Cheap Amigas and costly QNX
John Bayko
Joe Cosby <joec...@lontyb.seatac.net> wrote:
>** To reply in e-mail, remove "lontyb." from address **
>
>Clyde Coffey hunched over his computer, typing feverishly;
>thunder crashed, Clyde Coffey laughed madly, then wrote:
>> On 10 Feb 1999 00:20:01 GMT, ba...@aristotle.cs.uregina.ca (John
>> Bayko) wrote:
>>
>>> Basically, the Amiga API (as much as possible), the Amiga GUI (or
>>>some adaptation of it), and user-oriented utilities ported from AmigaOS.
>>>QNX is an embedded OS, meaning almost all software for it is either
>>>developer-oriented, or custom made for individual instalations. It
>>>will be used essentially to replace the AmigaOS kernel.
>>
>> Okay, then, what does the QNX kernel offer that the Amiga kernel
>> doesn't? They work in pretty much the same way. What is gained by
>> switching?
>
>VMem, memory protection, EXTREME scalability and
>flexibility.
>
>I don't have a URL handy, but look around for
>accounts of the demo at the Cologne show.
"http://www.qnx.com/" would be a good place to start.
--
John Bayko (Tau).
ba...@cs.uregina.ca
http://www.cs.uregina.ca/~bayko
David Cardinal
> >> Bayko) wrote:
> >>
> >> Okay, then, what does the QNX kernel offer that the Amiga kernel
> >> doesn't? They work in pretty much the same way. What is gained by
> >> switching?
What does the QNX microkernel offer that the *BeOS* microkernel doesn't?
Weren't they both being considered by Amiga? How did they pick QNX?
I see the title of this thread says "costly QNX"... How much does QNX cost? I
can't find that info on the QNX website. Please email me at
techn...@geocities.com.
Aram Iskenderian
In article <36CE2F62...@geocities.com>
David Cardinal <techn...@geocities.com> wrote:
>> >> On 10 Feb 1999 00:20:01 GMT, ba...@aristotle.cs.uregina.ca (John
>> >> Bayko) wrote:
>> >>
>> >> Okay, then, what does the QNX kernel offer that the Amiga kernel
>> >> doesn't? They work in pretty much the same way. What is gained by
>> >> switching?
>
>What does the QNX microkernel offer that the *BeOS* microkernel doesn't?
Real time operation.
>Weren't they both being considered by Amiga?
Be was approached first.
>How did they pick QNX?
Be got greedy and pulled out of the deal.
Qnx was considered even at the days of Commodore, according to Dr. Alan
Havemose head of Engineering and development at Amiga Inc.
>I see the title of this thread says "costly QNX"... How much does QNX cost? I
>can't find that info on the QNX website.
I'm not sure exactly how much, but it was said several times that the
cheapest is $800 and can cost up to $3000 depending on the setup.
--
Aram Iskenderian.
ar...@ix.netcom.com, ar...@mailexcite.com.
Steve Giovenella
> > >> On 10 Feb 1999 00:20:01 GMT, ba...@aristotle.cs.uregina.ca (John
> > >> Bayko) wrote:
> > >>
> > >> Okay, then, what does the QNX kernel offer that the Amiga kernel
> > >> doesn't? They work in pretty much the same way. What is gained by
> > >> switching?
>
> What does the QNX microkernel offer that the *BeOS* microkernel doesn't?
> Weren't they both being considered by Amiga? How did they pick QNX?
All were reviewed by Amiga Inc. QNX was picked as the basis for the
new kernal for many reasons. Since a lot of people have been asking,
I have included a "Next Generation" Amiga section in the AmigaOS
drawer on my web site. Please have a look for more details.
-Steve.
-http://www.amigapro.com
-The website for the Amiga performance enthusiast.
A. Steinhoff
>
>On Sat, 20 Feb 1999 00:43:31 -0300,
>In article <36CE2F62...@geocities.com>
>David Cardinal <techn...@geocities.com> wrote:
>
>>> >> Bayko) wrote:
>>> >>
>>> >> Okay, then, what does the QNX kernel offer that the Amiga kernel
>>> >> doesn't? They work in pretty much the same way. What is gained by
>>> >> switching?
>>
>>What does the QNX microkernel offer that the *BeOS* microkernel doesn't?
>
>
>>Weren't they both being considered by Amiga?
>
>
>>How did they pick QNX?
>
>
>Qnx was considered even at the days of Commodore, according to Dr. Alan
>Havemose head of Engineering and development at Amiga Inc.
>
>>I see the title of this thread says "costly QNX"... How much does QNX cost? I
>>can't find that info on the QNX website.
>
>I'm not sure exactly how much, but it was said several times that the
>cheapest is $800 and can cost up to $3000 depending on the setup.
>
More or less correct for QNX4 ... but I don't believe that the Amiga Inc.
will buy QNX/Neutrino in single licenses :-)
I'm sure that theire license agreement will give Amiga Inc. a very reasonable
low price.
Armin
Geoff Oltmans
> > What does the QNX microkernel offer that the *BeOS* microkernel doesn't?
> > Weren't they both being considered by Amiga? How did they pick QNX?
>
> All were reviewed by Amiga Inc. QNX was picked as the basis for the
> new kernal for many reasons. Since a lot of people have been asking,
> I have included a "Next Generation" Amiga section in the AmigaOS
> drawer on my web site. Please have a look for more details.
Not the least of which I'm sure was greater exclusivity with QNX. BeOS
has already been promised as a competing product for MS Windows/MacOS on
PC's and Macs.
How many end-users do YOU know that use QNX? I can't think of any.
*Geoff!*
Luca Filipozzi
says...
> > It wouldn't surprise me if _most_ people have "used" QNX. It is hidden
> > in all sorts of imbedded systems.
> You are correct Sir.
> Not to mention the a variety of high end equipment too.
How about enumerating some? We use QNX in Remotely Operated underwater
Vehicles (ROVs) and robotics.
--
Luca Filipozzi <luca...@ise.bc.ca.spamsucks>
Daryl Low
> How about enumerating some? We use QNX in Remotely Operated underwater
> Vehicles (ROVs) and robotics.
The Advanced Space Vision System (ASVS) is used by NASA to guide the
shuttle's Canadarm and help it dock with space stations. It was done in
QNX 2.x and later ported to QNX 4.x.
Ian Street
>
> ...
>
> How many end-users do YOU know that use QNX? I can't think of any.
>
It wouldn't surprise me if _most_ people have "used" QNX. It is hidden
in all sorts of imbedded systems.
Ian.
Aram Iskenderian
In article <36D085B5...@anca.com.au>
Ian Street <ians+p...@anca.com.au> wrote:
>Geoff Oltmans wrote:
>> How many end-users do YOU know that use QNX? I can't think of any.
>>
>
> It wouldn't surprise me if _most_ people have "used" QNX. It is hidden
>in all sorts of imbedded systems.
>
You are correct Sir.
Not to mention the a variety of high end equipment too.
--
Aram Iskenderian.
ar...@ix.netcom.com, ar...@mailexcite.com.
Ian Street
> > > in all sorts of imbedded systems.
> > You are correct Sir.
> > Not to mention the a variety of high end equipment too.
> Vehicles (ROVs) and robotics.
We use QNX in our Tool and Cutter Grinding machines.
Ian.
Guy Macon
>>What does the QNX microkernel offer that the *BeOS* microkernel doesn't?
>
In addition, QNX has extreme high reliability - good enough for
applications such as heart-lung machines or nuclear power plant
controls. The microkernal is very, very small, very well debugged,
and hasn't changed in a long time. The QNX microkernel keeps the
same memory footprint, so changing a video driver won't transform
a nasty bug that overwites a nonused memory location to a nasty
bug that overwites a critical memory location. This class of bug
tends to either work every time or not work every time. I don't
think that BeOS act this way (I know for a fact that Win95/98,
Win NT, and Linux don't act this way).
As for cost, someone like me who uses one to ten copies per year
as part of multimillion dollar production lines naturally pays
top dollar, while someone who wants to sell many, many Amigas
to consumers will of course pay far less per copy. You really
can't comment on the price for Amiga use unless you are in on the
negotiations.
Guy Macon
>How many end-users do YOU know that use QNX? I can't think of any.
On the contrary, I can almost guarantee that you are using QNX now.
If it's not in the embedded code for your CD-ROM drive or printer,
and it's not in your VCR or your car's engine control, it is most
probably in several of the stages that get AC power to your home,
and also most certainly in some of the routers, etc. between your
PC and the rest of the internet.
I have the full developers version - There are no general purpose
applications. Nothing like Word or Excel. That's fine with me,
because my target systems don't have keyboards, mice, or video cards.
Obviously, if Amiga wants to build in QNX, they need to get the
standard apps. I wonder how hard a Linux to QNX port would be?
Guy Macon
>How about enumerating some?
We use QNX in manufacturing lines that produce 100,000 CD-ROMS per day.
We also use it for a nice demo for visiting investors, executives, etc.
"Now watch as I pull that control window over to my (radio LAN equipped)
laptop. Now I can work on it outside of the cleanroom..." For the real
techies, I show them QNX running on a 486 DX2/66 with 12MB of RAM. The
look and feel is like Win NT on a Quad Xeon with 1GB of RAM.
John Birch
>>How many end-users do YOU know that use QNX? I can't think of any.
>On the contrary, I can almost guarantee that you are using QNX now.
>If it's not in the embedded code for your CD-ROM drive or printer,
>and it's not in your VCR or your car's engine control, it is most
>probably in several of the stages that get AC power to your home,
>and also most certainly in some of the routers, etc. between your
>PC and the rest of the internet.
I'd be very surprised if QNX was used in a CD ROM drive or printer
(although possibly a very large business laser / copier fax machine).
It certainly wouldn't be used in a VCR or in an engine controller.
It _would_ however be used in a set top internet box, a dedicated
internet phone, routers, and very many larger hard / soft real time
applications (including the nuclear power industry and defence).
The smaller applications you mention would almost certainly be
implemented without an operating system or with a primitive home
rolled scheduler if necessary.
<snip>
regards John B.
Derek Leach
>On 24 Feb 1999 13:29:23 PST, guym...@deltanet.com (Guy Macon) wrote:
>
>>>How many end-users do YOU know that use QNX? I can't think of any.
>
>>On the contrary, I can almost guarantee that you are using QNX now.
>>If it's not in the embedded code for your CD-ROM drive or printer,
>>and it's not in your VCR or your car's engine control, it is most
>>probably in several of the stages that get AC power to your home,
>>and also most certainly in some of the routers, etc. between your
>>PC and the rest of the internet.
>
>I'd be very surprised if QNX was used in a CD ROM drive or printer
>(although possibly a very large business laser / copier fax machine).
>It certainly wouldn't be used in a VCR or in an engine controller.
Actually I _believe_ it is used in engine controllers.
DL
Stephen J. Tremblett
http://www.qnx.com/realworld/cust_list.html
--
-=> Steve Tremblett
-=> Memorial University Computer Science
-=> s...@cs.mun.ca; www.cs.mun.ca/~sjt
Mark Sandrock
>Ian Street <ians+p...@anca.com.au> wrote:
>>Geoff Oltmans wrote:
>>> How many end-users do YOU know that use QNX? I can't think of any.
>>
>> It wouldn't surprise me if _most_ people have "used" QNX. It is hidden
>>in all sorts of imbedded systems.
>You are correct Sir.
>Not to mention the a variety of high end equipment too.
I believe Fujitsu uses QNX in some of their PBX systems.
Mark
--
(Not speaking for)
Wolfram Research, Inc.
Voice: 217-398-0700/x107
E-mail: sand...@wolfram.com
John Birch
_Car_ engine controllers??
How do you get QNX to fit onto an 80c196 (the almost de facto standard
chip for car engine management)?
I can well believe it might be used for aviation engine management or
combine harvesters but not your average saloon.
regards John B.
Pat Ford
John Birch <jo...@invision.co.uk> wrote in article
<36d56447...@news.demon.co.uk>...
| On 25 Feb 1999 13:33:42 GMT, dle...@qnx.com (Derek Leach) wrote:
|
| >In article <36d51df9...@news.demon.co.uk>, John Birch <nospam>
wrote:
|
| >>I'd be very surprised if QNX was used in a CD ROM drive or printer
| >>(although possibly a very large business laser / copier fax machine).
| >>It certainly wouldn't be used in a VCR or in an engine controller.
|
| >Actually I _believe_ it is used in engine controllers.
|
| _Car_ engine controllers??
mine ( 386sx laptop grafted to bosch-> renault efi install on a 61
landrover)
|
| How do you get QNX to fit onto an 80c196 (the almost de facto standard
| chip for car engine management)?
most gms use a 68hc11 variant ( I think they are the most wide spread)
|
| I can well believe it might be used for aviation engine management or
| combine harvesters but not your average saloon.
yes in the to the aviation engine test stands and increasingly in the
automotive
manufacturing area. There is automotive application that uses a gui comming
up
| regards John B.
|
|
--
***********************************
* Pat Ford R&D developer *
* QNX Software Systems *
* www.qnx.com *
* Any opinions expressed *
* are my own and not those*
* of my employer *
***********************************
Maury Markowitz
> I can well believe it might be used for aviation engine management
Maybe in the FADEC's on the last-gen turbines, but the sad reality is
that the GA market is based almost entirely on 1950's technology, designs
and materials. Your car engine is considerably more advanced in
practically all ways, and this is certainly the case for computerization,
fuel flow, and monitoring.
Maury
Steve Giovenella
In large part, that's why general aviation aircraft engines are so
tough and reliable. You can't just coast into the Circle K parking
lot when your timing belt slips.
The csaa point being, simplicity and power are very good qualities.
And in defense of the QNX point, it takes a tremendous amount of power
and consistency to run any real time engine management system,
especially one where lives may be lost from a hiccup.
-Steve.
-http://www.amigapro.com
-Bandwidth intensive, and built to stay that way.
Ken Schumm
> Let me give you another example. All of my friends have cell phones. Of
> all of them, I know one that had it break. On the other hand, I've been
> piloting FOUR airplanes where the comms went dead. For this you get to pay
> thousands of dollars. Cell phones are far more complex than av radios.
>
> Volume is the answer.
If volume were the answer, then M$ would be shipping the highest
quality software around.
[...]
Maury Markowitz
Giovenella" wrote:
> In large part, that's why general aviation aircraft engines are so
> tough and reliable.
Aero engines are *not* reliable. I've done *zero* maint on my car in
about 1000 hours of running time. In an average 50 hours on a 185 (IO-540)
something is broken, from a cylinder head, to a hose, to a clamp on a hose,
to an engine monitor cable, to a oil leak. Such is expected, normal, maint.
This is the Linux vs. NT issue again, no matter how overdesigned aero
engines are, they are still less reliable overall because there's less
people using them.
Aero engines are also *not* tough. If I take my foot off the gas in my
car and roll down a hill I don't worry about the block cracking. In a
plane I do. I worry about minor amounts of dirt, clogs in the cowl
openings, outside air temp, humidity, parking on a hill, water in the gas
line etc. Simplicity only leads to reliability if the device is
*inherently simple*. If you take all the goodies out of a car engine they
get *less* reliable.
