QNX4 Updates System Updated!

Thanks - I didn’t know about this list and I have used ICS before - I’ll call
them regarding QNX. Also you are right - you can’t have it both ways -
design/development is always a balance of trade offs.

KenR

Robert Rutherford wrote:

There are already many hardware vendors that are prepared to commit to QNX
support. For a start, take a look at
http://www.qnx.com/partners/vendors/embedded.html> .

In my experience (for example), ICS Advent are happy to supply rackmount PC
systems that are guaranteed to work with QNX. Of course, these systems are
at least 2-3 times more expensive than your average desktop PC.

IMHO you can’t have it both ways: either you are prepared to pay a premium
to a manufacturer that is prepared to put in the effort to support an
“obscure” OS like QNX, or you take a chance with a $5 NE2000 clone from
Fry’s.

You get what you pay for.

Rob Rutherford

“Ken Recchia” <> rectech@nctimes.net> > wrote in message
news:> 39899261.28E7F6FB@nctimes.net> …
Why don’t you guys start selling your own hardware systems, that work
great with
QNX? You could be like Apple, making a superb joining of OS and
hardware. Sell
everything from PC104 boards to full blown desktops, and then there
would be a
choice to go to for complete compatability. A web store with a build to
order
system would be a great choice. You could still continue with the way
things are,
supporting as many of the varied PC choices out there as you can, as
well. But it
would be great to have a choice that eliminates the guess work from
buying
computer hardware for QNX.

Do it! I wouldn’t buy from anyone else. I don’t get paid much to search
for
compatible adapters and the long term support is no fun.

Don’t be surprised when you find out that the video drivers for Photon aren’t
compatible. ICS/Advent uses an SBC line with Intel video chipsets that aren’t
supported by any QSSL product. If you don’t need graphics, they are very solid,
reliable machines.

—Thom

Robert Rutherford wrote:

There are already many hardware vendors that are prepared to commit to QNX
support. For a start, take a look at
http://www.qnx.com/partners/vendors/embedded.html> .

In my experience (for example), ICS Advent are happy to supply rackmount PC
systems that are guaranteed to work with QNX. Of course, these systems are
at least 2-3 times more expensive than your average desktop PC.

IMHO you can’t have it both ways: either you are prepared to pay a premium
to a manufacturer that is prepared to put in the effort to support an
“obscure” OS like QNX, or you take a chance with a $5 NE2000 clone from
Fry’s.

You get what you pay for.

Rob Rutherford

“Ken Recchia” <> rectech@nctimes.net> > wrote in message
news:> 39899261.28E7F6FB@nctimes.net> …
Why don’t you guys start selling your own hardware systems, that work
great with
QNX? You could be like Apple, making a superb joining of OS and
hardware. Sell
everything from PC104 boards to full blown desktops, and then there
would be a
choice to go to for complete compatability. A web store with a build to
order
system would be a great choice. You could still continue with the way
things are,
supporting as many of the varied PC choices out there as you can, as
well. But it
would be great to have a choice that eliminates the guess work from
buying
computer hardware for QNX.

Do it! I wouldn’t buy from anyone else. I don’t get paid much to search
for
compatible adapters and the long term support is no fun.

Is it just Photon? Will the installer work? Have you run any ncurses programs?
KenR

Thom Skrtich wrote:

Don’t be surprised when you find out that the video drivers for Photon aren’t
compatible. ICS/Advent uses an SBC line with Intel video chipsets that aren’t
supported by any QSSL product. If you don’t need graphics, they are very solid,
reliable machines.

—Thom

Robert Rutherford wrote:

There are already many hardware vendors that are prepared to commit to QNX
support. For a start, take a look at
http://www.qnx.com/partners/vendors/embedded.html> .

In my experience (for example), ICS Advent are happy to supply rackmount PC
systems that are guaranteed to work with QNX. Of course, these systems are
at least 2-3 times more expensive than your average desktop PC.

IMHO you can’t have it both ways: either you are prepared to pay a premium
to a manufacturer that is prepared to put in the effort to support an
“obscure” OS like QNX, or you take a chance with a $5 NE2000 clone from
Fry’s.

You get what you pay for.

Rob Rutherford

“Ken Recchia” <> rectech@nctimes.net> > wrote in message
news:> 39899261.28E7F6FB@nctimes.net> …
Why don’t you guys start selling your own hardware systems, that work
great with
QNX? You could be like Apple, making a superb joining of OS and
hardware. Sell
everything from PC104 boards to full blown desktops, and then there
would be a
choice to go to for complete compatability. A web store with a build to
order
system would be a great choice. You could still continue with the way
things are,
supporting as many of the varied PC choices out there as you can, as
well. But it
would be great to have a choice that eliminates the guess work from
buying
computer hardware for QNX.

Do it! I wouldn’t buy from anyone else. I don’t get paid much to search
for
compatible adapters and the long term support is no fun.

It is just Photon. All of the ICS/Advent SBCs that I use work fine in VGA mode.
Ncurses works fine. I’ve never used the install GUI on one of these units, so I can’t
help you there.

From talking to QSSL, they have made it clear that it just isn’t possible to keep up
with all of the different chipset manufacturers in the market today. I’m sure that
some are easier to work with than others. It may just be that Intel isn’t interested
in developing a relationship with them.

—Thom

Ken Recchia wrote:

Is it just Photon? Will the installer work? Have you run any ncurses programs?
KenR

Thom Skrtich wrote:

Don’t be surprised when you find out that the video drivers for Photon aren’t
compatible. ICS/Advent uses an SBC line with Intel video chipsets that aren’t
supported by any QSSL product. If you don’t need graphics, they are very solid,
reliable machines.

