Re:Poor article & microkernels arch. are dead
on
JKH on OS X
·
· Score: 1
> My own *first hand* experience completely contradicts your opinion.
Well, my opinion was true experience. We had
also NeXT workstation, that run on 86040 and
also some Intel Boxes.
And of course, in the same room the were also
Solaris and Linux-Boxes with 8MB.
Running those systems with 8MB is not a fun at
all, and maybe you already tried to run some
application. That was really time expense looking
at the spinning disk in front of you.
Anyway, the some Networking applications were
faster and more stable on Solaris and the
pre1.0 versions of Linux.
Anyway, the confort on NeXT was a real advantage,
in comparision with the OpenLook or the naked-X
on Linux.
> You use the term "emulation". This is misleading
You are right, because the correct term of this
is "personality". Anyway, it waste its time
on transforming the calls from BSD to the
underlying mikorkernel architecture.
And especially this is time expense. Running
a BYTE-Bench on the top of the BSD personality
shows the efficiency of Linux, because the penguin
is 2 times faster !!
> Ths is not "emulation" but just good software design. The fact that Linus don't get it is another problem.
> Mach is much more advanced more clean and easier to use good defined abstraction
The mostly quoted word from a mk-Fan is the good
"architecture". You may be right, but nobody is
interested with this, because it is the unique
selling point of a microkernel !!! But a slow
performance, that is recognized !!!
> Mach is also much more flexible than monolyth
> systems. For instance it provide NKE (Network
> kernel extension), the ability to add, remove
> or configure network module or entire stack
> while the kernel is running.
> Show me how to do that on other systems !
Yam, here is one !! Never heard of kernel
modules, really. You can do exactly the same
on monolytic implementations. Especially these
modules, that can be loaded und unloaded
while the kernel is running, can be parametrized
during runtime.
This feature, also the others you mentioned
and more can be gazed in Linux and partially
also with Solaris.
> Please, becomes informed about mach before commenting too much
Yes, please type > insmod brain
Re:Poor article & microkernels arch. are dead
on
JKH on OS X
·
· Score: 1
Your argumentation is good, from Apple's point
of view !!
Hubbards also dreams of topping Mircosoft.
You can only win the match, iff you have an
competing product. Now see what has happend
to the MacOS-X. Again the birth of a
hydrocephalus !!!
They used a OS-design, that actually is only used
to run big serves, or did you know a use-case,
where the OSF/1 was installed as a desktop
system. Normally, they are used to run
Server-application, actually with decreasing
interest !!!
OK, we have got big OS, and now also a big
GUI component, that runs mostly old
Mac-applications, through a emulation layer.
And, this will be the most use-case, as you
also remaked.
So what do we have in reality:
A mikro-kernel, that emulates a monolyth and
a emulation, that bring down (up ??) off-the-
shelf apple application on a nice workstation.
Sorry, these process contains too much emulation
steps to say, that the performance difference
will be "not be an important issue".
We also did not talk about the price of Apple
hardware !!
> You're making mountains out of mole hills.
Better making moutains, than allowing Apple to
making a bigger kloof in my purse !!!!
Re:Poor article & microkernels arch. are dead
on
JKH on OS X
·
· Score: 1
> It's hard to imagine someone reading an
> article and so completely Not Getting It.
> Congratulations.
I did after all. After reading this article,
i ask myself, for what is it good for, including
that arcticle. The contents of this article celebrates the combination of a confortable GUI
and a well known BSD like interface.
I simply remarked, that this is not all, there
are also some disadvantages. Besides there
are better solutions for a good GUI and
BSD-API.
I like lean solutions. So MacOS-X is not
attractive to me at all.
Last not least. Reading an article is one thing.
But understanding and reflecting is more !!!
Re:Poor article & microkernels arch. are dead
on
JKH on OS X
·
· Score: 1
> Actually this is quite ignorant, many of the
> features needed to deliver the higherlevel
> services depend extensively upon the mach
> message passing (wich is extremely high
> prefromance).
No, absolutly not !! A simple UNIX-byte bench
shows the difference. A Mach system is half
as fast like a modern monolythic OS like Linux.
You need specific instructions, to profit
from a message-passing system (like transputers).
Today's processors are not well designed for
to support the message passing architectures,
so they are slower !!