If aero engines were built with technology from even the last _20_ years
they would be tremendously better than they are today. I know of two that
fit into this catagory (both from Canada oddly enough). The first is the
Rotax, who's reliability continues to be superb (even though it's a brand
new engine), is quieter, WAY more fuel effecient, easier to run, and has a
power to weight (IIRC) something on the order of twice that of the old 150
engines (or such). The other is the new O-600's from Orenda, a piston
engine so advanced that people are pulling out turbines and putting them in
instead, the first time in history this has happened.
> You can't just coast into the Circle K parking
> lot when your timing belt slips.
Let me give you another example. All of my friends have cell phones. Of
all of them, I know one that had it break. On the other hand, I've been
piloting FOUR airplanes where the comms went dead. For this you get to pay
thousands of dollars. Cell phones are far more complex than av radios.
Volume is the answer.
> And in defense of the QNX point, it takes a tremendous amount of power
> and consistency to run any real time engine management system,
> especially one where lives may be lost from a hiccup.
Which would be great, if they use them.
Maury
Guy Macon
>
>On 24 Feb 1999 13:29:23 PST, guym...@deltanet.com (Guy Macon) wrote:
>
>>>How many end-users do YOU know that use QNX? I can't think of any.
>
>>If it's not in the embedded code for your CD-ROM drive or printer,
>>and it's not in your VCR or your car's engine control, it is most
>>probably in several of the stages that get AC power to your home,
>>and also most certainly in some of the routers, etc. between your
>>PC and the rest of the internet.
>
>(although possibly a very large business laser / copier fax machine).
>It certainly wouldn't be used in a VCR or in an engine controller.
>
>internet phone, routers, and very many larger hard / soft real time
>applications (including the nuclear power industry and defence).
>
>The smaller applications you mention would almost certainly be
>implemented without an operating system or with a primitive home
>rolled scheduler if necessary.
You might be suprised. There are occasions where one can reduce
hardware costs in ways that make the software's job much harder.
The classic example is whether to use a hardware or software UART
on your serial port. QNX has a nice ratio between great realtime
capabilities and memory/speed resources used. No one tool meets
all needs, but QNX does fit a subset of the low end market, and I
suspect that those who use QNX in low end applications pay a very
low per-unit price.
Sam E. Trenholme
I am confused. More people use NT, yet Linux is more reliable.
- Sam
--
Email address here: http://www.samiam.org/ssi/mailme.shtml
Music I write here: http://www.mp3.com/sam http://www.samiam.org/mp3
Mp3 reviews here: http://www.samiam.org/music
Steve Giovenella
> In <36D5CDCA.MD-0.1...@spamfree.pn.nettuno.it> "Steve
> Giovenella" wrote:
> > In large part, that's why general aviation aircraft engines are so
> > tough and reliable.
Well, since we hardly ever stay on topic anyway...
> Aero engines are *not* reliable. I've done *zero* maint on my car in
> about 1000 hours of running time. In an average 50 hours on a 185 (IO-540)
> something is broken, from a cylinder head, to a hose, to a clamp on a hose,
> to an engine monitor cable, to a oil leak. Such is expected, normal, maint.
Leaks and cracks happen, its got a much tougher job. I've never had
an AEIO-540 quit on me airborne, knock on wood.
> This is the Linux vs. NT issue again, no matter how overdesigned aero
> engines are, they are still less reliable overall because there's less
> people using them.
You can't compare the two things really. Aero engines aren't Linux,
just the opposite, they are old, rigorously proven designs, they're
expensive too.
> Aero engines are also *not* tough. If I take my foot off the gas in my
> car and roll down a hill I don't worry about the block cracking.
How often do you use gyroscopic precession to twist the entire weight
of your Honda on its ear, from the driveshaft? How many times does a
Honda engine go from running flat out to a stone cold stop, in an
instant, when a blade contacts the ground with no significant damage?
How many times does a Honda get a 4, 5, 6, 7g prang onto the runway?
> In a
> plane I do.
I won't ask. Don't ask me either. :^)
> If aero engines were built with technology from even the last _20_ years
> they would be tremendously better than they are today. I know of two that
> fit into this catagory (both from Canada oddly enough). The first is the
> Rotax, who's reliability continues to be superb (even though it's a brand
> new engine), is quieter, WAY more fuel effecient, easier to run, and has a
> power to weight (IIRC) something on the order of twice that of the old 150
> engines (or such). The other is the new O-600's from Orenda, a piston
> engine so advanced that people are pulling out turbines and putting them in
> instead, the first time in history this has happened.
To some extent you are right, this is the age old debate about the
venerable a/c engine series. Honestly, at the root of this aren't
technology issues so much as liability issues, and how the resulting
legal environment impacts corporate descion making.
I was simply making the point that many aircraft engines are indeed
simpler than cars, but there's a lot to be said for simple, tough
designs, especially in this case.
> > You can't just coast into the Circle K parking
> > lot when your timing belt slips.
>
> Let me give you another example. All of my friends have cell phones. Of
> all of them, I know one that had it break. On the other hand, I've been
> piloting FOUR airplanes where the comms went dead. For this you get to pay
> thousands of dollars. Cell phones are far more complex than av radios.
Again, the reason is certification and liability, and the associated
costs incurred.
> Volume is the answer.
I'm assuming you are bringing this back to the OS analogy, but I'm
afraid aviation involves too many other factors to make such a
straight forward comparision.
Geoff Roberts
> In <36d56447...@news.demon.co.uk> John Birch wrote:
> > I can well believe it might be used for aviation engine management
>
> Maybe in the FADEC's on the last-gen turbines, but the sad reality is
> that the GA market is based almost entirely on 1950's technology, designs
> and materials. Your car engine is considerably more advanced in
> practically all ways, and this is certainly the case for computerization,
> fuel flow, and monitoring.
As a pilot, and also having been around complicated technology for a while, I would
be a little concerned if the relatively simple but still unreliable GA internal
combustion engine (compared to jet turbines) were to start being dependent on a
brain! ;-)
And not just for safety reasons!
OTOH, a QNX based engine performance and monitoring system would tend to be a rather
good idea. I would simply worry about the brain actually "controlling" the engine, and
what this technology would cost. As a General Aviator.... who is often broke as a result.
Geoff.
>
> Maury
>
>
>
--
Realtime Technology Systems Pty Ltd.
2 Hadleigh Circuit,
Isabella Plains,
Canberra,
ACT 2905,
AUSTRALIA.
Phone: 61-2-6291 3833
Fax: 61-2-6291 3838
email: ge...@rtts.com.au
CR91LJL
> Aero engines are *not* reliable. I've done *zero* maint on my car in
>about 1000 hours of running time. In an average 50 hours on a 185 (IO-540)
>something is broken, from a cylinder head, to a hose, to a clamp on a hose,
>to an engine monitor cable, to a oil leak. Such is expected, normal, maint.
>engines are, they are still less reliable overall because there's less
>people using them.
>
>plane I do. I worry about minor amounts of dirt, clogs in the cowl
>openings, outside air temp, humidity, parking on a hill, water in the gas
>line etc. Simplicity only leads to reliability if the device is
>*inherently simple*. If you take all the goodies out of a car engine they
>get *less* reliable.
>
>they would be tremendously better than they are today. I know of two that
>fit into this catagory (both from Canada oddly enough). The first is the
>Rotax, who's reliability continues to be superb (even though it's a brand
>new engine), is quieter, WAY more fuel effecient, easier to run, and has a
>power to weight (IIRC) something on the order of twice that of the old 150
>engines (or such). The other is the new O-600's from Orenda, a piston
>engine so advanced that people are pulling out turbines and putting them in
>instead, the first time in history this has happened.
>
>> lot when your timing belt slips.
>
> Let me give you another example. All of my friends have cell phones. Of
>all of them, I know one that had it break. On the other hand, I've been
>piloting FOUR airplanes where the comms went dead. For this you get to pay
>thousands of dollars. Cell phones are far more complex than av radios.
>
Reliability of a product is related to:
1) Design integrity
2) Quality control
3) Operating history - load factors, duty cycle, etc.
Volume of production is only indirectly related in that a product of high
volume is more likely to produce a large cash flow, thus enabling a company to
afford better design and quality control. The reason older designs like
aircraft engines or military products may be more reliable is that their
designs have been interated on extensively (de-bugged) whereas many consumer
products are continually being replaced by new designs. This doesn't mean that
there aren't companies like Honda who are pretty good at getting it right the
first time. Operating history is a major factor for mechanical products like
engines where load factors and maintenance patterns have a strong impact on
life. We refer to the impact of these factors as "durability" instead of
"reliability". Electronic (solid-state) and software products don't have the
same durability issues.
QSSL doesn't have any better record than their competitors at getting it right
the first time, but they rapidly iterate on the fundamental design and have
produced a product of high integrity and reliability (if you don't use the
"betas").
Len Logterman
Bjørnar Bolsøy
Maury Markowitz <maury@remove_this.istar.ca> wrote:
> In <36D5CDCA.MD-0.1...@spamfree.pn.nettuno.it> "Steve
> Giovenella" wrote:
>
> > In large part, that's why general aviation aircraft engines are so
> > tough and reliable.
>
> Aero engines are *not* reliable.
They also have a couple of more moving parts.
> Aero engines are also *not* tough. If I take my foot off the gas in my
> car and roll down a hill I don't worry about the block cracking. In a
> plane I do. I worry about minor amounts of dirt, clogs in the cowl
> openings, outside air temp, humidity, parking on a hill, water in the gas
> line etc.
Let's not forget "unleded".
> Simplicity only leads to reliability if the device is
> *inherently simple*. If you take all the goodies out of a car engine
they
> get *less* reliable.
Take all the goodies out of a turbofan and what do you got, a
ramjet. Simple and reliable though pretty darn scary to push up
to speed in case you need to pull over for a leak in the woods.
> If aero engines were built with technology from even the last _20_ years
> they would be tremendously better than they are today. I know of two
that
> fit into this catagory (both from Canada oddly enough).
Well doens't surprise me, you made the Otters as well.
> The first is the
> Rotax, who's reliability continues to be superb (even though it's a brand
> new engine), is quieter, WAY more fuel effecient, easier to run, and has
a
> power to weight (IIRC) something on the order of twice that of the old
150
> engines (or such).
What about purchase cost?
> The other is the new O-600's from Orenda, a piston
> engine so advanced that people are pulling out turbines and
> putting them in instead, the first time in history this has happened.
Reminds me of a story I read years ago about how the
Germas pulled turbines out of a few HE162s, allegedly
because they caught fire within a few minutes of warming up.
> On the other hand, I've been
> piloting FOUR airplanes where the comms went dead.
Ouch.
Regards...
Kenneth Xu
Could you please tell me how you measure the look and feel? give some
example!
Kenneth Xu
>
> It wouldn't surprise me if _most_ people have "used" QNX. It is hidden
>in all sorts of imbedded systems.
I don't think we should mix "used an OS" and "used a black box built on an
OS"! It is not an interesting game at all!
Luca Filipozzi
says...
Macon.
Thanks.
--
Luca Filipozzi <luca...@ise.bc.ca.spamsucks>
John Birch
I don't see how having QNX on board would make implementing a software
UART any easier :-).
I agree that in general the reduction in hardware costs means that it
is cheap enough to provide the RAM and ROM space to hold an OS like
QNX on smaller and smaller systems, mind you this same argument also
means that the use of software to replace hardware is far less likely
- even PICs have UARTS :-) I doubt even QSSL's legendary massive
discounts for huge purchases of QNX would make it cheap enough for a
toaster, microwave or even the average saloon car engine management
system. It has to be appreciated that the total cost of the
controllers in these cases is _extremely_ low, so low in fact that it
is little more than the cost of the microcontrollers used. I would
guesstimate at costs in the region of 50 cents, <1$ and <20 dollars
respectively.
I'm not knocking QNX, I enjoy using it, but it's not as ubiquitous as
some might have us believe!. Please let's not indulge in the same kind
of hype as M$ are over CE. If I get asked to design a product that
needs an OS I'll choose the appropriate one, probably QNX given my
line of work. It it doesn't _need_ an OS, then I won't use one.
regards John B.
Armin Steinhoff
>
>On 25 Feb 1999 17:47:33 PST, guym...@deltanet.com (Guy Macon) wrote:
>
>>In article <36d51df9...@news.demon.co.uk>, jo...@invision.co.uk wrote:
>>>The smaller applications you mention would almost certainly be
>>>implemented without an operating system or with a primitive home
>>>rolled scheduler if necessary.
>
>>You might be suprised. There are occasions where one can reduce
>>hardware costs in ways that make the software's job much harder.
>>The classic example is whether to use a hardware or software UART
>>on your serial port. QNX has a nice ratio between great realtime
>>capabilities and memory/speed resources used. No one tool meets
>>all needs, but QNX does fit a subset of the low end market, and I
>>suspect that those who use QNX in low end applications pay a very
>>low per-unit price.
>
>I don't see how having QNX on board would make implementing a software
>UART any easier :-).
It makes the usage of e.g. dumb CAN boards easier ... that means you can replace
an onboard MCU by the power of the PC CPU :-)
>I agree that in general the reduction in hardware costs means that it
>is cheap enough to provide the RAM and ROM space to hold an OS like
>QNX on smaller and smaller systems, mind you this same argument also
>means that the use of software to replace hardware is far less likely
>- even PICs have UARTS :-)
There is a PIC compatible MCU with 50-100 MIPS processing power(www.scenix.com).
This 8 bit MCU allows software UARTS ... so called virtual devices :-)
Regards
Armin
Maury Markowitz
> > Aero engines are *not* reliable.
>
> They also have a couple of more moving parts.
Incorrect, they have way _less_ moving parts. There is typically no
cooling system, aside perhaps from cowl flaps with are external to the
engine itself, no RPM timing adjustment, trivial carbs. This is even more
true of turbines.
> Let's not forget "unleded".
Indeed, this is another problem, avgas is HEAVILY leaded. This is
because of the old materials, just as it is with old cars. Interestingly
most GA planes were designed to run on 87 octane, which is the same as
mogas here in NA. Yet, of course, the planes can't burn that without an
expensive overhaul.
> Well doens't surprise me, you made the Otters as well.
I got a chance to put some time in the CAT3 Twotter sim up at DeHaviland a
few years back. The instructor said "this is a big 150". And sure enough
that's exactly how you fly it, it drives around the same, rotates the same,
lands the same.
> What about purchase cost?
Don't have it handy, but low. As I understand it it's basically an
engine from the series that they use in snowmobiles and motorcycles scaled
up. Thus the advanced materials.
> Reminds me of a story I read years ago about how the
> Germas pulled turbines out of a few HE162s, allegedly
> because they caught fire within a few minutes of warming up.
What would they replace them with? That back-mounted engine makes other
thing hard to put up there.
> Ouch.
The funnest was over Atlantic City, because the same thing happened to
four other planes at the same time. The ATC guy was going ballistic!
Maury
Maury Markowitz
> I am confused. More people use NT
...but can't do anything to help fix it.
Maury
Maury Markowitz
Giovenella" wrote:
> Leaks and cracks happen
They don't in car engines. They don't have to in aero engines either.
> its got a much tougher job.
No it doesn't, a car engine has a a FAR harder job, because of the power
cycling.
> I've never had an AEIO-540 quit on me airborne, knock on wood.
I've never had a car engine quit on the road.
> You can't compare the two things really.