—Thom

Robert Rutherford wrote:

There are already many hardware vendors that are prepared to commit to QNX
support. For a start, take a look at
http://www.qnx.com/partners/vendors/embedded.html> .

In my experience (for example), ICS Advent are happy to supply rackmount PC
systems that are guaranteed to work with QNX. Of course, these systems are
at least 2-3 times more expensive than your average desktop PC.

IMHO you can’t have it both ways: either you are prepared to pay a premium
to a manufacturer that is prepared to put in the effort to support an
“obscure” OS like QNX, or you take a chance with a $5 NE2000 clone from
Fry’s.

You get what you pay for.

Rob Rutherford

“Ken Recchia” <> rectech@nctimes.net> > wrote in message
news:> 39899261.28E7F6FB@nctimes.net> …
Why don’t you guys start selling your own hardware systems, that work
great with
QNX? You could be like Apple, making a superb joining of OS and
hardware. Sell
everything from PC104 boards to full blown desktops, and then there
would be a
choice to go to for complete compatability. A web store with a build to
order
system would be a great choice. You could still continue with the way
things are,
supporting as many of the varied PC choices out there as you can, as
well. But it
would be great to have a choice that eliminates the guess work from
buying
computer hardware for QNX.

Do it! I wouldn’t buy from anyone else. I don’t get paid much to search
for
compatible adapters and the long term support is no fun.

Thom Skrtich wrote:

It is just Photon. All of the ICS/Advent SBCs that I use work fine in VGA mode.
Ncurses works fine. I’ve never used the install GUI on one of these units, so I can’t
help you there.

From talking to QSSL, they have made it clear that it just isn’t possible to keep up
with all of the different chipset manufacturers in the market today. I’m sure that
some are easier to work with than others. It may just be that Intel isn’t interested
in developing a relationship with them.

That’s why QNX needs to develop relationships with these manufactures, and ask them to
build the drivers for QNX for their chipsets themselves. How hard could it be?

—Thom

Ken Recchia wrote:

Is it just Photon? Will the installer work? Have you run any ncurses programs?
KenR

Thom Skrtich wrote:

Don’t be surprised when you find out that the video drivers for Photon aren’t
compatible. ICS/Advent uses an SBC line with Intel video chipsets that aren’t
supported by any QSSL product. If you don’t need graphics, they are very solid,
reliable machines.

—Thom

Robert Rutherford wrote:

There are already many hardware vendors that are prepared to commit to QNX
support. For a start, take a look at
http://www.qnx.com/partners/vendors/embedded.html> .

In my experience (for example), ICS Advent are happy to supply rackmount PC
systems that are guaranteed to work with QNX. Of course, these systems are
at least 2-3 times more expensive than your average desktop PC.

IMHO you can’t have it both ways: either you are prepared to pay a premium
to a manufacturer that is prepared to put in the effort to support an
“obscure” OS like QNX, or you take a chance with a $5 NE2000 clone from
Fry’s.

You get what you pay for.

Rob Rutherford

“Ken Recchia” <> rectech@nctimes.net> > wrote in message
news:> 39899261.28E7F6FB@nctimes.net> …
Why don’t you guys start selling your own hardware systems, that work
great with
QNX? You could be like Apple, making a superb joining of OS and
hardware. Sell
everything from PC104 boards to full blown desktops, and then there
would be a
choice to go to for complete compatability. A web store with a build to
order
system would be a great choice. You could still continue with the way
things are,
supporting as many of the varied PC choices out there as you can, as
well. But it
would be great to have a choice that eliminates the guess work from
buying
computer hardware for QNX.

Do it! I wouldn’t buy from anyone else. I don’t get paid much to search
for
compatible adapters and the long term support is no fun.

Previously, J. Scott Franko wrote in qdn.public.qnx4:

Thom Skrtich wrote:

From talking to QSSL, they have made it clear that it just isn’t possible to keep up
with all of the different chipset manufacturers in the market today. I’m sure that
some are easier to work with than others. It may just be that Intel isn’t interested
in developing a relationship with them.

That’s why QNX needs to develop relationships with these manufactures, and ask them to
build the drivers for QNX for their chipsets themselves. How hard could it be?

You mean that as a joke, right? It’s pretty hard to convince other people
to write software for you for free, they do it for microsoft, but microsoft
has a slightly larger userbase than QNX.

How many vendors build drivers for Macs? Apple gets away with it because they
build and spec their own hardware, and charge a premium for the service, and
don’t let other companies build clones, thus eroding their profits.

Sam


Sam Roberts (sam@cogent.ca), Cogent Real-Time Systems (www.cogent.ca)

Of course, no one gets the support that WinXX gets, but many companies write Macintosh drivers
for their products. ATI, and 3dfx come to mind in graphics cards. Many other devices from
other hardware vendors have Macintosh drivers.

Comtrol is majorly a WinXX serial card (rocketport), but they have QNX drivers.

If it weren’t for the stupid WinXX monopoly, these hardware vendors wouldn’t survive writing
their drivers for only one platform. If they can’t write drivers for multiple platforms, then
then they at least need to put their driver code in open source so it can be ported, like
nVidia does.

Yeah, I’m joking, but I’m seriously wishing too.

Scott

Sam Roberts wrote:

Previously, J. Scott Franko wrote in qdn.public.qnx4:


Thom Skrtich wrote:

From talking to QSSL, they have made it clear that it just isn’t possible to keep up
with all of the different chipset manufacturers in the market today. I’m sure that
some are easier to work with than others. It may just be that Intel isn’t interested
in developing a relationship with them.

