Interesting paper about micro kernel design

bridged with qnx.cafe
Post Reply
Armin Steinhoff

Interesting paper about micro kernel design

Post by Armin Steinhoff » Fri Oct 28, 2005 1:22 pm

This paper describs the design of the new 'true' micro kernel of
Minix 3.0

http://www.minix3.org/doc/herder_thesis.pdf

--Armin

John Nagle

Re: Interesting paper about micro kernel design

Post by John Nagle » Fri Oct 28, 2005 4:47 pm

Armin Steinhoff wrote:
This paper describs the design of the new 'true' micro kernel of
Minix 3.0

http://www.minix3.org/doc/herder_thesis.pdf

--Armin
It's a mediocre paper about a mediocre design. Which is sad.

That paper is by a master's student who took the drivers
out of the existing kernel and added wierd system calls so they could
do the things they were already doing. The architecture is far
inferior to that of QNX 6. No threads. Networking and file systems
still seem to be in the kernel. Messaging/scheduler interaction
has not been worked out. No modern devices (USB, FireWire,
etc.) are supported. The security is no better than that of
classic Unix/Linux.

Fundamentally, the problem is that it's a retrofit of a microkernel
architecture. That never works. See Mach.

I was hoping for something better.

John Nagle

Armin Steinhoff

Re: Interesting paper about micro kernel design

Post by Armin Steinhoff » Fri Oct 28, 2005 5:22 pm

John Nagle wrote:
Armin Steinhoff wrote:


This paper describs the design of the new 'true' micro kernel of
Minix 3.0

http://www.minix3.org/doc/herder_thesis.pdf

--Armin


It's a mediocre paper about a mediocre design. Which is sad.
The design of QNX6 and Minix 3.0 are based on the multi server
architecure ... so you mean this design is a mediocre one?
That paper is by a master's student who took the drivers
out of the existing kernel and added wierd system calls so they could
do the things they were already doing. The architecture is far
inferior to that of QNX 6. No threads. Networking and file systems
still seem to be in the kernel.
Wrong ... Networking and file systems are based on servers and drivers
running in users space. That's new for Minix 3.0 ...
Messaging/scheduler interaction has not been worked out.
.... but is implemented. Synchronous message passing ...
No modern devices (USB, FireWire,
etc.) are supported.
Minix 3.0 is a fresh release ... there will lots of things follow.

The security is no better than that of
classic Unix/Linux.

Fundamentally, the problem is that it's a retrofit of a microkernel
architecture. That never works. See Mach.
You don't know the history of Minix .... this system works.

--Armin
I was hoping for something better.

John Nagle

Malte Mundt

Re: Interesting paper about micro kernel design

Post by Malte Mundt » Tue Nov 08, 2005 10:28 am

Who needs Minix when we got QNX? :)

"Armin Steinhoff" <a-steinhoff@web.de> schrieb im Newsbeitrag
news:djtmgt$mg1$1@inn.qnx.com...
John Nagle wrote:
Armin Steinhoff wrote:


This paper describs the design of the new 'true' micro kernel of
Minix 3.0

http://www.minix3.org/doc/herder_thesis.pdf

--Armin


It's a mediocre paper about a mediocre design. Which is sad.

The design of QNX6 and Minix 3.0 are based on the multi server
architecure ... so you mean this design is a mediocre one?

That paper is by a master's student who took the drivers
out of the existing kernel and added wierd system calls so they could
do the things they were already doing. The architecture is far
inferior to that of QNX 6. No threads. Networking and file systems
still seem to be in the kernel.

Wrong ... Networking and file systems are based on servers and drivers
running in users space. That's new for Minix 3.0 ...

Messaging/scheduler interaction has not been worked out.

... but is implemented. Synchronous message passing ...

No modern devices (USB, FireWire,
etc.) are supported.

Minix 3.0 is a fresh release ... there will lots of things follow.

The security is no better than that of
classic Unix/Linux.

Fundamentally, the problem is that it's a retrofit of a microkernel
architecture. That never works. See Mach.

You don't know the history of Minix .... this system works.

--Armin


I was hoping for something better.

John Nagle

Mario Charest

Re: Interesting paper about micro kernel design

Post by Mario Charest » Tue Nov 08, 2005 1:43 pm

"Malte Mundt" <mmundt@qnx.de> wrote in message
news:dkpun4$rkc$1@inn.qnx.com...
Who needs Minix when we got QNX? :)
You must be kidding, right?

Armin Steinhoff

Re: Interesting paper about micro kernel design

Post by Armin Steinhoff » Tue Nov 08, 2005 6:38 pm

Malte Mundt wrote:
Who needs Minix when we got QNX? :)
The issue isn't to choose between Minix or QNX.

The interesting point is just the design of the
micro kernel of Minix 3.0.

And ... you have access to the sources!

--Armin



"Armin Steinhoff" <a-steinhoff@web.de> schrieb im Newsbeitrag
news:djtmgt$mg1$1@inn.qnx.com...

John Nagle wrote:

Armin Steinhoff wrote:


This paper describs the design of the new 'true' micro kernel of
Minix 3.0

http://www.minix3.org/doc/herder_thesis.pdf

--Armin


It's a mediocre paper about a mediocre design. Which is sad.

The design of QNX6 and Minix 3.0 are based on the multi server
architecure ... so you mean this design is a mediocre one?


That paper is by a master's student who took the drivers
out of the existing kernel and added wierd system calls so they could
do the things they were already doing. The architecture is far
inferior to that of QNX 6. No threads. Networking and file systems
still seem to be in the kernel.

Wrong ... Networking and file systems are based on servers and drivers
running in users space. That's new for Minix 3.0 ...


Messaging/scheduler interaction has not been worked out.

... but is implemented. Synchronous message passing ...


No modern devices (USB, FireWire,
etc.) are supported.

Minix 3.0 is a fresh release ... there will lots of things follow.

The security is no better than that of

classic Unix/Linux.

Fundamentally, the problem is that it's a retrofit of a microkernel
architecture. That never works. See Mach.

You don't know the history of Minix .... this system works.

--Armin


I was hoping for something better.

John Nagle


Post Reply

Return to “qnx.cafe”