Fair enough.
> How often do you use gyroscopic precession to twist the entire weight
> of your Honda on its ear, from the driveshaft?
I don't do this on planes either.
> How many times does a
> Honda engine go from running flat out to a stone cold stop, in an
> instant
How many times does this happen to an aero engine? Once, then you throw
it away. I've stalled my VTEC at the lights a few times though...
> when a blade contacts the ground with no significant damage?
Have you ever heard of a prop strike that didn't cost at least $5000 to
fix? I haven't. It typically requires the engine being pulled and torn
down looking for propshaft damage, and I can't say I've ever heard of one
that didn't get at _least_ pulled.
> How many times does a Honda get a 4, 5, 6, 7g prang onto the runway?
Cars often get into 20g prangs on the road, and the owner walks away from
it. The car is a rightoff, sure, that rightoff costs about as much as
getting the 150 checked after 7g's.
> To some extent you are right, this is the age old debate about the
> venerable a/c engine series. Honestly, at the root of this aren't
> technology issues so much as liability issues, and how the resulting
> legal environment impacts corporate descion making.
Sure. But we all hate lawers already, so there's no point harping about
it I suppose. Or maybe there is, did you ever hear why the 150's needed a
new fuel checking port drilled into the cabin end of the wing?
> Again, the reason is certification and liability, and the associated
> costs incurred.
I know, but the issue as I see it is that the technology has advanced so
quickly that it's possible to make far more reliable, cost effective, low
price and better performing technologies. This goes almost across the
board in aviation, where Cessna can start selling at 40 year old design and
it's "news".
The issue, as I see it, is that aviation is more expensive and less safe
than it could be.
> I'm assuming you are bringing this back to the OS analogy, but I'm
> afraid aviation involves too many other factors to make such a
> straight forward comparision.
Fair enough, I was being a little silly.
Maury
Guy Macon
>
>>In article <MPG.113a5b7b2...@news.bctel.net>,
>luca...@ise.bc.ca.spamsucks wrote:
>>For the real
>>techies, I show them QNX running on a 486 DX2/66 with 12MB of RAM. The
>>look and feel is like Win NT on a Quad Xeon with 1GB of RAM.
>>
>
>Could you please tell me how you measure the look and feel? give some
>example!
Take one engineer, familiar with Windows NT. Sit him down at an old 486
with a full installation of QNX and Photon. Leave the architecture guide
nearby - he WILL be interested in a few moments. Say "Go ahead and try
anything. Doesn't matter what you do because I can reload with a couple
of keystrokes". Walk away. Go tu cubicle. Wait for amazed person to
come by, ranting and raving over fast performance, plus a few minor
"confused by the GUI". problems. Followup with comments about ANSI and
Posix standards vs. proprietary shifting standards from the M$ Borg.
Guy Macon
>
>In article <36d63...@news.cyberway.com.sg>, xu_...@NOSPAM.irel.com.sg
>> >In article <MPG.113a5b7b2...@news.bctel.net>,
>> luca...@ise.bc.ca.spamsucks wrote:
>> >For the real
>> >techies, I show them QNX running on a 486 DX2/66 with 12MB of RAM. The
>> >look and feel is like Win NT on a Quad Xeon with 1GB of RAM.
>> >
>>
>> Could you please tell me how you measure the look and feel? give some
>> example!
>Macon.
>
>Thanks.
>--
>Luca Filipozzi <luca...@ise.bc.ca.spamsucks>
Yeah! Luca is MUCH more intellegent than me. How could you be so
insulting? <big grin> ;)
Guy Macon
>I'm not knocking QNX, I enjoy using it, but it's not as ubiquitous as
>some might have us believe!. Please let's not indulge in the same kind
>of hype as M$ are over CE. If I get asked to design a product that
>needs an OS I'll choose the appropriate one, probably QNX given my
>line of work. It it doesn't _need_ an OS, then I won't use one.
Wise words indeed.
dl...@lehigh.edu
<snip>
>Take one engineer, familiar with Windows NT. Sit him down at an old 486
>with a full installation of QNX and Photon. Leave the architecture guide
>nearby - he WILL be interested in a few moments. Say "Go ahead and try
>anything. Doesn't matter what you do because I can reload with a couple
>of keystrokes". Walk away. Go tu cubicle. Wait for amazed person to
>come by, ranting and raving over fast performance, plus a few minor
>"confused by the GUI". problems. Followup with comments about ANSI and
>Posix standards vs. proprietary shifting standards from the M$ Borg.
I am familiar with Windows NT. I am familiar with Windows 98/95/3.x, too.
I am familiar with both DOS and Linux. I am marginally familiar with BSD,
Solaris, IRIX, and so, I suppose, then, most all *nix variants. I am not
very
much familiar with QNX. I have used the "IAT Demo Disk - Version 3.03",
and, while reasonably nice, it didn't overly impress me. Other than that
one demo, I have never, to my knowledge, either seen or used QNX.
To a complete QNX newbie, please explain how well QNX can serve as a
home operating system. How much does it cost, for what functionality, that
is how usable?
Why might I like QNX more, or less, than, e.g., BSD or Linux?
Derrick Shearer
dl...@Lehigh.EDU
Joe Cosby
John Birch hunched over his computer, typing feverishly;
thunder crashed, John Birch laughed madly, then wrote:
> On 25 Feb 1999 17:47:33 PST, guym...@deltanet.com (Guy Macon) wrote:
>
>
> >QNX does fit a subset of the low end market, and I
> >suspect that those who use QNX in low end applications pay a very
> >low per-unit price.
>
> I agree that in general the reduction in hardware costs means that it
> is cheap enough to provide the RAM and ROM space to hold an OS like
> QNX on smaller and smaller systems, mind you this same argument also
> means that the use of software to replace hardware is far less likely
> - even PICs have UARTS :-) I doubt even QSSL's legendary massive
> discounts for huge purchases of QNX would make it cheap enough for a
> toaster, microwave or even the average saloon car engine management
> system.
But if QNX are trying to break into a completely different -kind- of
market, they are going to need a whole different philosophy in their
pricing. I hope this will be obvious to them.
Being competitive in a -huge- consumer market will require paper-thin
profit margins. Just as in the desktop industry.
They will be competing with CE, I assume.
> It has to be appreciated that the total cost of the
> controllers in these cases is _extremely_ low, so low in fact that it
> is little more than the cost of the microcontrollers used. I would
> guesstimate at costs in the region of 50 cents, <1$ and <20 dollars
> respectively.
>
> I'm not knocking QNX, I enjoy using it, but it's not as ubiquitous as
> some might have us believe!. Please let's not indulge in the same kind
> of hype as M$ are over CE. If I get asked to design a product that
> needs an OS I'll choose the appropriate one, probably QNX given my
> line of work. It it doesn't _need_ an OS, then I won't use one.
>
But in this case, we're talking about -one- OS, which will be active
in digicon devices, STBs, and (if everything goes as planned) desktops.
And inter-communicating, between them.
I would think there would be large advantages in that, in design,
flexibility of application, and compatibility between devices.
--
----------------------------------------------------
Joe Cosby
Devout member of the Church of Amiga since 1990
"Whatever you can do, or dream you can, begin it.
Boldness has genius, power and magic in it" - Goethe
----------------------------------------------------
Joe Cosby
If everything goes as they plan, it will -become-
a very interesting game :^)
Joe Cosby
"Pat Ford" hunched over his computer, typing feverishly;
thunder crashed, "Pat Ford" laughed madly, then wrote:
>
> yes in the to the aviation engine test stands and increasingly in the
> automotive
> manufacturing area. There is automotive application that uses a gui comming
> up
>
Dashboard Quake, anyone?
...and you thought car phones were bad...
Clyde Coffey
Markowitz) wrote:
>In <36D5FDA2.MD-0.1...@spamfree.pn.nettuno.it> "Steve
>Giovenella" wrote:
>> its got a much tougher job.
>
> No it doesn't, a car engine has a a FAR harder job, because of the power
>cycling.
I disagree. Car engines are pikers, so to speak. Cars may use 50%
power to leave a stop sign, or 75% to get on the freeway. Both
operations last a few seconds. The rest of the time they are just
loafing. It takes the average little 100hp car about 15hp to cruise at
60mph. Aero engines are run at 100% or max availble power for minutes
on end, and 60-70% power for hours at a time.
Joe Cosby
<dl...@Lehigh.EDU hunched over his computer, typing feverishly;
thunder crashed, <dl...@Lehigh.EDU laughed madly, then wrote:
> To a complete QNX newbie, please explain how well QNX can serve as a
> home operating system. How much does it cost, for what functionality, that
> is how usable?
>
> Why might I like QNX more, or less, than, e.g., BSD or Linux?
>
It will talk to your toothbrush.
--
----------------------------------------------------
Joe Cosby
Waste of bandwidth extraordinaire.
----------------------------------------------------
Armin Steinhoff
says...
>
>** To reply in e-mail, remove "tascuj." from address **
>
>John Birch hunched over his computer, typing feverishly;
>thunder crashed, John Birch laughed madly, then wrote:
>> On 25 Feb 1999 17:47:33 PST, guym...@deltanet.com (Guy Macon) wrote:
>>
>>
>> >QNX does fit a subset of the low end market, and I
>> >suspect that those who use QNX in low end applications pay a very
>> >low per-unit price.
>>
>> I agree that in general the reduction in hardware costs means that it
>> is cheap enough to provide the RAM and ROM space to hold an OS like
>> QNX on smaller and smaller systems, mind you this same argument also
>> means that the use of software to replace hardware is far less likely
>> - even PICs have UARTS :-) I doubt even QSSL's legendary massive
>> discounts for huge purchases of QNX would make it cheap enough for a
>> toaster, microwave or even the average saloon car engine management
>> system.
>
>But if QNX are trying to break into a completely different -kind- of
>market, they are going to need a whole different philosophy in their
>pricing. I hope this will be obvious to them.
>
>Being competitive in a -huge- consumer market will require paper-thin
>profit margins.
> Just as in the desktop industry.
How can Bill Gates live with these 'paper-thin' profit margins ??
Some US consumers organisations have a very different view !
>They will be competing with CE, I assume.
Nonsens ... QNX4 and QNX/Neutrino covers the whole range from embedded
RT apps to high-performance RT workstations ! Do you really know what CE is ?
CE has nothing to do with true real-time !
Armin
Steve Giovenella
> In <36D5FDA2.MD-0.1...@spamfree.pn.nettuno.it> "Steve
> Giovenella" wrote:
> > Leaks and cracks happen
>
> They don't in car engines. They don't have to in aero engines either.
Well, of course they do, but not as much because car engines
are designed for relatively benign conditions. A/c engines are often
built to leak on the ground, because the seals and tolerances etc
have to be much looser to accomodate massive temp fluctuations. As
far as cracks and whatnot, car engines wouldn't last very long if you
subjected them to the regular punishment flying inflicts.
> > its got a much tougher job.
>
> No it doesn't, a car engine has a a FAR harder job, because of the power
> cycling.
I still disagree. A/c engines regularly sustain forces that are just
insane compared to autos. A compressor fan blade, experiences no
kidding 100,000 g's at the tips during normal operation--better blend
those nicks properly... Even general aviation engines have to resist
forces trying to rip them apart while airborne that car engines
never experience. Prop balance takes a big toll. Every time you
change your aircraft attitude, or even get bumped around a little in
turbies, gyroscopic forces on the major moving parts are immense.
Everytime you fly uncoordinated, or rotate or flare or pitch or yaw,
assymetric AOA on the prop blades also inflicts heavy shear forces.
Not to mention the obvious stresses and lubrication problems
associated with greater than, or less than 1 g operation.
> > I've never had an AEIO-540 quit on me airborne, knock on wood.
>
> I've never had a car engine quit on the road.
I've had several. Pulling into my driveway, my ultra-reliable (up to
then) Toyota's timing belt slipped and stopped--good timing, but still
unstartable. I'll never forget driving up the mountains on HWY 24
from C.Springs when a friend's Honda Civic ignitor failed on a steep
grade--also not restartable. When the tow truck never showed, we
rolled back down into C.Springs--pushed it only 100 yards or so into a
service station at the bottom. At first we were tenative, then we
started passing slow pokes. Later, we measured that we coasted 13
miles! :^)
> > You can't compare the two things really.
>
> Fair enough.
>
> > How often do you use gyroscopic precession to twist the entire
weight
> > of your Honda on its ear, from the driveshaft?
>
> I don't do this on planes either.
Maybe not as violently, but every day IO-540s do that repeatedly
without a hitch. Yours experiences many of the same dynamics that
make such manuevers possible.
> > How many times does a
> > Honda engine go from running flat out to a stone cold stop, in an
> > instant
>
> How many times does this happen to an aero engine? Once, then you
throw
> it away. I've stalled my VTEC at the lights a few times though...
>
> > when a blade contacts the ground with no significant damage?
>
> Have you ever heard of a prop strike that didn't cost at least
$5000 to
> fix? I haven't. It typically requires the engine being pulled and
torn
Not to put a damper on the conversation or anything, but I've pulled
a still servicable IO-540 (pilot unfortunately wasn't) out from 3-4
feet of earth. They are pretty stout machines.
> > How many times does a Honda get a 4, 5, 6, 7g prang onto the
runway?
>
> Cars often get into 20g prangs on the road, and the owner walks
away from
> it. The car is a rightoff, sure, that rightoff costs about as much
as
> getting the 150 checked after 7g's.
We both agree a/c components are expensive, I'm just saying there are
lots of reasons for that, even a few good ones.
> > To some extent you are right, this is the age old debate about the
> > venerable a/c engine series. Honestly, at the root of this aren't
> > technology issues so much as liability issues, and how the
resulting
> > legal environment impacts corporate descion making.
>
> Sure. But we all hate lawers already, so there's no point harping
about
> it I suppose. Or maybe there is, did you ever hear why the 150's
needed a
> new fuel checking port drilled into the cabin end of the wing?
No, go ahead...
> > Again, the reason is certification and liability, and the
associated
> > costs incurred.
>
> I know, but the issue as I see it is that the technology has
advanced so
> quickly that it's possible to make far more reliable, cost
effective, low
> price and better performing technologies. This goes almost across
the
> board in aviation, where Cessna can start selling at 40 year old
design and
> it's "news".
>
> The issue, as I see it, is that aviation is more expensive and
less safe
> than it could be.
I'd like to say, well, its a free market, so if the was a better deal
than the 172 going, it would do just as well. But we both know that
there are, only the FAA and our sometimes nutty legal system generally
tries to allow as little of that as possible.
> > I'm assuming you are bringing this back to the OS analogy, but I'm
> > afraid aviation involves too many other factors to make such a
> > straight forward comparision.
>
> Fair enough, I was being a little silly.
Good break form the normal computer geek stuff though :^)
-Steve.
-http://www.amigapro.com
-The website for the Amiga performance enthusiast.
Maury Markowitz
<EF3B10247F8E936D.9CE5BE44...@library-proxy.airnews.ne
t> Clyde Coffey wrote:
> I disagree. Car engines are pikers, so to speak. Cars may use 50%
> power to leave a stop sign, or 75% to get on the freeway. Both
> operations last a few seconds. The rest of the time they are just
> loafing.
