny.
>
> And who says Mach is a reasonable microkernel :)

Yup, more like a maxikernel :)

That was my reaction on reading the code years ago and it hasn't changed.
I used to know one of the main guys who did the QNX microkernel (Dan
Hildebrandt, RIP 1998) and he talked about how a real microkernel was
never touched by more than 3 people and each of them spent as much
time removing stuff as adding it.

Mach is kinda on the bloated side, I always questioned the wisdom of
the GNU HURD being based on Mach, seemed like a bad call.  But then,
unless you have an extremely well controlled dev team, any micro kernel
is a bad call, it's going to bloat out.
--
---
Larry McVoy              lm at bitmover.com          http://www.bitmover.com/lm


Newsgroups: fa.linux.kernel
From: Larry McVoy <lm@bitmover.com>
Subject: Re: [OT] HURD vs Linux/HURD
Original-Message-ID: <20030719220306.GE24197@work.bitmover.com>
Date: Sat, 19 Jul 2003 22:05:10 GMT
Message-ID: <fa.glln2bs.1rkevpm@ifi.uio.no>

> Stop lying. No one at the GNU project ever claimed a code to be his if
> he didn't  write it.

Nonsense.  Go look at the set of code actually funded by the FSF and it
is tiny.  The FSF tries to get everyone to sign over their copyright to
the FSF so they can "protect" the code and then they rename it to GNU
this that or the other thing.  Start reading those copyrights you are
talking about, of the set of things described as GNU something, I would
guess than less than 1% of it was paid for by the FSF.  The rest of it
is all stuff they slapped their name on after convincing people to sign
over copyrights.
--
---
Larry McVoy              lm at bitmover.com          http://www.bitmover.com/lm


Newsgroups: fa.linux.kernel
From: "Theodore Ts'o" <tytso@mit.edu>
Subject: Re: [OT] HURD vs Linux/HURD
Original-Message-ID: <20030720000716.GA1085@think>
Date: Sun, 20 Jul 2003 02:57:11 GMT
Message-ID: <fa.c84shfo.18huob4@ifi.uio.no>

On Sat, Jul 19, 2003 at 04:03:55PM +0100, John Bradford wrote:
> > Given that large chunks of HURD come from Linux, please refer to it as
> > Linux/HURD.
>
> What HURD code comes from Linux?  GNU/Mach uses code from Linux, but
> not HURD as far as I know.

As far as I know, HURD is using ext2fs code.  It should definitely be
called HURD/Linux.  :-)

						- Ted

Index Home About Blog