Still no kmod for new nvidia
James McKenzie
jjmckenzie51 at earthlink.net
Thu Jul 29 03:37:04 UTC 2010
Rick Stevens wrote
> It would also be nice if the manufacturers would release better
> information on the devices themselves.
Yes, this has been hashed, rehashed and burnt to the ground.
Manufacturers don't want to give away technical secrets and will go to
great lengths to keep them that way. Their bottom lines depend on it.
Yes, AMD/ATI/nVidia/Intel are in business to make money. In order to
make money you have to have a feature your competition does not have or
you have to be super cheap.
Several independent developers were sued by a now defunct manufacturer
for reverse 'black box' engineering their chipset just to build drivers
that the manufacturer PUBLICLY stated they would not build.
> Much of the time spent in FOSS driver development is reverse engineering drivers for Windows because
> of a lack of adequate information--and sometimes bits get left out or don't work well because of it.
>
>
Describe 'reverse engineer'. There is looking at the input/outputs of a
module with no knowledge of the code of the module (called 'black' box)
and then there is decoding to the machine code level ('white' box or
'gray' box.) There is a reason that Tandy versus IBM stuck. Tandy
watched what happened when an IBM PC booted on power up with a DOS disk
in the machine (DOS remember was stolen and had been around as source
code) and found that the Operating System queried a specific location in
memory for the letters IBM. They did that and made a better BIOS. The
company is now known as Phoenix. Other companies did a line-by-line
decode of the BIOS and that got them in a bunch of trouble. That is why
'black box' is acceptable, but 'white' or 'gray' box is not.
> A few years back, Texas Instruments created a wireless NIC that a number
> of PCMCIA and USB dongles used (mostly from D-Link). We ended up with
> hack very similar to nVidia kmods that'd use a binary blob from Windows
> for the guts of the driver.
>
Again, black boxing the 'blob' is ok. Reverse engineering the code is
not.
Now for the problem with RPMFusion. If they ain't on the stick, someone
gotta take over. That means if we can get the 'propriatary' bits and
then black box them to see what happens and then write our own drivers.
ATI released a 'blob' for a very famous, but almost defunct, operating
system. Maybe they will do this again for those cards they cannot
financially justify supporting. Has anyone in the FOSS environment
approached them? This might put Fedora and the upstream company above
their competition if they did.
James McKenzie
More information about the users
mailing list