But short bursts of power are far harder on the engine than continual
output. If you put your car engine at 65% power, as you do in a plane, and
leave it there, it would do just fine. At the same time if you go
65%/25%65% in an airplane engine, you're going to reduce it's lifetime
tremendously.
Maury
Maury Markowitz
Giovenella" wrote:
> Well, of course they do, but not as much because car engines
> are designed for relatively benign conditions. A/c engines are often
> built to leak on the ground
I'm personally familiar with about 10 engine types, and I've never heard
of this. Is this something that you can point to, ie, the 540 is designed
specifically to do this?
> I still disagree. A/c engines regularly sustain forces that are just
> insane compared to autos. A compressor fan blade
I am talking pistons though, so this is not entirely fair. Tubines cost
into the millions, so that's not a great comparison either (although
pistons are likely to as well).
> Maybe not as violently, but every day IO-540s do that repeatedly
> without a hitch. Yours experiences many of the same dynamics that
> make such manuevers possible.
Which manuevers are you referring to?
> Not to put a damper on the conversation or anything, but I've pulled
> a still servicable IO-540 (pilot unfortunately wasn't) out from 3-4
> feet of earth. They are pretty stout machines.
This is the first I've heard of anything like this. Are you saying that
you pulled the engine out, stuck it on a plane, and flew it? If not, how
much did the checkout cost you?
> > new fuel checking port drilled into the cabin end of the wing?
>
> No, go ahead...
Perhaps an urban rumor, but here goes: 150 is parked outside on a hill in
the winter. The owner goes out and flies off without even doing a
walkaround. Plane crashes shortly after takeoff. Widow sues and wins,
saying that you couldn't see the water in the tanks because of the
placement of the check valves. Refit on 150s then follows.
> Good break form the normal computer geek stuff though :^)
Indeed, these types of conversations don't come up in the aviatio forums,
where we all talk about computers!
Maury
joec...@seatac.net
Volume.
> Some US consumers organisations have a very different view !
>
> >They will be competing with CE, I assume.
>
> Nonsens ... QNX4 and QNX/Neutrino covers the whole range from embedded
> RT apps to high-performance RT workstations ! Do you really know what CE is ?
>
I'm referring to the use of QNX in the Amiga OS5.
> CE has nothing to do with true real-time !
>
--
Joe Cosby
Devout member of the Church of Amiga since 1990
"Whatever you can do, or dream you can, begin it.
Boldness has genius, power and magic in it" - Goethe
-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own
joec...@seatac.net
Gene Poole
> ... I have used the "IAT Demo Disk - Version 3.03", and, while reasonably
> nice, it didn't overly impress me.
You're either very hard to impress, or you didn't understand what you were
seeing. To do the equivalent with Internet Explorer and NT 4.0, how many
diskettes would you need: 20, 40, more?
> Other than that one demo, I have never, to my knowledge, either seen or used
> QNX.
QNX generally works on mission critical real-time applications behind the
scenes: nuclear reactors, ATM's, shuttle arms. You know, dull stuff like that,
that has no effect on your life.
--- Msged 4.00
# Origin: Variation on a Theme *** Gloucester ON CA (510:563/1.1)
Igor Kovalenko
>
> On Feb 26 13:39, 1999, <dl...@Lehigh.EDU> of 510:563/1 wrote:
>
> > ... I have used the "IAT Demo Disk - Version 3.03", and, while reasonably
> > nice, it didn't overly impress me.
>
> You're either very hard to impress, or you didn't understand what you were
> seeing. To do the equivalent with Internet Explorer and NT 4.0, how many
> diskettes would you need: 20, 40, more?
>
He is probably referring to slow video performance. But being familar
with all kinds of Windows and *nixes he should know that no one of those
systems can provide support with high resolution for all vesa2
compatible chipsets in single driver (Pg.flat). I remember somebody
suggested to develop such a driver to XFree86 team and they answered
that it is technically impossible.
But you can't buy something for nothing, so Pg.flat doesn't do any
acceleration.
> > Other than that one demo, I have never, to my knowledge, either seen or used
Try to install real QNX/Photon with an accelerated driver and feel the
difference.
- Igor
Igor Kovalenko
>
> Indeed, these types of conversations don't come up in the aviatio forums,
> where we all talk about computers!
>
So, let's vote for comp.os.qnx.aero-engines where you should continue
this discussion :)
- Igor
dl...@lehigh.edu
>On Feb 26 13:39, 1999, <dl...@Lehigh.EDU> of 510:563/1 wrote:
>
> > ... I have used the "IAT Demo Disk - Version 3.03", and, while
reasonably
> > nice, it didn't overly impress me.
>
>You're either very hard to impress, or you didn't understand what you were
>seeing. To do the equivalent with Internet Explorer and NT 4.0, how many
>diskettes would you need: 20, 40, more?
Consider me rather hard to impress. Other than the size of the demo, what
is there to it? Its speed is hardly impressive, considering the fact that
it
was uncompressed into a RAM drive. Its interface is crude. There is not
much in the way of multitasking or realtime response with which to test it.
While it can be useful to some in certain circumstances, it isn't useful to
me,
under any circumstances of which I am currently aware.
Microsoft Windows does not make any pretenses as to its smallness of
size for the very simple reason that it expected to be run on systems having
multiple gigabyte hard disk drives. Size is not an issue, there.
> > Other than that one demo, I have never, to my knowledge, either seen or
> > used QNX.
>
>QNX generally works on mission critical real-time applications behind the
>scenes: nuclear reactors, ATM's, shuttle arms. You know, dull stuff like
that,
>that has no effect on your life.
I am aware of where QNX primarily sees use. However, you are wrong.
Read the topic. "Cheap Amigas and costly QNX"... QNX seems to be
making a move towards the desktop scene, and, that being the case, I
am curious about how it is for a desktop system. I am aware that QNX is
expensive, but I do not know how much for exactly what I want. How can I
know when I am not exactly sure what I want, because I am not even sure
what is available, and the quality of that which is available.
Will someone in this newsgroup please provide a comparrison between
QNX and Linux as PC type systems. Cost is not a consideration when
comparing functionality and performance. How complete a system does
QNX provide? What does QNX require to allow for a complete system?
Derrick Shearer
dl...@Lehigh.EDU
dl...@lehigh.edu
>Gene Poole wrote:
>> On Feb 26 13:39, 1999, <dl...@Lehigh.EDU> of 510:563/1 wrote:
>>
>> > ... I have used the "IAT Demo Disk - Version 3.03", and, while
reasonably
>> > nice, it didn't overly impress me.
>>
>> You're either very hard to impress, or you didn't understand what you
were
>> seeing. To do the equivalent with Internet Explorer and NT 4.0, how many
>> diskettes would you need: 20, 40, more?
>
>with all kinds of Windows and *nixes he should know that no one of those
>systems can provide support with high resolution for all vesa2
>compatible chipsets in single driver (Pg.flat). I remember somebody
>suggested to develop such a driver to XFree86 team and they answered
>that it is technically impossible.
The demo ran OK, and I was not disappointed with the video performance,
due to the fact that it was acceptably fast, even though using only generic
drivers. I would not, however, expect DOS to allow for less performance,
given similar supporting functionality, such as TCP/IP, a windowing system,
and similar programs, running on sufficiently fast hardware. I would expect
an equivalent DOS based setup to possibly be a little larger, but, other
than
that, what differences to be seen?
>But you can't buy something for nothing, so Pg.flat doesn't do any
>acceleration.
It's not just acceleration that I would like to know about. There is that
supposed ability to migrate processes, as I understand it, from one
QNX machine, to another, by dragging across a shared desktop.
While the description sounds interesting, how am I to know that the
implementation lives up to the claims? Furthermore, what other
nifty features exist for QNX which are not readily available for other
platforms? I would like for someone to compare QNX as a desktop
operating system to Linux as a desktop operating system. I can't. I
have never, to my knowledge, either seen or used QNX.
>> > Other than that one demo, I have never, to my knowledge, either seen
>> > or used QNX.
>
>Try to install real QNX/Photon with an accelerated driver and feel the
>difference.
I am sure that the video performance would be much better. But what
would the numbers be? Running Quake, or Doom, or some other
multiple platform game, e.g., how many fps difference might be expected?
That would certainly depend upon the drivers, but if the drivers were well
matched, then other aspects of system performance could be compared
as a rough increase in system throughput, identified by fps.
Consider one system. At one time loaded with DOS, using generic video
drivers, running Doom. At another time loaded with QNX, using generic
video drivers, running Doom. Since the video drivers and the game are,
for all practical purposes, the same, then the only difference measurable
is the operating system's ability to cope with demand.
If QNX can outperform DOS in terms of throughput, how much would it be
by? If QNX cannot outperform DOS in terms of throughput, then why might
I be interested in it, when other systems offer equivalent functionality,
albeit
for larger size?
A better grounds for comparison might be QNX running Doom with
unaccelerated video drivers contrasted to Linux running Doom with
unaccelerated video drivers, both run on a single machine so as not
to let machine differences interfere with testing results.
If I had the money to try QNX on a whim, then I suppose that I might buy it,
however, I don't, so I won't. Therefore, I must rely upon the opinions and
reviews of others. Getting answers is like pulling teeth; no fun. :(
>- Igor
Derrick Shearer
dl...@Lehigh.EDU
Steve Giovenella
Ok, I managed to find a brilliant way to bring this back to the OS
discussion at the end, but if you want to continue, how about we go
email?
> In <36D7E4A2.MD-0.1...@spamfree.pn.nettuno.it> "Steve
> Giovenella" wrote:
> > Well, of course they do, but not as much because car engines
> > are designed for relatively benign conditions. A/c engines are often
> > built to leak on the ground
>
> I'm personally familiar with about 10 engine types, and I've never heard
> of this. Is this something that you can point to, ie, the 540 is designed
> specifically to do this?
I was actually referring to higher performance stuff. Pretty much
any military tech order will establish maximum "normal" drops/minute
rates from various places of aircraft and engines running on the
ground. Some go higher than 1/sec.
> > I still disagree. A/c engines regularly sustain forces that are just
> > insane compared to autos. A compressor fan blade
>
> I am talking pistons though, so this is not entirely fair. Tubines cost
> into the millions, so that's not a great comparison either (although
> pistons are likely to as well).
You're right, I was just illustrating the high end.
> > Maybe not as violently, but every day IO-540s do that repeatedly
> > without a hitch. Yours experiences many of the same dynamics that
> > make such manuevers possible.
>
> Which manuevers are you referring to?
Many of the modern airshow tumbling maneuvers require precession to
do right. The Lomcevak being the most popular example. Its tough to
turn the initial outside snap roll into an end over end tumble without
a meaty prop to cause precession.
> > Not to put a damper on the conversation or anything, but I've pulled
> > a still servicable IO-540 (pilot unfortunately wasn't) out from 3-4
> > feet of earth. They are pretty stout machines.
>
> This is the first I've heard of anything like this. Are you saying that
> you pulled the engine out, stuck it on a plane, and flew it? If not, how
> much did the checkout cost you?
Actually no. Its a long story but, for accounting purposes the engine
was accessed, to our astonishment, to be servicable. It was never
returned to service (AFAIK, this was in 95) because it was part of a
greater three crash investigation.
> > > new fuel checking port drilled into the cabin end of the wing?
> >
> > No, go ahead...
>
> Perhaps an urban rumor, but here goes: 150 is parked outside on a hill in
> the winter. The owner goes out and flies off without even doing a
> walkaround. Plane crashes shortly after takeoff. Widow sues and wins,
> saying that you couldn't see the water in the tanks because of the
> placement of the check valves. Refit on 150s then follows.
>
> > Good break form the normal computer geek stuff though :^)
>
> Indeed, these types of conversations don't come up in the aviatio forums,
> where we all talk about computers!
Hey Maury, ask them what aircraft they think best repesents WindowsXX
for me, would you please? I vote for the Super Guppy:
http://station.nasa.gov/station/assembly/superguppy/gallery/sg1_low.jpg
Armin Steinhoff
>
>Igor Kovalenko wrote in message <36D8E352...@comm.mot.com>...
>>dl...@Lehigh.EDU wrote:
><snip><Mars demo><snip>
>>Now about size. The _whole_thing_ was 20 (twenty) Kb. No kidding and no
>>compression whatsoever. I think I still have that demo somewhere if
>>somebody is intrigued. Most people used to think that video driver alone
>>must be bigger :)
>
>If its demos you want: http://www.hornet.org
>How about 4Kb demos?...
Just kiddy ... a 'hello world demo' is much smaller :-) .. if you not
use M$ VC++
>One of the reasons why I'm difficult to impress is that I've
>already seen a fair bit of what ingenuity is capable of.
You are just impressed about how small a specialized piece of code
can be if is designed for a single purpose !
>> It's not just acceleration that I would like to know about. There is that
>>> supposed ability to migrate processes, as I understand it, from one
>>> QNX machine, to another, by dragging across a shared desktop.
>>
>>That is misunderstanding. They don't migrate processes, just windows.
>
>Oh, now that is a disappointment. Process migration would be something
>to write home about.
QNX as a DISTRIBUTED RTOS allows the networkwide transparently distribution
of processes ... that happens when a PHOTON window 'migrates' !
>I've migrated windows plenty of times, and continue to do so using VNC and X.
This functionality is not integrated in the underlaying OS.
Allows VNC to SLIDE a window from screen to screen ? AFAIK it can't do it !
> Photon might be a bit more efficient
It is more efficient because it's using an already distributed OS platform !
> in its implementation, but how much so,
> I'm not sure I could very well guess; too many factors to account for.
>
>>> While the description sounds interesting, how am I to know that the
>>> implementation lives up to the claims?
>>
>>Pretty close. The magic thing is, you don't have to do anything special
>>in application for that. QNX and Photon implement that effect
>>transparently. So, you can migrate windows not because they programmed
>>an app to do so, but just because in QNX/Photon it is as natural as move
>>them inside your desktop.
>
>
>A matter of integration.
That's the key point. This integration must starts with the architecture of the
OS and not on top of it! A microkernel design for a windowing system makes no
sense if the underlaying OS is a monolithic one.
The AMIGA OS5 will be based on QNX/Neutrino. OS5 inherits the distributed
microkernel architecture, which means:
- network transperant IPC
- true micro kernel architecture
- extreamly high scalabilty from STBs to high-performance OS5 workstations
- distribution of applications
- SMP support for powerful real-time applications
- simplified development of drivers
- a.s.o.
>Which, while nice, doesn't sell me entirely on QNX being considerably better,
>and, hence, worth the price.
Take the attributes above and add the robustness and stability of QNX ... and
you will see it's more than 'worth the price' ... it pays back after some years
of usage in the field. We have applications running since years in water power
plants and steel productions without any reboot actions !
> I already know how to 'effortlessly' (?)
>export the displays of programs which weren't specifically designed for such
>on operating systems which weren't specifically designed for such.
>What else does QNX offer?
Not simple desktop apps for the office ... but see above.
>>When I wanted to impress people, I used to connect 2 boxes by ethernet
>>and null-modem, then start copying a large file using any program which
>>provides some sort of 'progress bar'. In the middle of process, I unplug
>>the ethernet. Then see how copying continues with slower speed. Then
>>plug ethernet back and see how it becomes faster again and completes as
>>nothing happened.