That’s why QNX needs to develop relationships with these manufactures, and ask them to
build the drivers for QNX for their chipsets themselves. How hard could it be?

You mean that as a joke, right? It’s pretty hard to convince other people
to write software for you for free, they do it for microsoft, but microsoft
has a slightly larger userbase than QNX.

How many vendors build drivers for Macs? Apple gets away with it because they
build and spec their own hardware, and charge a premium for the service, and
don’t let other companies build clones, thus eroding their profits.

Sam


Sam Roberts (> sam@cogent.ca> ), Cogent Real-Time Systems (> www.cogent.ca> )

J. Scott Franko <jsfranko@switch.com> wrote:

Yeah, I’m joking, but I’m seriously wishing too.

Anyone who attended the QNX 2000 conference may have had a
chance to see Epson demonstrate some of their embedded video
devices running drivers that were developed entirely by
Epson.

It’s taken us a very long time to change our graphics
driver architecture to make it possible for third parties
to write graphics drivers.

Everyone put on some ruby red slippers and start clicking
your heels together…

J. Scott Franko <jsfranko@switch.com> wrot

then they at least need to put their driver code in open source
so it can be ported, like nVidia does.

The nVidia source gives us an easy way to port the feature set
the have exported.

Otherwise, their chipset is a black hole, and we cannot make
it do video overlay or 3D, nor fix half the problems anyone is
likely to report.

On the other hand, 3DFX released the full register set docs for
their chips, and three of us in the video group had a race to
see who could get a driver going. In about four hours, we had
three drivers going, and then it was a race to see whose driver
was fastest by the end of the day.

Source code is nice in theory, but complete disclosure under
NDA gives you the consumer the best product in the shortest
period of time.

eric@qnx.com (Eric Johnson) writes:

In Debbie’s message she mentioned that the old-style updates will be
available at_least until Jan 2001. i.e. we had not conclusively
decided to shut down the old-style updates for everyone on that date.

I’ve conferred with both Debbie and our product release group and we
have agreed that we will continue to create the old-style updates
for QNX 4 products when new versions are released, and will continue to
make them available to those existing customers who have a legitimate
need for them.

That’s a really bad answer, actually. Now you are maintaining two
different types of archives. One is new and “supported”, and the
other is available only to people who have identified themselves as
“really needing them” !? How long will it take until the old style
archives are unavailable for download, or one or two versions older
than the new style archives? There is a Really Simple Solution here.
Just create a command-line installer for the new style licenses. Give
it a non-interactive mode, like we currently have with “install -u”.
Now QNX only has to maintain one type of archive, and those who need
to install remotely or automatically can do so. Everybody wins. The
decision to maintain the old style archives is just making work for
QNX, and reducing the chance that the users will be happy with the
result.

Cheers,
Andrew


Andrew Thomas, President, Cogent Real-Time Systems Inc.
2430 Meadowpine Boulevard, Suite 105, Mississauga, Ontario, Canada L5N 6S2
Email: andrew@cogent.ca WWW: http://www.cogent.ca

pete@qnx.com wrote:

Source code is nice in theory, but complete disclosure under
NDA gives you the consumer the best product in the shortest
period of time.

I’ve heard people state the opposite. Even under NDA some vendors are
stingy with their information (ie complete disclosure is nice in theory,
but rarely happens), but with the source code, you can find it out for
yourself, by example.

But you’ve proven my point! In 4 hours, with the information you
needed, you guys had a fun little contest to see who could write the
best driver. Now if you guys can do it in 4 with information supplied
to you by the vendor, just think how fast a basic driver could be put
together by the people who are giving the information!

Example: 3dfx had a good model for writing a mac driver. They through
together some beta drivers (which I’m sure someone through together in 4
hours like you guys), put them out for Mac users to get, the Mac users
by PCI cards made for the PC, use the beta drivers, critique the
drivers, the vendor improves the drivers based on the feedback, and now
are releases actual voodoo5 cards and drivers made specifically for the
Mac!

That’s why I’m wishing and plodding for vendor written drivers! Not
that your drivers don’t work great with the boards you support, but for
all the cards you can’t get to, I think manufactures should step up and
provide drivers! If QNX is going to run on the ever changing PC
platform, its a fact of life that the cards are going to improve and
change in what is now an almost 6 month cycle!

It would be interesting to know what percentage of sales of PC’s are
driven by QNX. Is it so small that vendors can’t even think about it?

If you can't write it (because there are too many vendors, with too many revisions), and vendor's won't write it (cause the market is too small and 4 hours of their time is too much), and the boards change every 6 months, meaning you can't find the boards that have already had drivers written for them, in a new PC, and QSSL won't sell it's own PC's with OEM cards from vendors you have relations with, then its a crap shoot to buy new PC hardware that will work with QNX!

What’s the solution? Is world domination the only way that you can get
support for this fractured vendor card market?

So a parts list of products that you support just doesn’t work, unless
one decides one has to get into the business of building ones own PC’s
to build products upon. It’s frustrating!

Forgive me for venting my frustration on the group, but I’m sure many of
you have been there. Please, someone figure out how to support rapidly
improving vendor hardware across multiple architectures!

Scott

Its not only QNX and Windows that run on PCs. Thus as long as the PC can
boot and run a non-Windows OS (say Linux) then chances are it will work
with QNX code … I have rarely seen a box that can’t boot QNX4 and
Neutrino, and the only one I remember was from SGI which came to
existence solely to run NT and its BIOS reflected this fact.