Re:Poor article & microkernels arch. are dead
on
JKH on OS X
·
· Score: 1
> Take QNX RTP/4 for example. It can
This is a different market !!! OK, i accept, that
QNX may be a good solution for real-time
application.
But hardware installation is not a feature, that
is only available on mikro-kernel architetures.
The old SVR4 based reliant system made by
pyramid (now SIEMENS) offered a similar feature
for devices (i think also the CPU). You can be
sure, that this will be also available for
LINUX, cause this kernel support for late-device
driver binding, so called kernel-modules.
> It's NetBSD! Not FreeBSD
OK, sometime i get confused with the exisiting
BSD implementations.
> On the MainFrame issue... S
Actually, i can not! Be sure in future you will
find some users of Linux-390. IBM is willing to
bring this solution to the customer, by offering
a low price CPU bundeld with a Linux distribution.
They want to sell hardware, and the problem is
the license-model of third party software
suppliers. In order to bring down the high prises
they support LINUX and -you can be sure- customers
will buy it. They are looking forward to have
more choices in future !!!
((()))
Ahem, you also recognized, that we got out of
the origin scope of discussion. Maybe you can
life with a more moderate argument, that there
exist some use-cases for mikro-kernel architecture, i.e. real-time...
... and there is also a product available called
BeOS, that seems to be also a micro-kernel
implementation with remarkable advantages over
traditional UNIX'e.
But, my old these still remains !!! The
MacOS-X is only used (and also presented within
the article) as a symbiosis between a modern
GUI and a well known BSD interface. OK, and my
stricture is found upon the conclusion, that
the use of Mach is a architecual overkill, with
no real gain(profit) for the customer.
Instead of this, he has to pay for the costly
architecutr by losing memory ressources and
performance.
Why did not use Apple a simple (and faster)
OS like Linux or xxxBSD ???
Re:Poor article & microkernels arch. are dead
on
JKH on OS X
·
· Score: 1
Hy #172771 Info,
(1) ------------------------------------------
i know what's a mircokernel and the motivation
to develop one. I just changed the point of
view and tried to find the the practical impact
(or let's say) the advantages for the user of
a MC system like the new Mac-OS.
You may sit in front of your favourate GUI and
may say to yourself: "Great, while i am waiting
for the completion of the invoked request, the call may run through a proper OS-design"
This idea is not beeing meant to be cynical, but
i like to point out, that there exist a more
pragmatic view of this topic. The design issue
is only fascinating the tech people, but does not
offer a profit for a real use-case.
Another point is the fact, that today's
monolythic systems performed some architectual
improvements like kernel modules or the
fantastic STREAMS architecture (oh i really
like this!!)
(2) -----------------------------------------------
OK, FreeBSD is also very portable, but it is also
a monolythic kernel design, like LINUX.
(is FreeBSD also available on Mainframes ??)
But do not assume, that Mach3 gains the
protability of FreeBSD !! At last, please only
count real distributions not possible hacks !!
(3)--------------------------------------------
Please be more concrete. On Mac-X almost everybody use the BSD personality. There does not exist any other personality, so they spent
their time in emulation of a BSD interface.
This cause a performance loss and nobody
comforts me with the idea, that i could use
a actually not existing personality or any
feature, that i do not need !!
Ok, your argument with QNX is good, QNX may
be (or is) a proper solution for real-time
environments or embedded systems.
But the internet-applications, desktops systems
or middleware environments are the focus
of our discussion here, and this is Apples
focus use-case for their new OS !!!
(xx) ------------------------------------------
Please also take a look into Hubbards paper.
He pointed out the advantages of the Mac-X OS
and put focus on the GUI and the UNIX layer.
Hey and i said, why should i carry out the
burden of a performance loss, they could have
also use LINUX or FreeBSD instead of this
hydrocephalus.
Re:Poor article & microkernels arch. are dead
on
JKH on OS X
·
· Score: 1
I know of some advantages of a microkernel
architecture. But look at these adavantages,
this is -my critisim- only a design issue.
From the practical point of view, there are only
disadvantages, esp. higher ressource consumption
with regard to the memory capacity and processor
power.
Last not least, the most portable OS is Linux.
Hard to say, but this is the reality, and last
not least it is faster.