>
>
>That strikes me as being a bit more impressive. I am curious as to what
>is involved enabling QNX to work like that. I can imagine setting similar
>up for some *nix to do the same, or similar, but that wouldn't happen "right
>out of the box." Is the requisite process for QNX similarly involved, or
>not? Details?
See the architecture guide of Neutrino 2.0 at
http://www.qnx.com/literature/nto_sysarch/intro.htm
and you will see what the BASE of OS5 will be.
Again, it will be the BASE of OS5 !!! ... and not the desktop OS for the AMIGA!
[ clip ..]
>So run some dummy background applications to highlight the differences
>between single tasking performance, dual tasking performance, triple
>tasking performance, quad...unto 9999. Do the same for Linux, or BSD,
>or some commercial *nix. If QNX can be shown to significantly outstrip
>the competition in throughput, then I believe that I would be significantly
>impressed enough to buy into QNX as a desktop type system.
150.000 task switches/s (Send/Reply) between two user level tasks ... is
it fast enough for you ? (K6/266Mhz)
Armin
Gene Poole
> Consider me rather hard to impress.
OK. I won't try too hard to overcome your perceptions. You are welcome to
keep them or not.
> Other than the size of the demo, what is there to it?
But you see, size is *very* important when you are turning out a million of an
embedded application. When you can put the OS, GUI and application in about 2M
of ROM, you can *eliminate* the need for a hard disk. Can you say "profit".
> Its speed is hardly impressive, considering the fact that it was
> uncompressed into a RAM drive.
As was pointed out in another message, the video driver is a "universal video
card" driver that maps to a "flat" screen and has to do all the 3D effects in
software on the main processor instead of in the graphics card.
> Its interface is crude.
OK. That's your opinion.
> There is not much in the way of multitasking or realtime response
> with which to test it.
Did you try starting about a dozen "QNX is Cool" windows?
> While it can be useful to some in certain circumstances, it isn't useful to
> me, under any circumstances of which I am currently aware.
It's *only* a demo; it's not meant to be useful.
> Microsoft Windows does not make any pretenses as to its smallness of
> size for the very simple reason that it expected to be run on systems having
> multiple gigabyte hard disk drives. Size is not an issue, there.
No comment. <big grin>
> I am aware of where QNX primarily sees use. However, you are wrong.
> Read the topic. "Cheap Amigas and costly QNX"... QNX seems to be
> making a move towards the desktop scene, and, that being the case, I
> am curious about how it is for a desktop system. I am aware that QNX is
> expensive, but I do not know how much for exactly what I want. How can I
> know when I am not exactly sure what I want, because I am not even sure
> what is available, and the quality of that which is available.
QNX is *not* in, or headed toward, the desktop market. For what you want,
knowing the cost of QNX won't help you.
The Amiga NG will use QNX Neutrino as the foundation OS and then have an Amiga
personality layered on top. QNX SSL will give a good quantity based price to
Amiga. Amiga will release a family of products, but the desktop is not in
their immediate plans and there has been no announcement of prices.
> Will someone in this newsgroup please provide a comparrison between
> QNX and Linux as PC type systems. Cost is not a consideration when
> comparing functionality and performance. How complete a system does
> QNX provide? What does QNX require to allow for a complete system?
Linux is not a real time OS; QNX is.
QNX is POSIX.1 compliant and comes with the full set of POSIX.2 utilities.
QNX has three choices when it comes to GUI, one of which is X11.R5
TCP/IP is available.
Linux applications generally will port to QNX with just a few tweaks and a
recompile.
dl...@lehigh.edu
>>One of the reasons why I'm difficult to impress is that I've
>>already seen a fair bit of what ingenuity is capable of.
>
>You are just impressed about how small a specialized piece of code
>can be if is designed for a single purpose !
Why shouldn't I be? And, that being the case, why am I still not overly
impressed by the IAT demo disk? Size, while important, is not all.
>>> It's not just acceleration that I would like to know about. There is
that
>>>> supposed ability to migrate processes, as I understand it, from one
>>>> QNX machine, to another, by dragging across a shared desktop.
>>>
>>>That is misunderstanding. They don't migrate processes, just windows.
>>
>>Oh, now that is a disappointment. Process migration would be something
>>to write home about.
>
>QNX as a DISTRIBUTED RTOS allows the networkwide transparently distribution
>of processes ... that happens when a PHOTON window 'migrates' !
Distributed processes are not in the same category as migrated processes.
A distributed process only ever runs upon one machine, somewhere in a
given network, unless it is migrated. A migrated process is one which has
been started on one machine, and has entirely switched operations to one or
more other machines.
>>I've migrated windows plenty of times, and continue to do so using VNC and
X.
Migrating windows is not quite the same as migrating processes. Migrating
windows is merely a redirection of output. The process which is having its
output redirected is still hosted on the same machine, and so, hasn't
migrated.
>This functionality is not integrated in the underlaying OS.
So what? The difference between microkernels, like QNX, and macrokernels,
a.k.a. monolithic kernels, is one of distributed functionality. What does
it really
matter where the functionality is located?
>Allows VNC to SLIDE a window from screen to screen ? AFAIK it can't do it !
Just because it hasn't been written to, doesn't mean that it can't be done.
I'm not exactly sure what it means to slide a window from screen to screen.
Under X, with virtual desktop space, no problem. However, I'm almost
certain
that you, and others, don't mean that. So what does this screen migration
allow for?
>> Photon might be a bit more efficient
>
>It is more efficient because it's using an already distributed OS platform
!
Whatever works. ;p
<snip><comments on performance><snip>
>>A matter of integration.
>
>That's the key point. This integration must starts with the architecture of
the
>OS and not on top of it! A microkernel design for a windowing system makes
no
>sense if the underlaying OS is a monolithic one.
Why not? I can't see you as being able to justify your claim.
>The AMIGA OS5 will be based on QNX/Neutrino. OS5 inherits the distributed
>microkernel architecture, which means:
>
> - network transperant IPC
> - true micro kernel architecture
> - extreamly high scalabilty from STBs to high-performance OS5
workstations
> - distribution of applications
> - SMP support for powerful real-time applications
> - simplified development of drivers
> - a.s.o.
>
>>Which, while nice, doesn't sell me entirely on QNX being considerably
better,
>>and, hence, worth the price.
>
>Take the attributes above and add the robustness and stability of QNX ...
and
>you will see it's more than 'worth the price' ... it pays back after some
years
>of usage in the field. We have applications running since years in water
power
>plants and steel productions without any reboot actions !
That is an argument which I am willing to consider. Does someone have a
version history, along with release dates, for QNX? I am curious as to just
how immutable QNX has been, and, potentially, will be in the future.
<snip>
>>>When I wanted to impress people, I used to connect 2 boxes by ethernet
>>>and null-modem, then start copying a large file using any program which
>>>provides some sort of 'progress bar'. In the middle of process, I unplug
>>>the ethernet. Then see how copying continues with slower speed. Then
>>>plug ethernet back and see how it becomes faster again and completes as
>>>nothing happened.
>>
>>That strikes me as being a bit more impressive. I am curious as to what
>>is involved enabling QNX to work like that. I can imagine setting similar
>>up for some *nix to do the same, or similar, but that wouldn't happen
"right
>>out of the box." Is the requisite process for QNX similarly involved, or
>>not? Details?
>
>See the architecture guide of Neutrino 2.0 at
>
> http://www.qnx.com/literature/nto_sysarch/intro.htm
http://www.qnx.com/literature/whitepapers/archoverview.html#fleet provides
an overview, but not detail. What about details?
>and you will see what the BASE of OS5 will be.
>
>Again, it will be the BASE of OS5 !!! ... and not the desktop OS for the
AMIGA!
I don't know the differences between OS5 and QNX. What are they?
>[ clip ..]
>>So run some dummy background applications to highlight the differences
>>between single tasking performance, dual tasking performance, triple
>>tasking performance, quad...unto 9999. Do the same for Linux, or BSD,
>>or some commercial *nix. If QNX can be shown to significantly outstrip
>>the competition in throughput, then I believe that I would be
significantly
>>impressed enough to buy into QNX as a desktop type system.
>
> 150.000 task switches/s (Send/Reply) between two user level tasks ... is
> it fast enough for you ? (K6/266Mhz)
I don't know. How do I know that your results are accurate? Also, finding
similar benchmarks of 'realtime' Linuxes is seemingly difficult. Where
might I find statistics on "the competition?"
> Armin
Derrick Shearer
dl...@Lehigh.EDU
Igor Kovalenko
>
> One of the reasons why I'm difficult to impress is that I've
> already seen a fair bit of what ingenuity is capable of.
>
Lucky guy :)
> > It's not just acceleration that I would like to know about. There is that
> >> supposed ability to migrate processes, as I understand it, from one
> >> QNX machine, to another, by dragging across a shared desktop.
> >
> >That is misunderstanding. They don't migrate processes, just windows.
>
> Oh, now that is a disappointment. Process migration would be something
> to write home about. I've migrated windows plenty of times, and continue
> to do so using VNC and X. Photon might be a bit more efficient in its
> implementation, but how much so, I'm not sure I could very well guess; too
> many factors to account for.
Now this is misconception. You don't need to migrate processes when any
process can access any resource over network just as it is local. You
can sit on node 1, lauch a program stored on node 2 to run on node 3 and
redirect its i/o to node 4. Not just stdin/out but anything it tries to
access may be transparently redirected to another node (files, devices,
whatever).
The network is just one big distributed computer. What is Sun always
dreamed about :)
> That strikes me as being a bit more impressive. I am curious as to what
> is involved enabling QNX to work like that. I can imagine setting similar
> up for some *nix to do the same, or similar, but that wouldn't happen "right
> out of the box." Is the requisite process for QNX similarly involved, or
> not?
> Details?
It is out of box functionality. You have to setup network, of course.
I'm talking about QNX network, not IP. That whole job will be made my
Net manager. In cooperation with kernel it provides fault-tolerant
network transparent message passing. If you have more than one link,
when one fails packates will be auto redirected to other available. And
when more than one available, it will automatically balance the load as
well.
> I admit that Doom is perhaps not the best choice for a benchmark. I
> would still like to see numbers for it, though. Even moreso with other
> processes running, so as to encourage multitasking, under both QNX
> and Linux, or BSD, or some commercial *nix. Indication as to fps, when
> observed in an interactive context, provides, IMO, the best performance
> feedback possible.
>
Context switching time alone is pretty indicative too. For instance, a
DB server will be able to perform more transactions per second when
context switch time is fast.
- Igor
Armin Steinhoff
>
>Armin Steinhoff wrote in message <7bb7ul$b...@drn.newsguy.com>...
>>In article <7batvs$1b...@fidoii.cc.Lehigh.EDU>, <dl...@Lehigh.EDU says...
><snip>
>>>One of the reasons why I'm difficult to impress is that I've
>>>already seen a fair bit of what ingenuity is capable of.
>>
>>You are just impressed about how small a specialized piece of code
>>can be if is designed for a single purpose !
>
>
>Why shouldn't I be? And, that being the case, why am I still not overly
>impressed by the IAT demo disk? Size, while important, is not all.
It is important for embedded systems ... for that reason was the IAT demo disk
produced.
[ clip ..]
>
>>This functionality is not integrated in the underlaying OS.
>
>
>So what? The difference between microkernels, like QNX, and macrokernels,
>a.k.a. monolithic kernels, is one of distributed functionality. What does
>it really matter where the functionality is located?
You can't build an Indy car out of a family car :-)
[ clip ..]
>>The AMIGA OS5 will be based on QNX/Neutrino. OS5 inherits the distributed
>>microkernel architecture, which means:
>>
>> - network transperant IPC
>> - true micro kernel architecture
>> - extreamly high scalabilty from STBs to high-performance OS5
>workstations
>> - distribution of applications
>> - SMP support for powerful real-time applications
>> - simplified development of drivers
>> - a.s.o.
>>
>>>Which, while nice, doesn't sell me entirely on QNX being considerably
>>>better, and, hence, worth the price.
>>
>>Take the attributes above and add the robustness and stability of QNX ...
>>and you will see it's more than 'worth the price' ... it pays back after some
>>years of usage in the field. We have applications running since years in water
>>power plants and steel productions without any reboot actions !
>
>
>That is an argument which I am willing to consider. Does someone have a
>version history, along with release dates, for QNX? I am curious as to just
>how immutable QNX has been, and, potentially, will be in the future.
QNX is 18 years young :-) and we have seen lots of improvements in the past and
I'm sure we will see lot of improvements in the future.
[ .. ]
>http://www.qnx.com/literature/whitepapers/archoverview.html#fleet provides
>an overview, but not detail. What about details?
You should ask QSSL about details.
>>and you will see what the BASE of OS5 will be.
>>
>>Again, it will be the BASE of OS5 !!! ... and not the desktop OS for the
>AMIGA!
>
>
>I don't know the differences between OS5 and QNX. What are they?
OS5 'is under construction' and QNX exist since 18 years :-)
Again ... there are two systems from QSSL: QNX4 and QNX/Neutrino.
QNX/Neutrino will be the base of OS5.
[ clip ..]
>> 150.000 task switches/s (Send/Reply) between two user level tasks ... is
>> it fast enough for you ? (K6/266Mhz)
>
>
>I don't know. How do I know that your results are accurate?
OK ... make your own tests. But don't use a Pentium II with 450Mhz or so ... the
result could impress you :-)
>Also, finding similar benchmarks of 'realtime' Linuxes is seemingly difficult. >
>Where might I find statistics on "the competition?"
When I remember right RT LINUX is at least ~20 times slower.
Armin
Daryl Low
> Why shouldn't I be? And, that being the case, why am I still not overly
> impressed by the IAT demo disk? Size, while important, is not all.
It's like money, it's not everything but still can be very useful.
Because QNX can fit in small places (and maintain full functionality),
you can cram it into anything from a palmtop up to big super-servers.
Also, small = less code = less execution overhead = fewer bugs = less
grief. But at the same time, all the core OS features of NT and other
*NIXs are there (ie. preemptive multitasking, memory protection,
etc.....).
> Distributed processes are not in the same category as migrated processes.
> A distributed process only ever runs upon one machine, somewhere in a
> given network, unless it is migrated. A migrated process is one which has
> been started on one machine, and has entirely switched operations to one or
> more other machines.
That would be nice, but alas QNX has not entered the process load
balancing world yet. I'm not sure if the RT world is looking for that
right now (due to non-deterministic behaviour of migrating entire
processes over an interconnect). But I'm sure it would be possible to
port existing cluster software (Beowulf etc...) to work under QNX.
> So what? The difference between microkernels, like QNX, and macrokernels,
> a.k.a. monolithic kernels, is one of distributed functionality. What does
> it really matter where the functionality is located?
What comes to mind when I say "upgrade video driver" or "change SCSI
driver"?
Winblowz: "Your computer will now restart"
Linux: make config, make clean, make dep, make bzImage, make install,
shutdown, reboot
QNX: kill <driver pid>, <new driver name> &
> Just because it hasn't been written to, doesn't mean that it can't be done.
> I'm not exactly sure what it means to slide a window from screen to screen.
> Under X, with virtual desktop space, no problem. However, I'm almost
> certain that you, and others, don't mean that. So what does this screen
> migration allow for?
They mean you can drag between any two montors (local or remote).
Winblowz 98 just came out with multi-monitor support and only because
they added that on, not because it was inherent in the OS/GUI.
