|
|
|
| Authored by: greed on Wednesday, June 20 2012 @ 10:46 AM EDT |
I always figure companies that won't even document interfaces do so because
you'd find out how little you were getting for your money.
Basically, you'll find out your special magic hardware board is really a fairly
generic processor with software making it do the magic commands. Like WinModems
and WinPrinters and ST-512 hard disks. (OK, that last one didn't even have its
own processor.)
Having worked for software companies for the last 20 years, I believe most
companies are opposed to open source because, with open source, their customers
would find out how badly they're being ripped off.
People coding for fun in their spare time generally do a much better job than
9-to-5 wage slaves.
[ Reply to This | Parent | # ]
|
| |
| Authored by: mbouckaert on Wednesday, June 20 2012 @ 02:10 PM EDT |
Unless one believes that hardware has to be open sourced
too, I do think that cursing this manufacturer because the
software layer that is "in direct contact" with hardware is
kept confidential is unwarranted.
A driver is where software commands make things happen in
the physical world. Software doesn't overheat; hardware
does. Software knows no duty cycles. Software doesn't use
or dissipate power. Etc.
Not even mentioning that the documentation of the hardware
cannot be completely specific about all such limitations,
the hardware manufacturer could sit with a piece of
equipment to replace under warranty, if some instruction
path causes a hardware limitation to be overstepped.
If the driver software is proprietary, themanufacturer knows
the risk, bites the bullet, and replaces the hardware. It
probably
(a) runs plenty of lab tests where the hardware is
instrumented and (b) asks its engineers to code
conservatively around the hardware limits.
How would you handle open-sourced drivers that can be
"tweaked" by some ? The tweaker does not have the option of
instrumenting the hardware. In stronger words: the tweaker
does no engineering, just hacking in the dark.
Actually, the above applies not only to NVIDIA, but to any
manufacturer whose hardware interface exposes serious
physical limitations and allows these bounds to be exceeded.
IMHO, they have the right to keep the lowest layer of code
close to their chest, *but* the obligation to state why
(otherwise it's Tivoization).
That layer of code should be regarded as an extension of the
hardware. And in the NVIDIA case, they provide the
appropriate lower-layer-to-Linux intercae, and update
regularly enough.
Now light up the flame torches, asbestos underwear in place.
---
bck[ Reply to This | Parent | # ]
|
|
|
|
|