Nobody takes profit from the "adavantages" of
the microkernel architecture, because erverybody
uses a SVR4 or a BSD personality. The Mach
seems only to be designed, to emulate off
the self environments, thats not an advantage !!!
Re:poor english aside, you're a moron
on
JKH on OS X
·
· Score: 1
Moron... , thanks for that compliment !!!
Beside i am working hard to improve my english,
the original message still remain, thats a fact.
Comparring NeXTStep with LINUX is right, cause
the Mach2.5 based was slower and less stable than
Linux, that's a fact.
I have some experience with OS, an Mach2.5 like
Mach3.0 (and NT too) use internally message
passing. That's a fact too.
You may pass the calls to an ordinary API, but
the basic foundation runs with messages, because
your BSD are a thinking to work with, is a simple
service a so called personality, that's the
truth.
I am a programmer and a design architect. The
KDE Api is known to be excellent, and i like
it. The KDE2 has in contrast to the Mac-GUI applications and a better acceptance.
From the time i was using and programming the
NeXT-Step GUI (that seems to be a preceding
system) i must say, that KDE looks better to me.
The technical internals are very exciting,
esspecially the DCOP architecture, it seems
that you never took a look at it !!
Oh, sad to say to you, that the functionality
of Quartz also resides in Gnome, and maybe
-i am not sure- also in KDE2. PDF and and PS
is absoultly not a problem, so why are you so
upset.
It seems that you did not understand what i
meant. This was a OS comparison and a more
realistic view to the thousand of features
that are usesless ballast in OS-X, that makes
it slow.
------
Nice to meet the hard-core marketing division of
Apple here. Maybe you try a more pollite and
sound mode of expression !!
Poor article & microkernels arch. are dead
on
JKH on OS X
·
· Score: 1
I expected a better quality of that article, but
unfortunately the author is only scratching
the very surface of the Mac-X, getting excitet
by simple eye-catcher. He absolutly fails to take
a critical look behind the curtains, so this
publication is simple propaganda.
Let me explain wy.
=(1)= The author seems not to be interested in the
kernel design, especially the differences between
a monolythic and a mikrokernel architecture. He
seemed, willing to sell us this system as a
feature mixture of FreeBSD and Mach3, ignoring
the adantages and disadvantages for the user.
He failed to explain, that a microkernel is
a message passing architecture, that deals with
mutlithreading, serveral basic kernel serices,
that are running as a kernel-process.
A simple Mach-Mirkokernel comes with very basic
functionality, on this ground there a futher
so called personalities, that provides a known
OS interface to the running applications.
The FreeBSD Code is split up into a Kernel-Driver
Part, and a BSD-personality, that provides a
system, that is similar to the well known BSD-API.
So up to know, there is no real advantage over a
traditional BSD System, because most Implementations
only suppots one personality, that is common
with a traditional off-the-self OS, so the arcticle
did not left out any advantages, cause there exist not
any !!!
All the time, UNIX-microkernels OS'ies do not
profit from their adavantages, they almost spent
their time, emulating known UNIX layer, congratulation !!
Now, lets face the disatvantages, and there are two
of very importance:
1) Memory usage: Hey in oder of keeping your inefficient
circus running, you need also memory. So thats a
real disadvantage !!!
2) Efficiency, Performance: Yeh, and also you need
computing ressources, to get these inefficent operatins
running, to get you simple UNIX-call running.
So wy, and that is the question Jorda barly forgot
to ask us, why should it not be better to use the
original ????? OK, he got excitet with the graphical
representation of the start-up, but this is only
a nice gimmick, but not a quality feature !!!
A simple performance comparison makes clear what
i mean. Some years ago, i had a Intel-i486 based
Workstation with NextStep (Mach2.5) and a similar
Linx-Box on a similar hardware configuration. With
the Byte-Becnmarks, the Linux box outperforms the
NextStep platform by 2 times !!!
Hey Jordan, thst's an argument !!!
=(2)= Let's see what he thinks abount GUI.
Nice to recognize, that he believes in the open source
community, especially he mentioned KDE and Gnome. The situation
can be better analysed by a visit of a professional computer
exposition like the Systems2000 (in Munich, Germany/Europe).