X can probably do the same, but I shudder to think of how I'd pull it
off under XFree86. I managed to rig the same demo at home in about 30
minutes (most of the time being spent reading docs on setting up
networking), the Photon part was dead easy.
> That is an argument which I am willing to consider. Does someone have a
> version history, along with release dates, for QNX? I am curious as to just
> how immutable QNX has been, and, potentially, will be in the future.
I think www.qnx.com has a timeline. That's the best I can offer. QNX
is about 20 years old. They're at version 4.25 now and NTO 2.0 is just
about caught up with QNX in terms of bells and whistles (drivers, GUI,
etc...). But growing pains are inevitable.
Also, just looking around, I notice *lots* of people still using QNX 2.x
in mission critical realtime stuff.
> I don't know the differences between OS5 and QNX. What are they?
QNX made by QSSL. OS5 under construction by Amiga Inc.
OS5 is probably what you want your comparison to be done with. It is a
real desktop/consumer OS built using the kernel from NTO.
> I don't know. How do I know that your results are accurate? Also, finding
> similar benchmarks of 'realtime' Linuxes is seemingly difficult. Where
> might I find statistics on "the competition?"
If you're looking for a Byte/PC Mag style review of QNX vs. M$ vs. Linux
vs. BSD you're gonna be hard pressed to find one. If you did, it would
be an apples to oranges comparison with M$ winning by sheer quantity of
BS (not to mention the editors are heavily M$ biased $$$$$$), Linux
second (by geek appeal) and then BSD close behind.
QNX does not have the office suites, dancing paperclips, Quakes,
StarCrafts and other consumer goodies because that was not what QNX was
designed for. QNX was designed to be small and unobtrusive ensuring
that important processes (like the ones for emergency reactor shutdowns)
get the CPU cycles they need to do their job. Adding fancy stuff on top
is an exercise to the developers who purchase QNX (like Amiga).
QSSL knows they don't have the resources to lock horns with M$ on M$'s
home turf (workstations, servers, desktop) and QNX would not be the
fault-tolerant RTOS it is today if it was open sourced (Linux/RT still
has a loooooong way to go). As it stands, QSSL is very profitable,
expanding on their home turf (embedded/RT) and digging in to squash
RT/WinCE (oxymoron in my opinion). In this industry, M$ style hoopla
doesn't hold up very well because the embedded/RT community comprise
(mainly) of competent engineers who know their stuff (and know that CE
just can't cut it).
If you sift through some Embedded/RT magazines, you might be able to
find QNX vs. VxWorks vs. psos vs. etc... that does an apples to apples
comparison (little good that may be to you since you probably never
heard of the other listed OSs).
If you want to see what QNX can do for the average Joe as a desktop OS,
you'll just have to wait for OS5 like the rest of us.
Gene Poole
dls2> macrokernels, a.k.a. monolithic kernels, is one of distributed
dls2> functionality. What does it really matter where the functionality is
dls2> located?
Yes it matters a great deal.
First, in QNX, distributing the function to seperate processes provides MMU
protection to the kernel and between processes. If a process is poorly behaved
(as will often be the case during development) the process will die with a
SIGSEGV and, at worst, a post mortem dump can be submitted to the symbolic
debugger and the offending code found. The kernel "takes no lickin' it just
keeps on tickin'". The offending process can be swept away and a replacement
process started 'on the fly', perhaps by a software watchdog process. The MMU
protection between processes is a large reason for QNX robustness.
With a monolithic kernel, if any thread gets a runaway pointer, the whole
system goes belly up (BSOD).
Second, QNX implements network transparency for Inter Process Communication
(IPC). By using seperate processes and IPC the OS is easily distributed across
the network. A QNX network behaves much more like a single large computer.
QNX applications benefit from this architecture as well. An application, which
was well designed as a team of processes on a single machine, can at a later
time be distributed across a network with zero change to the source, and not
even needing to be relinked with a set of network oriented libraries. The
binaries can be loaded onto the processor that makes the most sense for the
configuration.
Third, by using a set of processes, when software must be changed to meet new
application requirements, the changes can be made to the affected processes and
then "hot swapped" with the old. QNX gives a whole new meaning to 7*24.
Fourth, QNX resource managers (eg file systems) are all implemented with a
standard IPC message interface. This standard is published by QSSL and sample
code for resource managers available. It is relatively easy for a non-QSSL
developer to implement a special purpose file system WITHOUT EVER HAVING ACCESS
TO THE KERNEL SOURCE CODE OR HAVING TO SIGN A NDA. This is an effective
argument against the need for an Open Source kernel (if the kernel is QNX).
Regards
GEve...@escape.com
"Gene Poole" <Gene....@offsbbs.com> wrote:
> On Feb 26 13:39, 1999, <dl...@Lehigh.EDU> of 510:563/1 wrote:
>
> > ... I have used the "IAT Demo Disk - Version 3.03", and, while reasonably
> > nice, it didn't overly impress me.
>
> You're either very hard to impress, or you didn't understand what you were
> seeing. To do the equivalent with Internet Explorer and NT 4.0, how many
> diskettes would you need: 20, 40, more?
>
used
>
> > QNX.
>
> scenes: nuclear reactors, ATM's, shuttle arms. You know, dull stuff like
that,
> that has no effect on your life.
>
> # Origin: Variation on a Theme *** Gloucester ON CA (510:563/1.1)
I understand where the guy could say he wasn't overly empressed.
It dosn't meen that QNX isn't great, but comming from the Amiga
community, we all have experiance useing and OS and browser that
can fit on one disk, though to be fair, out disks were 1.7K instead
of 1.4K.
QNX neutrino will be a great kernal for the Amiga OS 5, because
while Commodore was waisting time and technology, QNX was slowly
doing all the things Amiga OS guys would have wanted to do if they
had a real company behind them.
Jerry Everett
Igor Kovalenko
>
> Igor Kovalenko wrote in message <36D88522...@infomarket.ru>...
> The demo ran OK, and I was not disappointed with the video performance,
> due to the fact that it was acceptably fast, even though using only generic
> drivers. I would not, however, expect DOS to allow for less performance,
> given similar supporting functionality, such as TCP/IP, a windowing system,
> and similar programs, running on sufficiently fast hardware. I would expect
> an equivalent DOS based setup to possibly be a little larger, but, other
> than
> that, what differences to be seen?
First, you have to realize difference between DOS and an OS.
DOS is utility to store files on hard drive.
Windows is utility to print them.
An OS is something different :)
The QNX demo is not just a little bit smaller, it includes a lot more
functionality than DOS provides. I don't want to bring a lesson about OS
here, just a little example to illustrate what kind of size you can get
if you strip out functionality which is invisible for you.
Some guys from Russia once made another demo to impress people. It was
targeted for DOS and used DOS only to load the executable. Then that
executable switch to protected mode and forget about DOS, implementing
their own multitasking environment and simple windowing system. There
were 3 threads with corresponding windows, each displaying simulated
flight over Mars surface in 3D (like a flight simulator game). There was
a way to switch beween windows and change direction of flight from
keyboard. The background could be either preloaded picture or
dynamically generated fractal. Fractal technology has been used for Mars
surface simulation as well. Another windows contained the "magic eye" (a
'recursive' picture of the whole screen including itself, etc, etc) also
updated in realtime along with flight. The video resolution was
800x600x256, about fps I'm not sure but it looked pretty much like 15fps
or more when I ran that thing on an old 486.
Now about size. The _whole_thing_ was 20 (twenty) Kb. No kidding and no
compression whatsoever. I think I still have that demo somewhere if
somebody is intrigued. Most people used to think that video driver alone
must be bigger :)
> It's not just acceleration that I would like to know about. There is that
> supposed ability to migrate processes, as I understand it, from one
> QNX machine, to another, by dragging across a shared desktop.
That is misunderstanding. They don't migrate processes, just windows.
> While the description sounds interesting, how am I to know that the
> implementation lives up to the claims?
Pretty close. The magic thing is, you don't have to do anything special
in application for that. QNX and Photon implement that effect
transparently. So, you can migrate windows not because they programmed
an app to do so, but just because in QNX/Photon it is as natural as move
them inside your desktop.
> Furthermore, what other
> nifty features exist for QNX which are not readily available for other
> platforms?
When I wanted to impress people, I used to connect 2 boxes by ethernet
and null-modem, then start copying a large file using any program which
provides some sort of 'progress bar'. In the middle of process, I unplug
the ethernet. Then see how copying continues with slower speed. Then
plug ethernet back and see how it becomes faster again and completes as
nothing happened.
> I am sure that the video performance would be much better. But what
> would the numbers be? Running Quake, or Doom, or some other
> multiple platform game, e.g., how many fps difference might be expected?
I don't think there will be big difference, unless you will play network
game.
> That would certainly depend upon the drivers, but if the drivers were well
> matched, then other aspects of system performance could be compared
> as a rough increase in system throughput, identified by fps.
>
> Consider one system. At one time loaded with DOS, using generic video
> drivers, running Doom. At another time loaded with QNX, using generic
> video drivers, running Doom. Since the video drivers and the game are,
> for all practical purposes, the same, then the only difference measurable
> is the operating system's ability to cope with demand.
>
Doom is particularly bad thing, it consumes almost all CPU time on its
priority level running READY most of time. That conflicts with
philosophy of QNX. If you want to get some real difference, wait until
it will be available for NTO, then try to run multiplayer game on SMP
system, comparing NTO and NT. There will be moment of truth.
> If QNX can outperform DOS in terms of throughput, how much would it be
> by? If QNX cannot outperform DOS in terms of throughput, then why might
> I be interested in it, when other systems offer equivalent functionality,
> albeit
> for larger size?
Yep, why would you need a Corvet when there is a pace bus? Where do you
think you'll feel yourself better? :)
> A better grounds for comparison might be QNX running Doom with
> unaccelerated video drivers contrasted to Linux running Doom with
> unaccelerated video drivers, both run on a single machine so as not
> to let machine differences interfere with testing results.
Singe application whatever it is, is not adequate test for a
multitasking OS. You basically will just test your hardware.
- Igor
dl...@lehigh.edu
>dl...@Lehigh.EDU wrote:
<snip><Mars demo><snip>
>compression whatsoever. I think I still have that demo somewhere if
>somebody is intrigued. Most people used to think that video driver alone
>must be bigger :)
If its demos you want: http://www.hornet.org
How about 4Kb demos?...
One of the reasons why I'm difficult to impress is that I've
already seen a fair bit of what ingenuity is capable of.
> It's not just acceleration that I would like to know about. There is that
>> supposed ability to migrate processes, as I understand it, from one
>> QNX machine, to another, by dragging across a shared desktop.
>
>That is misunderstanding. They don't migrate processes, just windows.
Oh, now that is a disappointment. Process migration would be something
to write home about. I've migrated windows plenty of times, and continue
to do so using VNC and X. Photon might be a bit more efficient in its
implementation, but how much so, I'm not sure I could very well guess; too
many factors to account for.
>> While the description sounds interesting, how am I to know that the
>> implementation lives up to the claims?
>
>Pretty close. The magic thing is, you don't have to do anything special
>in application for that. QNX and Photon implement that effect
>transparently. So, you can migrate windows not because they programmed
>an app to do so, but just because in QNX/Photon it is as natural as move
>them inside your desktop.
A matter of integration. Which, while nice, doesn't sell me entirely on
QNX being considerably better, and, hence, worth the price. I already know
how to 'effortlessly' (?) export the displays of programs which weren't
specifically designed for such on operating systems which weren't
specifically
designed for such. What else does QNX offer?
>> Furthermore, what other
>> nifty features exist for QNX which are not readily available for other
>> platforms?
>
>When I wanted to impress people, I used to connect 2 boxes by ethernet
>and null-modem, then start copying a large file using any program which
>provides some sort of 'progress bar'. In the middle of process, I unplug
>the ethernet. Then see how copying continues with slower speed. Then
>plug ethernet back and see how it becomes faster again and completes as
>nothing happened.
That strikes me as being a bit more impressive. I am curious as to what
is involved enabling QNX to work like that. I can imagine setting similar
up for some *nix to do the same, or similar, but that wouldn't happen "right
out of the box." Is the requisite process for QNX similarly involved, or
not?
Details?
>> I am sure that the video performance would be much better. But what
>> would the numbers be? Running Quake, or Doom, or some other
>> multiple platform game, e.g., how many fps difference might be expected?
>
>I don't think there will be big difference, unless you will play network
>game.
My opinion differs, but, unless its tried and tested, who knows for sure? I
understand that the QNX microkernel only operates when requested, but,
there is more to an operating *system* than just its kernel. Like it or
not,
overhead accumulates, is measurable, and is something which would be
nice for comparrison. Benchmarks, whatever they may be...
>> That would certainly depend upon the drivers, but if the drivers were
well
>> matched, then other aspects of system performance could be compared
>> as a rough increase in system throughput, identified by fps.
>>
>> Consider one system. At one time loaded with DOS, using generic video
>> drivers, running Doom. At another time loaded with QNX, using generic
>> video drivers, running Doom. Since the video drivers and the game are,
>> for all practical purposes, the same, then the only difference measurable
>> is the operating system's ability to cope with demand.
>
>Doom is particularly bad thing, it consumes almost all CPU time on its
>priority level running READY most of time. That conflicts with
>philosophy of QNX. If you want to get some real difference, wait until
>it will be available for NTO, then try to run multiplayer game on SMP
>system, comparing NTO and NT. There will be moment of truth.
I admit that Doom is perhaps not the best choice for a benchmark. I
would still like to see numbers for it, though. Even moreso with other
processes running, so as to encourage multitasking, under both QNX
and Linux, or BSD, or some commercial *nix. Indication as to fps, when
observed in an interactive context, provides, IMO, the best performance
feedback possible.
<snip>
>Singe application whatever it is, is not adequate test for a
>multitasking OS. You basically will just test your hardware.
So run some dummy background applications to highlight the differences
between single tasking performance, dual tasking performance, triple
tasking performance, quad...unto 9999. Do the same for Linux, or BSD,
or some commercial *nix. If QNX can be shown to significantly outstrip
the competition in throughput, then I believe that I would be significantly
impressed enough to buy into QNX as a desktop type system. Otherwise,
it just doesn't stand a chance of fitting the bill for general, or my,
demand.
>- Igor
Andrew Stuart
after;
http://www.qnx.com/literature/whitepapers/photonmicrogui.html
http://www.qnx.com/literature/whitepapers/archoverview.html
There are so many features that are stronger and more powerful than NT that
it just isn't possible to note them all here. You have to read the
whitepapers.
Andrew
ast...@werple.net.au
dl...@Lehigh.EDU wrote in message <7ba5ml$s...@fidoii.cc.Lehigh.EDU>...
>Igor Kovalenko wrote in message <36D88522...@infomarket.ru>...
>>Gene Poole wrote:
>>> On Feb 26 13:39, 1999, <dl...@Lehigh.EDU> of 510:563/1 wrote:
>>>
>>> > ... I have used the "IAT Demo Disk - Version 3.03", and, while
>reasonably
>>> > nice, it didn't overly impress me.
>>>
>>> You're either very hard to impress, or you didn't understand what you
>were
>>> seeing. To do the equivalent with Internet Explorer and NT 4.0, how
many
>>> diskettes would you need: 20, 40, more?