-Ahmed

J. Scott Franko <jsfranko@switch.com> wrote:

: pete@qnx.com wrote:

:> Source code is nice in theory, but complete disclosure under
:> NDA gives you the consumer the best product in the shortest
:> period of time.

: I’ve heard people state the opposite. Even under NDA some vendors are
: stingy with their information (ie complete disclosure is nice in theory,
: but rarely happens), but with the source code, you can find it out for
: yourself, by example.

: But you’ve proven my point! In 4 hours, with the information you
: needed, you guys had a fun little contest to see who could write the
: best driver. Now if you guys can do it in 4 with information supplied
: to you by the vendor, just think how fast a basic driver could be put
: together by the people who are giving the information!

: Example: 3dfx had a good model for writing a mac driver. They through
: together some beta drivers (which I’m sure someone through together in 4
: hours like you guys), put them out for Mac users to get, the Mac users
: by PCI cards made for the PC, use the beta drivers, critique the
: drivers, the vendor improves the drivers based on the feedback, and now
: are releases actual voodoo5 cards and drivers made specifically for the
: Mac!

: That’s why I’m wishing and plodding for vendor written drivers! Not
: that your drivers don’t work great with the boards you support, but for
: all the cards you can’t get to, I think manufactures should step up and
: provide drivers! If QNX is going to run on the ever changing PC
: platform, its a fact of life that the cards are going to improve and
: change in what is now an almost 6 month cycle!

: It would be interesting to know what percentage of sales of PC’s are
: driven by QNX. Is it so small that vendors can’t even think about it?

:
: If you can’t write it (because there are too many vendors, with too many
: revisions), and vendor’s won’t write it (cause the market is too small
: and 4 hours of their time is too much), and the boards change every 6
: months, meaning you can’t find the boards that have already had drivers
: written for them, in a new PC, and QSSL won’t sell it’s own PC’s with
: OEM cards from vendors you have relations with, then its a crap shoot to
: buy new PC hardware that will work with QNX!

: What’s the solution? Is world domination the only way that you can get
: support for this fractured vendor card market?

: So a parts list of products that you support just doesn’t work, unless
: one decides one has to get into the business of building ones own PC’s
: to build products upon. It’s frustrating!
:

: Forgive me for venting my frustration on the group, but I’m sure many of
: you have been there. Please, someone figure out how to support rapidly
: improving vendor hardware across multiple architectures!

: Scott

\

Ahmed Abed
QNX Software Systems Ltd.
(613) 591 0836 x139
ahmed@qnx.com

J. Scott Franko <jsfranko@switch.com> wrote:


pete@qnx.com > wrote:

Source code is nice in theory, but complete disclosure under
NDA gives you the consumer the best product in the shortest
period of time.

I’ve heard people state the opposite. Even under NDA some vendors are
stingy with their information (ie complete disclosure is nice in theory,
but rarely happens), but with the source code, you can find it out for
yourself, by example.

When we first took a look at the open' nVidia source code, we found that it was intentionally obfuscated. Many of the vital register addresses and register bits were globs of indecipherable numbers. I believe it was run through a C’ pre-processor before
release.

Open source like that I don’t need.

But you’ve proven my point! In 4 hours, with the information you
needed, you guys had a fun little contest to see who could write the
best driver. Now if you guys can do it in 4 with information supplied
to you by the vendor, just think how fast a basic driver could be put
together by the people who are giving the information!

Ha!

I bet it would take them a lot longer.

We’ve all seen many different video chipsets, but we also know
the QNX OS intimately.

There are Linux programmers at many of the video chip companies,
but that really just means they are a whole lot braver than they
would be if they didn’t think they knew how to program a device
driver in `UNIX’.

The similarities between other OS’s and any QNX OS becomes
greater the closer you get to writing things like device
drivers. A Photon driver is very unlike an `X’ driver or
a windows driver.

That’s why I’m wishing and plodding for vendor written drivers! Not
that your drivers don’t work great with the boards you support, but for
all the cards you can’t get to, I think manufactures should step up and
provide drivers! If QNX is going to run on the ever changing PC
platform, its a fact of life that the cards are going to improve and
change in what is now an almost 6 month cycle!

I agree with your forecast of changing weather in the hardware
market, but be careful what your wish for. Would you trust
every video chip vendor you know of to be able to deal with some
of the subtleties of a real time system (i.e. like learning not
to disable interrupts, and run at a high priority all the time).

How many of your Windows' problems are caused by a sea of device manufacturers cranking out crappy drivers that step on other drivers? I know that I have _often_ updated sound or video drivers and had that miraculously cure some kind of Windows’ problem or another.

stepping on soapbox
If you can’t write it (because there are too many vendors, with too many
revisions),

OK… so by this statement, you seem to imply that we should support
every revision of every vendors chip.

No. We will not support some of these chipsets that are so
cruddy that they take as much time to support as four or five
other chips.

and vendor’s won’t write it (cause the market is too small
and 4 hours of their time is too much),

Do you really want some guy at ATI who thinks he can make
a Photon driver because he can make a Linux X driver to put
out some (probably) unsupported driver he did one weekend in four
hours?

You may be mistaking the four hour 3DFX driver we did the first
day for the one that’s released now. It’s got far more than four
hours in it.

Frankly, I would rather have some QNX fan use the DDK to do one
by looking at some Linux source and actually thinking about it.

and the boards change every 6
months, meaning you can’t find the boards that have already had drivers
written for them,

Until the Voodoo 5.0 came out recently, our 3DFX driver worked with
pretty much every 3DFX card made in over two years.

We haven’t really had a large gap in our ATI card support for a very
long time.