Fact is the acceptance of the KDE-GUI (more than Gnome) several
exposition-booth of Software Tools AND some Application vendors.
Apple-GUI did not play a role on this exposition, in contrast
to especially KDE. So Jordan's GUI-argument is histroy, absolutely.
No question. Apple-Boxes have a good design, are reliabel and offer
a good performance. This is good platform for me, to install Linux
with KDE2 on it !!! (or BSD iff you really want this)
Slashdot Mirror
You can read it from my lips !
> My own *first hand* experience completely contradicts your opinion.
Well, my opinion was true experience. We had
also NeXT workstation, that run on 86040 and
also some Intel Boxes.
And of course, in the same room the were also
Solaris and Linux-Boxes with 8MB.
Running those systems with 8MB is not a fun at
all, and maybe you already tried to run some
application. That was really time expense looking
at the spinning disk in front of you.
Anyway, the some Networking applications were
faster and more stable on Solaris and the
pre1.0 versions of Linux.
Anyway, the confort on NeXT was a real advantage,
in comparision with the OpenLook or the naked-X
on Linux.
> You use the term "emulation". This is misleading
You are right, because the correct term of this
is "personality". Anyway, it waste its time
on transforming the calls from BSD to the
underlying mikorkernel architecture.
And especially this is time expense. Running
a BYTE-Bench on the top of the BSD personality
shows the efficiency of Linux, because the penguin
is 2 times faster !!
> Ths is not "emulation" but just good software design. The fact that Linus don't get it is another problem.
> Mach is much more advanced more clean and easier to use good defined abstraction
The mostly quoted word from a mk-Fan is the good
"architecture". You may be right, but nobody is
interested with this, because it is the unique
selling point of a microkernel !!! But a slow
performance, that is recognized !!!
> Mach is also much more flexible than monolyth
> systems. For instance it provide NKE (Network
> kernel extension), the ability to add, remove
> or configure network module or entire stack
> while the kernel is running.
> Show me how to do that on other systems !
Yam, here is one !! Never heard of kernel
modules, really. You can do exactly the same
on monolytic implementations. Especially these
modules, that can be loaded und unloaded
while the kernel is running, can be parametrized
during runtime.
This feature, also the others you mentioned
and more can be gazed in Linux and partially
also with Solaris.
> Please, becomes informed about mach before commenting too much
Yes, please type > insmod brain
Your argumentation is good, from Apple's point
of view !!
Hubbards also dreams of topping Mircosoft.
You can only win the match, iff you have an
competing product. Now see what has happend
to the MacOS-X. Again the birth of a
hydrocephalus !!!
They used a OS-design, that actually is only used
to run big serves, or did you know a use-case,
where the OSF/1 was installed as a desktop
system. Normally, they are used to run
Server-application, actually with decreasing
interest !!!
OK, we have got big OS, and now also a big
GUI component, that runs mostly old
Mac-applications, through a emulation layer.
And, this will be the most use-case, as you
also remaked.
So what do we have in reality:
A mikro-kernel, that emulates a monolyth and
a emulation, that bring down (up ??) off-the-
shelf apple application on a nice workstation.
Sorry, these process contains too much emulation
steps to say, that the performance difference
will be "not be an important issue".
We also did not talk about the price of Apple
hardware !!
> You're making mountains out of mole hills.
Better making moutains, than allowing Apple to
making a bigger kloof in my purse !!!!
> It's hard to imagine someone reading an
> article and so completely Not Getting It.
> Congratulations.
I did after all. After reading this article,
i ask myself, for what is it good for, including
that arcticle. The contents of this article celebrates the combination of a confortable GUI
and a well known BSD like interface.
I simply remarked, that this is not all, there
are also some disadvantages. Besides there
are better solutions for a good GUI and
BSD-API.
I like lean solutions. So MacOS-X is not
attractive to me at all.
Last not least. Reading an article is one thing.
But understanding and reflecting is more !!!
> Actually this is quite ignorant, many of the > features needed to deliver the higherlevel > services depend extensively upon the mach > message passing (wich is extremely high > prefromance). No, absolutly not !! A simple UNIX-byte bench shows the difference. A Mach system is half as fast like a modern monolythic OS like Linux. You need specific instructions, to profit from a message-passing system (like transputers). Today's processors are not well designed for to support the message passing architectures, so they are slower !!
