Kernel source -- INCORRECT ASSUMPTIONS!

[Bryan J. Smith]

On Mon, 2006-04-10 at 18:24 +0200, Arjan van de Ven wrote:
> this is one of the reason such (illegal) binary modules are bad. People
> don't report or work with the open driver to get that one fixed, but
> rather just jump to the binary junk.

*WRONG*!

But nVidia _does_ work with the GPL forcedeth team!  They _do_ report
what changes!  And it goes into the _next_ stock kernel, or gets patched
in the next Red Hat kernel.

They just offer a "quick fix" for those that don't want to upgrade their
kernel!  That's all!  It's _not_ nVidia, but mainboard vendors who use
different PHY combinations with the nForce MAC.

In fact, nVidia usually catches it, immediately rev's the nvnet driver,
and then works with the kernel team on the necessary forcedeth changes.

> But that way linux as a whole doesn't make progress and it doesn't get
> fixed for the future, giving again a crappy experience in the next
> release.

Again, *WRONG*!  100% of the nVidia nForce chipset not only has GPL
drivers, but nVidia *ACTIVELY* supports their development!

On Wed, 2006-04-12 at 06:29 +0530, Tamal Kanti Nath wrote:
> The real problem is NVIDIA does not distribute their drivers to Linux
> community like HP does.

*STOP*!  100% of the nVidia nForce chipset is GPL!  That include
reporting to and working with the kernel team.

The only kernel module nVidia doesn't make GPL is the memory code for
its GeForce cards.  Some of that code is actually IP of Intel (among
others), who nVidia has a NDA with.

At one time nVidia's AGPgart was also not released into the kernel, and
only with that driver.  But Intel lifted the NDA on that once
PCI-Express came close to release.

Unlike nVidia, Intel doesn't unify its drivers across all platforms.
That's why Intel puts a "crippled" driver on Linux, without that code.
nVidia doesn't, hence why Intel enforces its IP.

> So users have to download and install their driver manually.

No, you're thinking of the video driver.

And even then, nVidia _actively_ puts people on the 2D MIT "nv" driver.
They typically _beat_ ATI on 2D support in Xorg/XFree for newer cards.

Don't shoot nVidia because they offer a 3D GLX atop.  ATI _never_ has!
It was the National Weather Service that paid Precision Insight to write
DRI drivers for the R100 series, which works for the R200, but not the
R300+ because ATI has withheld those specs.

Intel gives you a rather crippled 3D driver, at least compared to
Windows, to avoid disclosing their own IP (among others).

> Users know very little about fixing a problem (if any error occurs).

Not any different than any other chipset -- ATI, Intel, etc...

nVidia is just as supportive of GPL driver development for the nForce
chipset as Intel is, although their SATA design in their MCP isn't as
good as Intel's ICH.

> So, they had to use the generic driver provided by the vendor.

No, that is _not_ true!

The "forcedeth" works fine.  The "nvnet" is just there if you don't want
to upgrade your kernel to work with new MAC+PHY combinations mainboard
vendors introduce (_not_ nVidia!).

> I do not blame NVIDIA because Linux kernel (and related products)
> updates rapidly. Also there are many many flavors of UNIX/Linux in the
> market. So, it is impossible to build drivers for each Hardware and
> software platforms.

Well nVidia _does_!
I rather _tire_ of the _ignorance_ on nVidia.

The only kernel driver they make proprietary is their memory driver for
their GeForce video cards -- that's _it_.  They support ALSA, forcedeth,
SATA, etc... in 100% GPL developments.


-- 
Bryan J. Smith             Professional, technical annoyance
mailto:b.j.smith at ieee.org       http://thebs413.blogspot.com
------------------------------------------------------------
****** Speed doesn't kill.  Difference in speed does! ******