I’m not saying the video card market isn’t volatile, but you’re
exaggerating.

in a new PC, and QSSL won’t sell it’s own PC’s with
OEM cards from vendors you have relations with, then its a crap shoot to
buy new PC hardware that will work with QNX!

Well it’s a crap shoot to buy it from certain large vendors who
cater to the Windows market and won’t let you choose which parts
go into it, I’ll grant you that.

What’s the solution? Is world domination the only way that you can get
support for this fractured vendor card market?

Not at all. If you want support for a particular card, you need only
make a business case to the appropriate sales contact at QNX.

So a parts list of products that you support just doesn’t work, unless
one decides one has to get into the business of building ones own PC’s
to build products upon. It’s frustrating!

We go through a lot of computers, and though we don’t build our own,
we get a local company to build them for us with parts we pre approve.

By `build’, I mean they hire monkeys to screw motherboards into cases.
If you can’t find a company that will do that for you where you live,
I’d be surprised.

Forgive me for venting my frustration on the group, but I’m sure many of
you have been there. Please, someone figure out how to support rapidly
improving vendor hardware across multiple architectures!

I have a great idea for doing exactly that. If all the OS vendors would band together and _not_ support hardware that information is not available for, then eventually the bad hardware will go away leaving only the good stuff.

In article <x74s4ue846.fsf@cogent.ca>, Andrew Thomas <andrew@cogent.ca> wrote:

eric@qnx.com > (Eric Johnson) writes:
In Debbie’s message she mentioned that the old-style updates will be
available at_least until Jan 2001. i.e. we had not conclusively
decided to shut down the old-style updates for everyone on that date.

I’ve conferred with both Debbie and our product release group and we
have agreed that we will continue to create the old-style updates
for QNX 4 products when new versions are released, and will continue to
make them available to those existing customers who have a legitimate
need for them.

That’s a really bad answer, actually. Now you are maintaining two
different types of archives. One is new and “supported”, and the
other is available only to people who have identified themselves as
“really needing them” !? How long will it take until the old style
archives are unavailable for download, or one or two versions older
than the new style archives? There is a Really Simple Solution here.
Just create a command-line installer for the new style licenses. Give
it a non-interactive mode, like we currently have with “install -u”.
Now QNX only has to maintain one type of archive, and those who need
to install remotely or automatically can do so. Everybody wins. The
decision to maintain the old style archives is just making work for
QNX, and reducing the chance that the users will be happy with the
result.

We have an established and reliable process for creating the existing
style of archives, and we don’t have a command line installer for
the new style archives. Customers also have a reliable tool -
/etc/install - to install the old archives, and by continuing to
support this mechanism we aren’t forcing anyone to change they
way they do things. The possibility of creating a command-line
installer for the new packages is something that we’ve kept in mind
over the years but it just isn’t near the top of the priority heap
right now. Sorry.


Eric Johnson
QA Mgr, QNX Software Systems Ltd.

pete@qnx.com wrote:

J. Scott Franko <> jsfranko@switch.com> > wrote:

pete@qnx.com > wrote:

Source code is nice in theory, but complete disclosure under
NDA gives you the consumer the best product in the shortest
period of time.

I’ve heard people state the opposite. Even under NDA some vendors are
stingy with their information (ie complete disclosure is nice in theory,
but rarely happens), but with the source code, you can find it out for
yourself, by example.

When we first took a look at the open' nVidia source code, we found that it was intentionally obfuscated. Many of the vital register addresses and register bits were globs of indecipherable numbers. I believe it was run through a C’ pre-processor before
release.

Open source like that I don’t need.

obviously. I didn’t know they obfusc’d the code. How stupid.

But you’ve proven my point! In 4 hours, with the information you
needed, you guys had a fun little contest to see who could write the
best driver. Now if you guys can do it in 4 with information supplied
to you by the vendor, just think how fast a basic driver could be put
together by the people who are giving the information!

Ha!

I bet it would take them a lot longer.

We’ve all seen many different video chipsets, but we also know
the QNX OS intimately.

And a board vendor knows their board intimately. You both have to learn each
others tech data to do the job. Are you saying that its easier to learn the
tech data of a graphics board, than it is to learn how to write a video
driver for QNX?

There are Linux programmers at many of the video chip companies,
but that really just means they are a whole lot braver than they
would be if they didn’t think they knew how to program a device
driver in `UNIX’.

I’m not sure I understand what you mean here. Are you saying that linux
programmers are only junior unix program, thinking they are working in a real
Unix, but not really?


The similarities between other OS’s and any QNX OS becomes
greater the closer you get to writing things like device
drivers. A Photon driver is very unlike an `X’ driver or
a windows driver.

If the similarities are greater at the device driver level, then why isn’t a
photon driver like an X driver or a windows driver?

That’s why I’m wishing and plodding for vendor written drivers! Not
that your drivers don’t work great with the boards you support, but for
all the cards you can’t get to, I think manufactures should step up and
provide drivers! If QNX is going to run on the ever changing PC
platform, its a fact of life that the cards are going to improve and
change in what is now an almost 6 month cycle!

I agree with your forecast of changing weather in the hardware
market, but be careful what your wish for. Would you trust
every video chip vendor you know of to be able to deal with some
of the subtleties of a real time system (i.e. like learning not
to disable interrupts, and run at a high priority all the time).

Perhaps not, but you can help them keep it straight. Give them enough
technical info and they’ll write it so that it doesn’t conflict.

How many of your Windows' problems are caused by a sea of device manufacturers cranking out crappy drivers that step on other drivers? I know that I have _often_ updated sound or video drivers and had that miraculously cure some kind of Windows’ problem or another.