> Take QNX RTP/4 for example. It can
...
This is a different market !!! OK, i accept, that
QNX may be a good solution for real-time
application.
But hardware installation is not a feature, that
is only available on mikro-kernel architetures.
The old SVR4 based reliant system made by
pyramid (now SIEMENS) offered a similar feature
for devices (i think also the CPU). You can be
sure, that this will be also available for
LINUX, cause this kernel support for late-device
driver binding, so called kernel-modules.
> It's NetBSD! Not FreeBSD
OK, sometime i get confused with the exisiting
BSD implementations.
> On the MainFrame issue... S
Actually, i can not! Be sure in future you will
find some users of Linux-390. IBM is willing to
bring this solution to the customer, by offering
a low price CPU bundeld with a Linux distribution.
They want to sell hardware, and the problem is
the license-model of third party software
suppliers. In order to bring down the high prises
they support LINUX and -you can be sure- customers
will buy it. They are looking forward to have
more choices in future !!!
((()))
Ahem, you also recognized, that we got out of
the origin scope of discussion. Maybe you can
life with a more moderate argument, that there
exist some use-cases for mikro-kernel architecture, i.e. real-time
... and there is also a product available called
BeOS, that seems to be also a micro-kernel
implementation with remarkable advantages over
traditional UNIX'e.
But, my old these still remains !!! The
MacOS-X is only used (and also presented within
the article) as a symbiosis between a modern
GUI and a well known BSD interface. OK, and my
stricture is found upon the conclusion, that
the use of Mach is a architecual overkill, with
no real gain(profit) for the customer.
Instead of this, he has to pay for the costly
architecutr by losing memory ressources and
performance.
Why did not use Apple a simple (and faster)
OS like Linux or xxxBSD ???
Hy #172771 Info,
(1) ------------------------------------------
i know what's a mircokernel and the motivation
to develop one. I just changed the point of
view and tried to find the the practical impact
(or let's say) the advantages for the user of
a MC system like the new Mac-OS.
You may sit in front of your favourate GUI and
may say to yourself: "Great, while i am waiting
for the completion of the invoked request, the call may run through a proper OS-design"
This idea is not beeing meant to be cynical, but
i like to point out, that there exist a more
pragmatic view of this topic. The design issue
is only fascinating the tech people, but does not
offer a profit for a real use-case.
Another point is the fact, that today's
monolythic systems performed some architectual
improvements like kernel modules or the
fantastic STREAMS architecture (oh i really
like this!!)
(2) -----------------------------------------------
OK, FreeBSD is also very portable, but it is also
a monolythic kernel design, like LINUX.
(is FreeBSD also available on Mainframes ??)
But do not assume, that Mach3 gains the
protability of FreeBSD !! At last, please only
count real distributions not possible hacks !!
(3)--------------------------------------------
Please be more concrete. On Mac-X almost everybody use the BSD personality. There does not exist any other personality, so they spent
their time in emulation of a BSD interface.
This cause a performance loss and nobody
comforts me with the idea, that i could use
a actually not existing personality or any
feature, that i do not need !!
Ok, your argument with QNX is good, QNX may
be (or is) a proper solution for real-time
environments or embedded systems.
But the internet-applications, desktops systems
or middleware environments are the focus
of our discussion here, and this is Apples
focus use-case for their new OS !!!
(xx) ------------------------------------------
Please also take a look into Hubbards paper.
He pointed out the advantages of the Mac-X OS
and put focus on the GUI and the UNIX layer.
Hey and i said, why should i carry out the
burden of a performance loss, they could have
also use LINUX or FreeBSD instead of this
hydrocephalus.
I know of some advantages of a microkernel
architecture. But look at these adavantages,
this is -my critisim- only a design issue.
From the practical point of view, there are only
disadvantages, esp. higher ressource consumption
with regard to the memory capacity and processor
power.
Last not least, the most portable OS is Linux.
Hard to say, but this is the reality, and last
not least it is faster.
Nobody takes profit from the "adavantages" of
the microkernel architecture, because erverybody
uses a SVR4 or a BSD personality. The Mach
seems only to be designed, to emulate off
the self environments, thats not an advantage !!!
