James Richard Tyrer wrote:
Jeff Garzik wrote:
James Richard Tyrer wrote:
Intel claims that they provide complete documentation. VIA isn't that
Intel has certainly -not- provided full docs.
An open source X driver with 3D acceleration exists. How was that
developed without documentation?
It was developed without published documentation... by Intel engineers.
The same way Intel did their ixgb 10gig ethernet driver, for which no
docs have been published.
False. Trivial inspection shows a full, open source, 2D/3D
accelerated driver for VIA is available. Taking two seconds to do "ls
drivers/char/drm" in the kernel source tree would have shown you reality.
Is that for the UniChorme chip? The dox says that it is for the
MVP3/Apollo chip set.
The driver which VIA supplies for the UniChrome chip (FBDev_Lite) is a
frame buffer
VIA has supplied full UniChrome port to the right people. Its not
committed upstream yet, but its open source. Just too ugly to merge
(the perils of vendor-written code).
The Linux market isn't asking them for an investment. What we have
been asking for is the full documentation of the chips. Note that
graphics
False. Full documentation is an investment that requires legal $$$,
plus opportunity cost, plus potential cost of your company's primary
product being supported at a sub-standard level.
chips are an anomaly here. All other chips I know of have full
documentation available.
Ok that is just laughable. There are so many counter-examples its not
even funny. Easy examples: Ethernet, SATA (areas I deal with every
day in the Linux kernel)
Say WHAT? How do they sell chips if they won't supply documentation?
Vendors supply the drivers that drive the chips.
Often in embedded situations, the chip designers and driver writers sit
next to each other, which lends itself to substandard or non-existent
documentation _even internally_.
and winmodems
A winmodem is not a chip, it is a board or module. Some may have custom
chips, but they started out as a DAC and an ADC -- better ones also have
a DSP or controller chip. What they won't give out is the driver for
the board, but then some of the manufacturers started giving out a
False. Winmodem boards are trivial, its the chips that are cloaked and
undocumented. It took two seconds with 'lspci' and some PCI bus
snooping to determine how to talk to the boards.
And once you've reverse engineered the DSP chip instructions, you still
have a patent minefield when it comes to modem protocols.
driver for Linux. This is a situation similar to the graphics drivers.
The driver is proprietary, but there is stuff in it that can't be
patented.
all have hardware vendors that refuse to give out docs.
I have not looked at SATA chips, but all the NIC chips I have looked at
are documented. Perhaps you are using the word "documentation" in some
different sense. If they tell you what all the registers are for, that
is full documentation.
Show me a 10gig ethernet NIC vendor that has provided documentation to
open source developers.
Any vendor will do.
There is also an endless list of gigabit NICs without disclosed
documentation: EN13xx, JATO, tons of embedded chips, ...
The law guarantees that they cannot go into the open source market in
a big way.
Explain this please!
I suspect that the problem is that they don't have a patent. They can
<blink> Um... ATI and NV have bunches of patents. Search the patent
database. Graphics is a patent minefield.
But, there is OpenGL which stands as prior art. Unfortunately,
companies patent everything that they can knowing full well that 90% or
more of it won't stand up if challenged.
Have you been reading this thread at all?
Perhaps a few more words would have made this clearer. They are
obviously trying to protect something which is not patented.
OpenGL is an API. Its the implementation details (software and
hardware) which are patented.
Trying to claim that OpenGL is prior art that invalidates ATI/NV patents
will get you laughed out of the room by any patent lawyer.
If everything were protected by valid patents, then it has all been
publicly disclosed and there is no reason not to provide documentation.
There is the cost of user confusion due to competing implementations,
the cost of volunteer support damaging your brand, support costs, tons
of other costs.
Attempting to directly correlate patent validity with documentation
disclosure is just silly. It ignores 1001 other real world factors.
copyright the drivers and copyright the chip, but that doesn't
prevent reverse engineering of the chip, or making a clean room clone.
Without a patent grant, you would be infringing if you distributed the
RE work product.
This issue is currently an area of contention. ARM has managed to get
all of the clones taken off the market. But, that is a processor, not
an ASIC. OTOH, clean room RE of the IBM-PC BIOS was determined not to
be a copyright infringement. If they tell me what all the registers are
for and what the commands do, and I independently design a chip that
does exactly the same thing -- produces exactly the same output -- with
NO knowledge of the inner workings of their chip, have I violated a
patent. I doubt it. You can't patent a register layout. You can't
patent an instruction set (specifically you can't patent a programing
language). The patents would be on the methods that the internals of
the chip are implemented with and if I do a clean room clone, I
deliberately don't know anything about that.
I agree its an area of contention. The courts have upheld certain MIPS
patented instructions, and also upheld patents where clones cannot
reproduce a certain behavior without duplicating patented logic.
We can certainly argue that the patent system is horribly broken, but we
must deal with the reality of case law.
Jeff
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)