How many problems on any OS are caused by people of varying levels of skills
cranking out crappy code. Its a fact of life. I can’t tell you the number of
times these supposed memory protected systems I run have crashed on me for no
reason, even on QNX, where the driver code is controlled so tightly.

stepping on soapbox
If you can’t write it (because there are too many vendors, with too many
revisions),

OK… so by this statement, you seem to imply that we should support
every revision of every vendors chip.

I’m not implying anything, just reitterating QSSL’s position.

No. We will not support some of these chipsets that are so
cruddy that they take as much time to support as four or five
other chips.

I’m not asking you to support the cruddy stuff. I’m just asking that there be
some support for the major cards that come out. I just want to be able to buy
a major brand computer and have it support QNX. Is that too much to ask?

and vendor’s won’t write it (cause the market is too small
and 4 hours of their time is too much),

Do you really want some guy at ATI who thinks he can make
a Photon driver because he can make a Linux X driver to put
out some (probably) unsupported driver he did one weekend in four
hours?

No, I want some guy at ATI to add QNX driver writing to his repetoire, and
become good at it, and write a driver to support Photon. What’s so elite
about QNX and it’s drivers that a guy who already writes drivers for windows
and unix and linux, couldn’t study for a while and write good drivers for QNX?

You may be mistaking the four hour 3DFX driver we did the first
day for the one that’s released now. It’s got far more than four
hours in it.

I don’t mistake it. I see the 4 hour deal as a fun hack just to prove you
could get it working. I think other good driver writers could do the same.
But of course you aren’t going to release a 4 hour hack. You clean it up,
alpha it, beta it, get feedback, and polish it into a release. This takes
time, but I don’t think it should take 6 months.

Frankly, I would rather have some QNX fan use the DDK to do one
by looking at some Linux source and actually thinking about it.

And why can’t you recruit and create fans of QNX at the vendors?

and the boards change every 6
months, meaning you can’t find the boards that have already had drivers
written for them,

Until the Voodoo 5.0 came out recently, our 3DFX driver worked with
pretty much every 3DFX card made in over two years.

We haven’t really had a large gap in our ATI card support for a very
long time.

I’m not saying the video card market isn’t volatile, but you’re
exaggerating.

Perhaps, but not by much.

in a new PC, and QSSL won’t sell it’s own PC’s with
OEM cards from vendors you have relations with, then its a crap shoot to
buy new PC hardware that will work with QNX!

Well it’s a crap shoot to buy it from certain large vendors who
cater to the Windows market and won’t let you choose which parts
go into it, I’ll grant you that.

But why should it be? If the large vendors are the most available computers
to buy, shouldn’t you be supporting their configurations with QNX? I guess
the problem lies in how scalable QNX is. And QNX seems to be aiming for the
embedded market now more than it is aiming for general purpose realtime
computing on larger server type computers. When you work with the embedded
products, by its very nature you are buying parts. But with general purpose
realtime servers that run large scale operations, you need to buy systems.

What’s the solution? Is world domination the only way that you can get
support for this fractured vendor card market?

Not at all. If you want support for a particular card, you need only
make a business case to the appropriate sales contact at QNX.

I can’t make a business case. We’re a small fish doing railyards here and
there. We do a primary and a redundant realtime server in QNX, and the rest
of the machines are NT display clients. We have this silly requirement to buy
all the same computers so any machine can be cannabilized into a replacement
for any other.

All I really wanted when I started this thread was the recomendation of a
system that I could by that had the parts in it that QNX supports. I have
more complicated realtime problems to think about. I don’t want to spend all
my life finding a product that works. I want one that “just works!” I think
I got one reco along those lines, out of all the many messages in this thread,
and even that was qualified with the recomedation that I buy it empty and get
the video and net cards separately.

So a parts list of products that you support just doesn’t work, unless
one decides one has to get into the business of building ones own PC’s
to build products upon. It’s frustrating!

We go through a lot of computers, and though we don’t build our own,
we get a local company to build them for us with parts we pre approve.

By `build’, I mean they hire monkeys to screw motherboards into cases.
If you can’t find a company that will do that for you where you live,
I’d be surprised


Forgive me for venting my frustration on the group, but I’m sure many of
you have been there. Please, someone figure out how to support rapidly
improving vendor hardware across multiple architectures!

soapbox
I have a great idea for doing exactly that. If all the OS vendors would
band together and not support hardware that information is not
available for, then eventually the bad hardware will go away leaving
only the good stuff.
soapbox off

Yeah, we know that will never happen.

Thanks for the feedback! I’m pretty much know where I stand now, with respect
to choosing a system for QNX, even if I didn’t get many real system names to
choose from.

J. Scott Franko <jsfranko@switch.com> wrote:


pete@qnx.com > wrote:

I bet it would take them a lot longer.

We’ve all seen many different video chipsets, but we also know
the QNX OS intimately.


And a board vendor knows their board intimately. You both have to learn each
others tech data to do the job. Are you saying that its easier to learn the
tech data of a graphics board, than it is to learn how to write a video
driver for QNX?

That’s my opinion.

Graphics chips are really all very similar. In the driver group here,
we have seen dozens of different chipsets, and we know how to debug
a driver while it’s undergoing development, so our only real learning
curve is the chip itself.

A chip vendor on the other hand would have to learn everything from
how to install our OS, how to put it on a network, how to use our
compilers, how to write QNX interrupt handlers, how to get physical
memory, etc. etc.