>>
>>He is probably referring to slow video performance. But being familar
>>with all kinds of Windows and *nixes he should know that no one of those
>>systems can provide support with high resolution for all vesa2
>>compatible chipsets in single driver (Pg.flat). I remember somebody
>>suggested to develop such a driver to XFree86 team and they answered
>>that it is technically impossible.
>
>The demo ran OK, and I was not disappointed with the video performance,
>due to the fact that it was acceptably fast, even though using only generic
>drivers. I would not, however, expect DOS to allow for less performance,
>given similar supporting functionality, such as TCP/IP, a windowing system,
>and similar programs, running on sufficiently fast hardware. I would
expect
>an equivalent DOS based setup to possibly be a little larger, but, other
>than
>that, what differences to be seen?
>
>>But you can't buy something for nothing, so Pg.flat doesn't do any
>>acceleration.
>
>
>It's not just acceleration that I would like to know about. There is that
>supposed ability to migrate processes, as I understand it, from one
>QNX machine, to another, by dragging across a shared desktop.
>While the description sounds interesting, how am I to know that the
>implementation lives up to the claims? Furthermore, what other
>nifty features exist for QNX which are not readily available for other
>platforms? I would like for someone to compare QNX as a desktop
>operating system to Linux as a desktop operating system. I can't. I
>have never, to my knowledge, either seen or used QNX.
>
>>> > Other than that one demo, I have never, to my knowledge, either seen
>>> > or used QNX.
>>
>>Try to install real QNX/Photon with an accelerated driver and feel the
>>difference.
>
>I am sure that the video performance would be much better. But what
>would the numbers be? Running Quake, or Doom, or some other
>multiple platform game, e.g., how many fps difference might be expected?
>That would certainly depend upon the drivers, but if the drivers were well
>matched, then other aspects of system performance could be compared
>as a rough increase in system throughput, identified by fps.
>
>Consider one system. At one time loaded with DOS, using generic video
>drivers, running Doom. At another time loaded with QNX, using generic
>video drivers, running Doom. Since the video drivers and the game are,
>for all practical purposes, the same, then the only difference measurable
>is the operating system's ability to cope with demand.
>
>If QNX can outperform DOS in terms of throughput, how much would it be
>by? If QNX cannot outperform DOS in terms of throughput, then why might
>I be interested in it, when other systems offer equivalent functionality,
>albeit
>for larger size?
>
>A better grounds for comparison might be QNX running Doom with
>unaccelerated video drivers contrasted to Linux running Doom with
>unaccelerated video drivers, both run on a single machine so as not
>to let machine differences interfere with testing results.
>
Andrew Stuart
Faster
Easier to program
(Much) more stable
More functionality
More flexible
More scalable
More exciting
Andrew
ast...@werple.net.au
dl...@Lehigh.EDU wrote in message <7b7sts$l...@fidoii.cc.Lehigh.EDU>...
>Guy Macon wrote in message <7b7hua$9...@chronicle.concentric.net>...
><snip>
>>Take one engineer, familiar with Windows NT. Sit him down at an old 486
>>with a full installation of QNX and Photon. Leave the architecture guide
>>nearby - he WILL be interested in a few moments. Say "Go ahead and try
>>anything. Doesn't matter what you do because I can reload with a couple
>>of keystrokes". Walk away. Go tu cubicle. Wait for amazed person to
>>come by, ranting and raving over fast performance, plus a few minor
>>"confused by the GUI". problems. Followup with comments about ANSI and
>>Posix standards vs. proprietary shifting standards from the M$ Borg.
>
>I am familiar with Windows NT. I am familiar with Windows 98/95/3.x, too.
>I am familiar with both DOS and Linux. I am marginally familiar with BSD,
>Solaris, IRIX, and so, I suppose, then, most all *nix variants. I am not
>very
>much familiar with QNX. I have used the "IAT Demo Disk - Version 3.03",
>and, while reasonably nice, it didn't overly impress me. Other than that
>one demo, I have never, to my knowledge, either seen or used QNX.
>
>To a complete QNX newbie, please explain how well QNX can serve as a
>home operating system. How much does it cost, for what functionality, that
>is how usable?
>
>Why might I like QNX more, or less, than, e.g., BSD or Linux?
>
>Derrick Shearer
>dl...@Lehigh.EDU
>
>
Pat Ford
Joe Cosby <joec...@pujnod.seatac.net> wrote in article
<r6KB2.329$%a5....@news6.ispnews.com>...
| ** To reply in e-mail, remove "pujnod." from address **
|
| "Pat Ford" hunched over his computer, typing feverishly;
I don't type fast enough for it to be called feverishly 8-)
| thunder crashed, "Pat Ford" laughed madly, then wrote:
| >
| > yes in the to the aviation engine test stands and increasingly in the
| > automotive
| > manufacturing area. There is automotive application that uses a gui
comming
| > up
| >
|
| Dashboard Quake, anyone?
would hace been fun this morning ( snow and bad drivers up here)
|
| ...and you thought car phones were bad...
|
| --
| ----------------------------------------------------
| Joe Cosby
|
| Devout member of the Church of Amiga since 1990
|
| "Whatever you can do, or dream you can, begin it.
| Boldness has genius, power and magic in it" - Goethe
| ----------------------------------------------------
|
--
***********************************
* Pat Ford R&D developer *
* QNX Software Systems *
* www.qnx.com *
* Any opinions expressed *
* are my own and not those*
* of my employer *
***********************************
Vadim Radionov
Armin Steinhoff wrote:
AS> A microkernel design for a windowing system makes no sense if the
AS> underlaying OS is a monolithic one.
No arguments.
(I see sense, to separate the graphics driver from the main windowing
magic, such as regions, events...)
[]
AS> 150.000 task switches/s (Send/Reply) between two user level tasks
AS> ... is it fast enough for you ? (K6/266Mhz)
643128 task switches/s (sched_yield)
...(P166 MMX)
Vadim
Vadim Radionov
dl...@Lehigh.EDU wrote:
d> I don't know. How do I know that your results are accurate? Also,
d> finding similar benchmarks of 'realtime' Linuxes is seemingly
d> difficult.
Following test have been adopted by mine for linux:
(this code is based on QNX test written by Mike Nefodov and sent me by
Vladimir Doshevsky)
=== Cut ===
#include <stdio.h>
#include <time.h>
#include <sched.h>
#include <stdlib.h>
#include <unistd.h>
#include <sys/wait.h>
#include <sys/mman.h>
#include <fcntl.h>
#include "outinfo.c"
int *init_shm();
void done_shm();
main(int argc, char *argv[])
{
long scount = 1000000, i;
clock_t c;
int fault = 0;
int status;
int *swtchno;
struct sched_param param;
if (argc > 1)
scount = atoi(argv[1]);
swtchno = init_shm();
scount /= 2; // Convert count of switches to count of iter.
param.sched_priority = 9;
if( sched_setscheduler(0, SCHED_FIFO, ¶m) == -1 )
printf( "warning: you aren't suser, so SCHED_OTHER will be used\n" );
if ((fork() > 0)) {
c = clock();
*swtchno = 0;
for (i = 0; i < scount; i++) {
if (*swtchno != 0) fault++;
*swtchno = 1;
sched_yield();
}
scount *= 2; // Restore original meaning of scount (count of
// switches)
outinfo( "Context switch", scount, clock() - c );
printf( "switch fault count %d\n", fault);
wait( &status );
} else {
for (i = 0; i < scount; i++) {
sched_yield();
*swtchno = 0;
}
sched_yield();
done_shm();
// pause();
}
return EXIT_SUCCESS;
}
char mo_name[] = { "shared_object" };
int *init_shm()
{
int fd;
int *addr;
fd = open( mo_name, O_RDWR | O_CREAT, 0600 );
if( fd < 0 )
{
perror( "shm_open" );
exit( EXIT_FAILURE );
}
if( ftruncate( fd, 1024) == -1 )
{
perror( "ftruncate" );
exit( EXIT_FAILURE );
}
addr = (int *)mmap(0, 1024, PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0);
if( addr == (int *)-1 )
{
perror( "mmap" );
exit( EXIT_FAILURE );
}
close( fd );
return addr;
}
void done_shm()
{
unlink( mo_name );
}
=== Cut ===
the result is near 371000 task switches/s. That's on Pentium 200 MMX.
to compare, the result in QNX is exactly 643128 switches/s on Pentium
166 MMX.
Vadim
Paul May
> > size for the very simple reason that it expected to be run on systems
> > having multiple gigabyte hard disk drives. Size is not an issue, there.
>No comment. <big grin>
ROTFL, yep i think he got his reasoning the wrong way around on that one.
still i'm not complaining, if it wasnt for (how shall i put this) SLACK
programing in the M$ world, we wouldn't have these cheap GIGs of HD and ram
as commodity items.
>The Amiga NG will use QNX Neutrino as the foundation OS and then have an
>Amiga personality layered on top. QNX SSL will give a good quantity based
>price to Amiga. Amiga will release a family of products, but the desktop is
>not in their immediate plans and there has been no announcement of prices.
theres a new high ranking gateway Captain at the Ainc helm now, and his
statment on the AMIGA site says there is a comitment to the desktop and
high end machines (later) hopefully he will get things moving and we
should hear some real news soon (perhaps the 12th may show ?).
>Linux applications generally will port to QNX with just a few tweaks and a
>recompile.
is this also true of Nto ?, hopefully yes, and presumably easyer to
port the future OS5/Nto that it is currently to AMIGA OS3.1.
Hmmm, has anyone ported the open source VNC client/server S/W
to Nto yet.
i'd like to see that on the OS5/Nto when the developer boxes ship
http://www.uk.research.att.com/vnc/ for anyone interested.
loads of platforms already have it ported, including AMIGA but i didnt
see mention of QNX or Nto.
--
Paul May, UK
Team *AMIGA* , *KOSH* , *OwlNet*
KO...@reedsweb.net
--
http://www.KOSH.net
Kommunity Operating System and Hardware
http://www.owlnet.net Open Platform Support Services (Now Open)
Paul May
>>
>reasonably
>> > nice, it didn't overly impress me.
>>
>>seeing. To do the equivalent with Internet Explorer and NT 4.0, how many
>>diskettes would you need: 20, 40, more?
>Consider me rather hard to impress. Other than the size of the demo, what
>is there to it? Its speed is hardly impressive, considering the fact that
>it
>was uncompressed into a RAM drive. Its interface is crude. There is not
>much in the way of multitasking or realtime response with which to test it.
>While it can be useful to some in certain circumstances, it isn't useful to
>me,
>under any circumstances of which I am currently aware.
>Microsoft Windows does not make any pretenses as to its smallness of
>size for the very simple reason that it expected to be run on systems having
>multiple gigabyte hard disk drives. Size is not an issue, there.
>> > Other than that one demo, I have never, to my knowledge, either seen or
>> > used QNX.
>>
>>QNX generally works on mission critical real-time applications behind the
>>scenes: nuclear reactors, ATM's, shuttle arms. You know, dull stuff like
>that,
>>that has no effect on your life.
>I am aware of where QNX primarily sees use. However, you are wrong.
>Read the topic. "Cheap Amigas and costly QNX"...
Arrrrrrrr, now i thought that this refering to Nto as QNX would come
back and bite someone.
the thing is the subj *should* have been called
Cheap Amigas and Nto (short for Neutrino) as that is infact the
base OS that AMIGA inc is useing NOT QNX.
right now, other than *maybe* a handful of people here and QSSL if you sign
an NDA with them can give you the information you want about the desktop OS.
the reason being simple, the production or even the Developer OS5/Nto
isn't out yet, so no one knows the full spec, or even if the generic
QSSL Neutrino and the AMIGA OS5/Nto version will be cross compatable.
i hope noone minds, i'v crossposted this to C.S.A.Misc so that they
can see all this very nice QNX/Nto info too.
its a shame they missed Igor's post about some of the QNX features
and i assume Nto has the same or better functionality ?.
<sliped Igor's interesting bits in here>
Igor:
Now this is misconception. You don't need to migrate processes when any
process can access any resource over network just as it is local. You
can sit on node 1, lauch a program stored on node 2 to run on node 3 and
redirect its i/o to node 4. Not just stdin/out but anything it tries to
access may be transparently redirected to another node (files, devices,
whatever).
The network is just one big distributed computer. What is Sun always
dreamed about :)
<and the other bit>
>>Igor:
>>When I wanted to impress people, I used to connect 2 boxes by ethernet
>>and null-modem, then start copying a large file using any program which
>>provides some sort of 'progress bar'. In the middle of process, I unplug
>>the ethernet. Then see how copying continues with slower speed. Then
>>plug ethernet back and see how it becomes faster again and completes as
>>nothing happened.
>dls2:
>That strikes me as being a bit more impressive. I am curious as to what
>is involved enabling QNX to work like that. I can imagine setting similar
>up for some *nix to do the same, or similar, but that wouldn't happen
>"right
>out of the box." Is the requisite process for QNX similarly involved, or
>not? Details?
Igor:
It is out of box functionality. You have to setup network, of course.
I'm talking about QNX network, not IP. That whole job will be made my
Net manager. In cooperation with kernel it provides fault-tolerant
network transparent message passing. If you have more than one link,
when one fails packates will be auto redirected to other available. And
when more than one available, it will automatically balance the load as
well.
<end of previous posts>
>QNX seems to be
>making a move towards the desktop scene, and, that being the case, I
>am curious about how it is for a desktop system. I am aware that QNX is
>expensive, but I do not know how much for exactly what I want. How can I
>know when I am not exactly sure what I want, because I am not even sure
>what is available, and the quality of that which is available.
we dont know the price of the new OS5/Nto so you will need to wait
and perhaps come back later when the info is available.
Guy Macon
>Linux applications generally will port to QNX with just a few tweaks and a
>recompile.
This is VERY interesting. What kind of tweaks? How hard is this to do?
Guy Macon
>That is an argument which I am willing to consider. Does someone have a
>version history, along with release dates, for QNX? I am curious as to just
>how immutable QNX has been, and, potentially, will be in the future.
Especially, how long has the microkernel been unchanged? I don't much
care about changes to the video code because I can crash the video
code or even change video drivers without any interruption of the rest
of what QNX is doing (robot wrangling in my case).
Guy Macon
>>That is misunderstanding. They don't migrate processes, just windows.
>
>Oh, now that is a disappointment. Process migration would be something
>to write home about. I've migrated windows plenty of times, and continue
>to do so using VNC and X. Photon might be a bit more efficient in its
>implementation, but how much so, I'm not sure I could very well guess; too
>many factors to account for.
You are still thinking "computers communicating with each other"
QNX isn't like that. QNX is "One big computer running lots of tasks".
If two tasks are on the same CPU, they communicate fast but steal CPU
cycles from each other. If two tasks are on differentmachines, they
communicate slower (not much with GigaBit networks!) and don't steal
CPU cycles from each other. Put the machines on differend hemisheres
and use a 300 bps modem to connect them and the communication gets
really slow but in all cases it feels like programing one big computer.
In article <7ba5ml$s...@fidoii.cc.Lehigh.EDU>, dl...@Lehigh.EDU wrote:
>
>It's not just acceleration that I would like to know about. There is that
>supposed ability to migrate processes, as I understand it, from one
>QNX machine, to another, by dragging across a shared desktop.