Moron ... , thanks for that compliment !!!
Beside i am working hard to improve my english,
the original message still remain, thats a fact.
Comparring NeXTStep with LINUX is right, cause
the Mach2.5 based was slower and less stable than
Linux, that's a fact.
I have some experience with OS, an Mach2.5 like
Mach3.0 (and NT too) use internally message
passing. That's a fact too.
You may pass the calls to an ordinary API, but
the basic foundation runs with messages, because
your BSD are a thinking to work with, is a simple
service a so called personality, that's the
truth.
I am a programmer and a design architect. The
KDE Api is known to be excellent, and i like
it. The KDE2 has in contrast to the Mac-GUI applications and a better acceptance.
From the time i was using and programming the
NeXT-Step GUI (that seems to be a preceding
system) i must say, that KDE looks better to me.
The technical internals are very exciting,
esspecially the DCOP architecture, it seems
that you never took a look at it !!
Oh, sad to say to you, that the functionality
of Quartz also resides in Gnome, and maybe
-i am not sure- also in KDE2. PDF and and PS
is absoultly not a problem, so why are you so
upset.
It seems that you did not understand what i
meant. This was a OS comparison and a more
realistic view to the thousand of features
that are usesless ballast in OS-X, that makes
it slow.
------
Nice to meet the hard-core marketing division of
Apple here. Maybe you try a more pollite and
sound mode of expression !!
I expected a better quality of that article, but
unfortunately the author is only scratching
the very surface of the Mac-X, getting excitet
by simple eye-catcher. He absolutly fails to take
a critical look behind the curtains, so this
publication is simple propaganda.
Let me explain wy.
=(1)= The author seems not to be interested in the
kernel design, especially the differences between
a monolythic and a mikrokernel architecture. He
seemed, willing to sell us this system as a
feature mixture of FreeBSD and Mach3, ignoring
the adantages and disadvantages for the user.
He failed to explain, that a microkernel is
a message passing architecture, that deals with
mutlithreading, serveral basic kernel serices,
that are running as a kernel-process.
A simple Mach-Mirkokernel comes with very basic
functionality, on this ground there a futher
so called personalities, that provides a known
OS interface to the running applications.
The FreeBSD Code is split up into a Kernel-Driver
Part, and a BSD-personality, that provides a
system, that is similar to the well known BSD-API.
So up to know, there is no real advantage over a
traditional BSD System, because most Implementations
only suppots one personality, that is common
with a traditional off-the-self OS, so the arcticle
did not left out any advantages, cause there exist not
any !!!
All the time, UNIX-microkernels OS'ies do not
profit from their adavantages, they almost spent
their time, emulating known UNIX layer, congratulation !!
Now, lets face the disatvantages, and there are two
of very importance:
1) Memory usage: Hey in oder of keeping your inefficient
circus running, you need also memory. So thats a
real disadvantage !!!
2) Efficiency, Performance: Yeh, and also you need
computing ressources, to get these inefficent operatins
running, to get you simple UNIX-call running.
So wy, and that is the question Jorda barly forgot
to ask us, why should it not be better to use the
original ????? OK, he got excitet with the graphical
representation of the start-up, but this is only
a nice gimmick, but not a quality feature !!!
A simple performance comparison makes clear what
i mean. Some years ago, i had a Intel-i486 based
Workstation with NextStep (Mach2.5) and a similar
Linx-Box on a similar hardware configuration. With
the Byte-Becnmarks, the Linux box outperforms the
NextStep platform by 2 times !!!
Hey Jordan, thst's an argument !!!
=(2)= Let's see what he thinks abount GUI.
Nice to recognize, that he believes in the open source
community, especially he mentioned KDE and Gnome. The situation
can be better analysed by a visit of a professional computer
exposition like the Systems2000 (in Munich, Germany/Europe).
Fact is the acceptance of the KDE-GUI (more than Gnome) several
exposition-booth of Software Tools AND some Application vendors.
Apple-GUI did not play a role on this exposition, in contrast
to especially KDE. So Jordan's GUI-argument is histroy, absolutely.
No question. Apple-Boxes have a good design, are reliabel and offer
a good performance. This is good platform for me, to install Linux
with KDE2 on it !!! (or BSD iff you really want this)
That's all folks !!!