Admittedly, once they have done one or two QNX drivers, they could
probably do a driver for a new chip as fast as or faster than us,
but right now I think we could definitely do QNX drivers faster
than any of them assuming we have access to the same information

There are Linux programmers at many of the video chip companies,
but that really just means they are a whole lot braver than they
would be if they didn’t think they knew how to program a device
driver in `UNIX’.

I’m not sure I understand what you mean here. Are you saying that linux
programmers are only junior unix program, thinking they are working in a real
Unix, but not really?

No… I’m saying that a Linux programmer who has done a video driver
for Linux has written an XWindows driver… not a Photon driver, and
that if they assume Photon is just our name for XWindows, they
are going to be completely lost.

The similarities between other OS’s and any QNX OS becomes
greater the closer you get to writing things like device
drivers. A Photon driver is very unlike an `X’ driver or
a windows driver.

If the similarities are greater at the device driver level, then why isn’t a
photon driver like an X driver or a windows driver?

Duh…

I meant the differences are greater, or that the similarities are
`lesser’


I agree with your forecast of changing weather in the hardware
market, but be careful what your wish for. Would you trust
every video chip vendor you know of to be able to deal with some
of the subtleties of a real time system (i.e. like learning not
to disable interrupts, and run at a high priority all the time).

Perhaps not, but you can help them keep it straight. Give them enough
technical info and they’ll write it so that it doesn’t conflict.

That’s what we’re trying to do, but I know that a couple of years
ago, there was a big flap in the Windows world because certain
companies disabled interrupts in their drivers in order to get
higher benchmark numbers.

We can give them what they need and ask them to do things nicely' but there is no way to stop them from doing bad things’.

No. We will not support some of these chipsets that are so
cruddy that they take as much time to support as four or five
other chips.


I’m not asking you to support the cruddy stuff. I’m just asking that there be
some support for the major cards that come out. I just want to be able to buy
a major brand computer and have it support QNX. Is that too much to ask?

It depends on your definition of major brand computer'. In my experience, many major brand computers’ do not do things the way everyone else does
them, and they do not release any information on how to make the machines
work in a non-Windows OS.

Basically, there are a lot of ``Windows’’ boxes out there that are
designed only to run Windows.

and vendor’s won’t write it (cause the market is too small
and 4 hours of their time is too much),

Do you really want some guy at ATI who thinks he can make
a Photon driver because he can make a Linux X driver to put
out some (probably) unsupported driver he did one weekend in four
hours?


No, I want some guy at ATI to add QNX driver writing to his repetoire, and
become good at it, and write a driver to support Photon. What’s so elite
about QNX and it’s drivers that a guy who already writes drivers for windows
and unix and linux, couldn’t study for a while and write good drivers for QNX?

There is nothing elite about it, but you’re no longer talking about four hours
of their time.

You can’t make a comment citing four hours being too much, when what you
really want is far more than four hours effort from them.

And just in case it got lost in the middle of these posts, I did drop
a broad hint that you may be seeing vendor written drivers
down the road. We do have a driver development kit now, and
we have given it to some of the vendors.

You may be mistaking the four hour 3DFX driver we did the first
day for the one that’s released now. It’s got far more than four
hours in it.

I don’t mistake it. I see the 4 hour deal as a fun hack just to prove you
could get it working. I think other good driver writers could do the same.
But of course you aren’t going to release a 4 hour hack. You clean it up,
alpha it, beta it, get feedback, and polish it into a release. This takes
time, but I don’t think it should take 6 months.

Agreed. I agree with most of what you’re saying, but I’ve learned
to pick out exaggerations and hyperbole since I’m so good
at generating it myself :slight_smile:

Frankly, I would rather have some QNX fan use the DDK to do one
by looking at some Linux source and actually thinking about it.

And why can’t you recruit and create fans of QNX at the vendors?

We can and we do. It takes time. As I said in an earlier post, Epson
demonstrated their own driver at the QNX 2000 conference. They aren’t
the only ones, but it’s not my place to tell you who the others
are… they’re still a bt shy.

in a new PC, and QSSL won’t sell it’s own PC’s with
OEM cards from vendors you have relations with, then its a crap shoot to
buy new PC hardware that will work with QNX!

Well it’s a crap shoot to buy it from certain large vendors who
cater to the Windows market and won’t let you choose which parts
go into it, I’ll grant you that.

But why should it be?

It shouldn’t be, but it is not us who is making the non standard
hardware.

If the large vendors are the most available computers
to buy, shouldn’t you be supporting their configurations with QNX?

Well I’m enjoying this conversation immensely, but I really don’t
have time to write an essay on many of these things.

I’ll just say that we do our best to keep up with things, and if
you ever wonder why we don’t support something that seems real
popular to you, it’s probably because we cannot get any documentation
on it.

Not at all. If you want support for a particular card, you need only
make a business case to the appropriate sales contact at QNX.

I can’t make a business case. We’re a small fish doing railyards here and
there. We do a primary and a redundant realtime server in QNX, and the rest
of the machines are NT display clients. We have this silly requirement to buy
all the same computers so any machine can be cannabilized into a replacement
for any other.

A business case isn’t always how many licenses you buy or wether
you can pay for a custom engineered driver. If you tell your sales
contact you would like to use a Brand X such and such, and lots
of other people do the same, there could be an aggregate business
case there.

QUICS isn’t really a good place to say stuff that gets heard by
the sales and business people though.

All I really wanted when I started this thread was the recomendation of a
system that I could by that had the parts in it that QNX supports. I have
more complicated realtime problems to think about. I don’t want to spend all
my life finding a product that works. I want one that “just works!” I think
I got one reco along those lines, out of all the many messages in this thread,
and even that was qualified with the recomedation that I buy it empty and get
the video and net cards separately.