>While the description sounds interesting, how am I to know that the
>implementation lives up to the claims? Furthermore, what other
>nifty features exist for QNX which are not readily available for other
>platforms? I would like for someone to compare QNX as a desktop
>operating system to Linux as a desktop operating system. I can't. I
>have never, to my knowledge, either seen or used QNX.
I can do that.
Linux as a desktop operating system: Pretty good. Getting better.
QNX as a desktop operating system: QNX is not a desktop operating system.
Possible QNX-based Amiga OS as a desktop operating system: ...Interesting.
In article <7ba28n$k...@fidoii.cc.Lehigh.EDU>, dl...@Lehigh.EDU wrote:
>I am aware of where QNX primarily sees use. However, you are wrong.
>Read the topic. "Cheap Amigas and costly QNX"... QNX seems to be
>making a move towards the desktop scene, and, that being the case, I
>am curious about how it is for a desktop system.
QNX is in no way making a move towards the desktop scene. They are
and will continue to be in the Real Time scene. Different world.
AMIGA (The new Amiga) is talking about the desktop with something
based on QNX. Think of Amiga as Dell or Compaq and QNX as Intel
or Pheonix Bios.
>Will someone in this newsgroup please provide a comparrison between
>QNX and Linux as PC type systems. Cost is not a consideration when
>comparing functionality and performance. How complete a system does
>QNX provide? What does QNX require to allow for a complete system?
Desktop applications and a price that won't choke a cow.
There are no apps. Write your own. There are some nice utilities,
but not many. There are no desktop customers. The price for a
full system for one user is pretty nice if you are shopping for a
car, but not if you are shopping for an operating system. In every
respect, Linux wins hands down as a desktop OS. Think of Linux and
Windows as GM and Ford. Think of QNX as an Otis escalator.
Igor Kovalenko
>
> Armin Steinhoff wrote:
>
> AS> A microkernel design for a windowing system makes no sense if the
> AS> underlaying OS is a monolithic one.
> No arguments.
>
> (I see sense, to separate the graphics driver from the main windowing
> magic, such as regions, events...)
>
To provide an example, remember times when video drivers on NT ran in
user space? That was before NT4.0. Then they changed it, now they run in
kernel space and that was the only way how they could provide a
reasonable performance.
- Igor
Darwin Ouyang
Guy Macon <guym...@deltanet.com> wrote:
>Take one engineer, familiar with Windows NT. Sit him down at an old 486
>with a full installation of QNX and Photon. Leave the architecture guide
>nearby - he WILL be interested in a few moments. Say "Go ahead and try
>anything. Doesn't matter what you do because I can reload with a couple
>of keystrokes". Walk away. Go tu cubicle. Wait for amazed person to
>come by, ranting and raving over fast performance, plus a few minor
>"confused by the GUI". problems. Followup with comments about ANSI and
>Posix standards vs. proprietary shifting standards from the M$ Borg.
I have both QNX and NT installed. The University of Waterloo has a site
license for QNX, and I got a copy free.
QNX is cool, but I still like NT better for day to day use.
QNX boots fast, but NT still feels snappier on my system. (I have 192 MB
RAM.) It's probably because the ATI drivers for QNX aren't nearly as
optimized. (And YES I'm using an accelerated QNX driver, not the generic
driver.)
I wouldn't use NT to build a real time control system, but QNX ain't the
best general purpose desktop OS either.
Darwin Ouyang
Igor Kovalenko
well, if we talk about QNX:
- avoid any SysV IPC calls (shared memory, semaphores, messages) to use
POSIX counterparts;
- eliminate calls dealing with supplementary groups;
- if it use Unix domain sockets, switch to INET domain;
- if it use memory mapped files, invent some substitute;
now if you mean NTO, things are better. Only SysV IPC is still big
concern. Other mentioned features are either already available, or will
be. Once I started to work on SysV IPC library for QNX, but then lost
interest. May be I'll do that for NTO.
- Igor
Inge Vabekk
>In article <7b7hua$9...@chronicle.concentric.net>,
>I wouldn't use NT to build a real time control system, but QNX ain't the
>best general purpose desktop OS either.
Really? I would claim that the OS is a lot better than NT when it comes to
speed and efficiency. It may be true that the QNX driver for your video card
is not perfect, but don't blame the whole OS for that. Getting a better card
would probably not cost a lot (But NT supports more cards because more
manufacturers have put their money on Microsoft than on QSSL, and the reason
for that is not that NT is a better OS!!!!).
What you're probably missing is applications, and no one disagrees with you.
There should have been more desktop apps for QNX. And when they're
available, QNX will be much more attractive than the old goliath, both for
desktop and for server use.
--
>>>>> Just try to compress me <<<<<
in...@arxi.nospam (remove spam....)
Bjørnar Bolsøy
Armin Steinhoff <Armin@Steinhoff_de> wrote:
> joec...@tascuj.seatac.net says...
>
> >They will be competing with CE, I assume.
>
> Nonsens ... QNX4 and QNX/Neutrino covers the whole range from embedded
> RT apps to high-performance RT workstations ! Do you really know what CE
is ?
>
> CE has nothing to do with true real-time !
How, exactly, is CE not realtime?
Regards...
Maury Markowitz
> the thing is the subj *should* have been called
> Cheap Amigas and Nto (short for Neutrino) as that is infact the
> base OS that AMIGA inc is useing NOT QNX.
Be that as it may, I'm curious about why they're doing it at all? Yes,
AmigaOS was a breakthrough for it's time, but that was a long time ago.
None of the cool things it could do at the time seem to be particularily
interesting in comparison to newer OS's these days, and it lacks a number
of features from others (OO base (MacOS-X), DB filesystems (Be)) etc. So
why do this at all? Why not just "do" Be? What am I missing?
Maury
Vadim Radionov
Igor Kovalenko wrote:
IK> Vadim Radionov wrote:
>>
>> Armin Steinhoff wrote:
>>
>> AS> A microkernel design for a windowing system makes no sense if
>> AS> the underlaying OS is a monolithic one.
>> No arguments.
>>
>> (I see sense, to separate the graphics driver from the main
>> windowing magic, such as regions, events...)
IK> To provide an example, remember times when video drivers on NT ran in
IK> user space? That was before NT4.0.
IK> Then they changed it, now they run
IK> in kernel space and that was the only way how they could provide
IK> a reasonable performance.
But what should we do to provide us with clustering window sistem?
Vadim
Steve Furr
Daryl Low <dl...@undergrad.math.uwaterloo.ca> wrote:
>dl...@Lehigh.EDU wrote:
>> Why shouldn't I be? And, that being the case, why am I still not overly
>> impressed by the IAT demo disk? Size, while important, is not all.
>
>It's like money, it's not everything but still can be very useful.
>
>Because QNX can fit in small places (and maintain full functionality),
>you can cram it into anything from a palmtop up to big super-servers.
>
>Also, small = less code = less execution overhead = fewer bugs = less
>grief. But at the same time, all the core OS features of NT and other
>*NIXs are there (ie. preemptive multitasking, memory protection,
>etc.....).
And size means a lot more if you are a consumer electronics device
manufacturer looking to add such capabilility to your product.
It adds up to millions of $s in savings that can help make or break
the product.
>
>> Distributed processes are not in the same category as migrated processes.
>> A distributed process only ever runs upon one machine, somewhere in a
>> given network, unless it is migrated. A migrated process is one which has
>> been started on one machine, and has entirely switched operations to one or
>> more other machines.
>
>That would be nice, but alas QNX has not entered the process load
>balancing world yet. I'm not sure if the RT world is looking for that
>right now (due to non-deterministic behaviour of migrating entire
>processes over an interconnect). But I'm sure it would be possible to
>port existing cluster software (Beowulf etc...) to work under QNX.
Yes, memory and CPU being finite resources. But I doubt if anybody
seriously considers migrating processes to be a solid answer to this
problem any more. The distributed community has put a lot of effort
into persistent objects and pass by value (CORBA, Java RMI, Enterprise
Java Beans, etc.). As long as a service (resource, whatever) can save
it state it can be restarted somewhere else and rediscovered. This yields
many benefits including interoperability on heterogeneous
networks and avoids a lot of bugbears that infest process migration,
like how to reacquire all the correct resources when you migrate the
process.
[snip]
--
-------------------------------------------------------------------------
Steve Furr email: fu...@qnx.com
QNX Software Systems, Ltd.
Steve Furr
> AS> underlaying OS is a monolithic one.
>No arguments.
>
The argument for using a microkernel windowing system is effectively
no different from the move from a kernel-based windowing/graphics system
to a client/server one. If your assertion were true, nobody would ever
have given much consideration to News and X. The benefits were pretty obvious
though. As long as suitable IPC is available in the monolithic OS,
it can be worthwhile. The benefit to a microkernel windowing system is
primarily that the separation extends to the graphics drivers. I can
now start and stop the correct drivers on the fly and connect to
an existing desktop session. All you need to do is compare the
convolutions required to dynamically load the correct driver for an X server
to recognize the benefit.
NOS...@isp.invalid
It isn't deterministic. That's why M$ is promising a future version
that is realtime.
Guy Macon
>The argument for using a microkernel windowing system is effectively
>no different from the move from a kernel-based windowing/graphics system
>to a client/server one. If your assertion were true, nobody would ever
>have given much consideration to News and X. The benefits were pretty obvious
>though. As long as suitable IPC is available in the monolithic OS,
>it can be worthwhile. The benefit to a microkernel windowing system is
>primarily that the separation extends to the graphics drivers. I can
>now start and stop the correct drivers on the fly and connect to
>an existing desktop session. All you need to do is compare the
>convolutions required to dynamically load the correct driver for an X server
>to recognize the benefit.
I use QNX, Linux, and NT on systems with no video card or monitor
(no keyboard or mouse either!) This is the environment where QNX
shines. Leave the desktop to desktop operating systems.
D J Hawkey Jr
Just to add to Igor's list...
Linux ports are the worst among the various Unices. Too many odd things
like declarations in weird headers, etc. BSD ports the easiest (be
prepared to convert the "old style" varargs to the "new style").
Generally speaking, the 'autoconf' way of doing things works well about
50% of the time against Watcom 10.6 on QNX. Small tweaks are needed
about 75% of the time on the remaining. The remaining packages are
usually so platform-specific or so broken as not to be worth my time.
Low-level TCP/IP things, like ntalk[d], need considerable tweaking due
to dissimilar sockaddr structures, but is easily worked around after you
recognise the difference.
X apps port nearly intact.
> - Igor
My two-cents' worth,
Dave
--
______________________ ______________________
\__________________ \ D. J. HAWKEY JR. / __________________/
\________________/\ haw...@visi.com /\________________/
http://www.visi.com/~hawkeyd/
A Shelton
I believe microsofts semi-official line is that if you make the computer
fast enough, and the time slice generous enough, then you can guarantee
meeting it...thus a real time system (in their eyes).
There are almost zero reasons why a desktop user should care though, which
is why RTlinux is actually a seperate development stream from normal linux.
A slow system, but which is deterministically so, can still be real time. It
doesn't *neccessarily* mean a real-time system is fast. I have no reason to
believe this of QNX though.
( I also miss the QNX sales guy who used to frequent the newsgroups (RIP))
--
Apparently I'm insane, but I'm one of the happy kinds. (dilbert)
Andrew Shelton ashe...@yallara.cs.rmit.edu.au
GCS(2.1)-d+H+sw+v-C++UL+>L+++E-N++WV--R++tv-b+D++e+fr*y?
Pat Ford
A Shelton <ashe...@cs.rmit.edu.au> wrote in article
<7bincj$s1t$1...@emu.cs.rmit.edu.au>...
| NOS...@isp.invalid writes:
|
snip
|
| ( I also miss the QNX sales guy who used to frequent the newsgroups
(RIP))
|
He is dearly missed here as well
| --
| Apparently I'm insane, but I'm one of the happy kinds. (dilbert)
| Andrew Shelton ashe...@yallara.cs.rmit.edu.au
| GCS(2.1)-d+H+sw+v-C++UL+>L+++E-N++WV--R++tv-b+D++e+fr*y?
|
Mike Davies
<ashe...@cs.rmit.edu.au> writes
>>>
>>> How, exactly, is CE not realtime?
>
>>It isn't deterministic. That's why M$ is promising a future version
>>that is realtime.
>
>I believe microsofts semi-official line is that if you make the computer
>fast enough, and the time slice generous enough, then you can guarantee
>meeting it...thus a real time system (in their eyes).
>
if it were true :-)
Deterministic != fast
Or not necessarily anyway.
Regards,
--
Mike Davies
Tom Sheppard
>The benefit to a microkernel windowing system is
>primarily that the separation extends to the graphics drivers. I can
>now start and stop the correct drivers on the fly and connect to
>an existing desktop session.
Yes, I remember fondly a process control system, designed in the early
'70s, running on an IBM mainframe in the mid '80s that allowed you to
replace the display driver that was displaying the screen that allowed you
to replace the driver. Enter the module to be replaced (the driver), press
<Enter> and blink, the new display appeared. Great fun!
Funny how some s/w today still can't do that 25 years later. I guess the
designers follow the "Del Monte" philosophy.
...Tom
Ric
The ever so slight inconvenience that Gassee and the folk at Menlo
Park just didn't want to know....
Ric
Maury Markowitz
> The ever so slight inconvenience that Gassee and the folk at Menlo
> Park just didn't want to know....
What? Didn't want to know what?
Maury
Ric
Didn't want anything to do with anything Amiga-related, like providing
a new base OS for Amiga Inc. to call Amiga OS. (There's a couple of
Gassee's columns in Be Newsletters of about a year ago (bit more,
probably), should be available on Be's web site.
Matt Sergeant
>
> In <36ddeaf8...@news.demon.co.uk> Ric wrote:
> > The ever so slight inconvenience that Gassee and the folk at Menlo
> > Park just didn't want to know....
>
> What? Didn't want to know what?
Be pulled out of the deal (actually, they were shoved out) at the last
minute (i.e. hours before the WoA announcement). They asked for
extortionate amounts of money that would have made the price of OS5
about the same as BeOS (i.e. around the $100 mark). Amiga Inc weren't
talking about a price like that (i.e. about half that) and so dropped
them like a stone. Be also tried to put all sorts of restrictions and
bits in their favour in the contract that would have made OS5 not what
AI wanted it to be.
I don't know why this has been covered up (re: the Gasse statements to
the contrary). It _is_ the truth.
--
<Matt email="mser...@ndirect.co.uk" />
| Fastnet Software Ltd | Perl in Active Server Pages |
| Perl Consultancy, Web Development | Database Design | XML |
| http://come.to/fastnet | Information Consolidation |
Maury Markowitz
> Didn't want anything to do with anything Amiga-related, like providing
> a new base OS for Amiga Inc. to call Amiga OS.
Ahhhh. Ok, but that's not really what I was asking. The question is why
would anyone want to use AmigaOS when there is, say, BeOS? You appear to
be answering another question, why Amiga Co. didn't "just" go with BeOS.
Seeing as that isn't much clearer, let me rephrase in a way that looks
like flame bait but isn't...
Why, after almost 15 years, should I care about AmigaOS? What does it do
that other OS's out there already don't do? Can it hold a candle to
OpenStep or Be?
Maury