I sympathize with you there, but if you look at if from my side, I’ve
been burned many times by recommending a card, and then finding out
that the vendor is using a different chip on it now. Their drivers
work because they shipped them with the card, but noone told me they
were going to pull a switcheroo.

Laptops are particularly bad for that kind of thing.

The other side of that coin is that I have to go to the chipset vendors
and try to get information from them to support their chips, so I have
to be very careful recommending one chip over another. That’s another
reason we’re a little vague here when it comes to hardware
recommendations.

Thanks for the feedback! I’m pretty much know where I stand now, with respect
to choosing a system for QNX, even if I didn’t get many real system names to
choose from.

No problem. Good luck finding suitable hardware, and if you do find
something that almost works, don’t hesitate to ask about it. We
may be able to come up with a quick patch that lets you move
forward.

pete@qnx.com wrote:

I sympathize with you there, but if you look at if from my side, I’ve
been burned many times by recommending a card, and then finding out
that the vendor is using a different chip on it now. Their drivers
work because they shipped them with the card, but noone told me they
were going to pull a switcheroo.

You know, I seem to remember that being what happend with the 3com 10/100 card now
that you mention it. The 3c905 and 3c905 B users a version I and version II of a
chip, but they used a completely different chip in the 3c905c. I think reports
finally ironed out that this was the only version of the board that worked well with
QNX.

We had our biggest problems with that card, and unfortuately, spent a bunch of money
on test cards before we finally settled on the netgear fa310tx. The info that the
3c905c was working, came out a bit later.

Pete, thanks for taking the time to discuss this with me. It’s really been fun to
have access to you guys like this. That’s one of the things I really like about
QSSL. I have a much better understanding of the difficulties you guys have in
trying to support or even recommend certain systems.

I have a love/hate relationship with the QNX OS right now, but maybe one day my
opinions will be all good! ;0)

Scott

J. Scott Franko <jsfranko@switch.com> wrote:

Pete, thanks for taking the time to discuss this with me. It’s really been fun to
have access to you guys like this. That’s one of the things I really like about
QSSL. I have a much better understanding of the difficulties you guys have in
trying to support or even recommend certain systems.

Thanks, and please understand that we do recognize your concerns
and are doing what we can to address them.

It’s a sore point for both of us… you wish I could recommend a
specific system for you, and so do I, but I’m just not in a position
to do anything more than recommend what components are in it.

Broader hardware support is always high on the list, but so is
better support for more hardware features, and unfortunately the
two things compete with each other for resources.

It’s our hope (and yours) that by publishing driver development
kits, we can spend more of our resources on implementing the
`deep’ feature support, and have a means of empowering those who
can’t neccessarily afford a custom driver solution to roll their
own, and maybe even donate it back to the QNX community.

In qdn.public.qnx4 James Gober <jgober@goisi.com> wrote:

It could be the floppy drive but it doesn’t seem too likely considering:

The drive is brand new (a Mitsumi D359M3).
The QNX floppy disk was not created on this drive but on another QNX system.
The floppy boots successfully on the original system.
The QNX boot loader is successfully being read from the floppy.
DOS will successfully boot from a floppy in this drive.

It is a relatively painless test to get another floppy drive and try it, but
I have my doubts…

I have my doubts too, but I’m not sure you know what he’s saying…

The drive creating the floppies could be off track, and if it was,
it would still be able to read any disks it creates. Only other
drives would have trouble reading it’s disks, wether they were
brand spanking new or not.

I can’t tell if this is appropriate in your case or not, since
it’s not clear what The drive' this drive’, another QNX system', the original system’ etc. all mean.

I think you are saying the floppy drive in your target system
is brand new, and that a boot floppy you create on your development
system boots OK in your development system, but not on your target
system.

I think you are also saying that your target system boots a DOS floppy,
but I don’t know if the DOS floppy was created on your development
system, or even if it was, wether it may have been created a year or
more ago before your development systems floppy drive might have
drifted out of spec.

As far as the QNX boot loader being successfully loaded, I don’t
remember if you said it would print dots forever or if it printed
a whole bunch and then stopped dead.

If it’s the former, then although the boot loader is being
successfully read, the image it needs to load is not.

If the latter, then your boot image may be too big for the BIOS to
load it, or something you are doing early in your image is locking
the target system up early on (but not your devl system).

Look at the size of the .boot file on the floppy. If it is around
580+ K, you may be getting into the size that will give you
problems.

Before you ask me what the exact size of boot image that is safe
is, I don’t know. It varies from computer to computer. If you
find out, please let us know.

If your image size is getting into that range, you could try
removing something large temporarily to see if it resolves
your booting problem. If it does, then you’ll have to look
at trimming the size of your boot image down, or fiddling
with any CMOS options your target system has that might
allow an increase in the boot image size.

If your image size is small, and you’re getting the dots then
a freeze, then the OS is coming up, but getting stuck on
something early on.

In that case, you’ll have to figure out what’s going on by
either taking stuff out of your build image until you find
the one that’s screwing the mix, or start with more or less
nothing, and add stuff to it until it breaks.

Previously, Per Akesson wrote in qdn.public.qnx4, comp.os.qnx:

When I run sysmon in a pterm I don’t get the
correct “graphic” characters as on a console,
even if I set the TERM to qnx.

But strange thing, if I ditto from a pterm into the console
the sysmon output looks OK.

Could someone explain this ?

My guess would be that sysmon doesn’t use the correct way of
displaying linedrawing characters.

(and perhaps give me a work-around)

Run your pterm in QNX mode instead of ANSI mode.