Linked by Eugenia Loli on Sat 12th Aug 2006 19:01 UTC, submitted by fish reloaded
3D News, GL, DirectX Statement by ATI: "For other markets, such as workstation and consumer, performance and feature differentiation are key metrics. Proprietary, patented optimizations are part of the value we provide to our customers and we have no plans to release these drivers to open source. In addition, multimedia elements such as content protection must not, by their very nature, be allowed to go open source."
Thread beginning with comment 151865
To read all comments associated with this story, please click here.
Big OS mistake
by vtolkov on Sun 13th Aug 2006 04:17 UTC
vtolkov
Member since:
2006-07-26

Big mistake, but Linux's one, not ATI's. OS world should reconsider it's driver strategy. Making driver internal and declaring driver to be derived from Linux does not allow to have normal drivers. Without drivers OSs are just students' toys, as it used to be for a few decades. Everyone seems to be happy with this position.

API is a point of integration for both drivers and applications. Using documented API should not mean to be derived.

Reply Score: 1

RE: Big OS mistake
by ajuc on Sun 13th Aug 2006 19:26 in reply to "Big OS mistake"
ajuc Member since:
2006-08-13

Binary drivers aren't normal drivers. It works only with stable ABI systems, such as windows (but not beetween all versions).

Open source programs, such as XOrg, take ABI unstable, and change it often, but keeps source API stable. When (as in new XOrg 7.1) ABI changes, every binary drivers that are supposed to work with it don't. So Manufacturer must compile it's drivers for new API, and keep old to support older systems.

Stable binary API isn't really good, becouse it keeps from improving program, and often ends in making ugly hacks to make things working (padding structures with reserver fields, making one variable meaning something different, such things). If only manufacturers take documentation to their products, or open their drivers, it will be no problem, people will recompile drivers when it ABI changes, and distribute it within days.

But when there are no documentation nor open drivers, it is a big problem, and will be bigger, as open source programs will change faster (as XOrg now).

So: documented API from OS side is must. But documented API for hardware is must, too. Without them, manufacturer must support all changing systems that his hardware work with. If there is no monoculture (like in windows world), he can't.

Reply Parent Score: 1

RE: Big OS mistake
by draethus on Mon 14th Aug 2006 07:47 in reply to "Big OS mistake"
draethus Member since:
2006-08-02

Big mistake, but Linux's one, not ATI's. OS world should reconsider it's driver strategy. Making driver internal and declaring driver to be derived from Linux does not allow to have normal drivers. Without drivers OSs are just students' toys, as it used to be for a few decades. Everyone seems to be happy with this position.

It's virtually impossible for Linux to reconsider its driver strategy (or change its licence to eg. GPLv3): there are over 500 contributors, some of which have died.

And no, many people are unhappy with this position, including me. It's highly unrealistic to expect everyone to write open-source Linux drivers for every device.

The best way out is probably to use user-space drivers. User-space drivers can be closed source, and the kernel's user-space interfaces are stable. libusb already lets you do full USB access this way, and there are several ways to get user-space PCI access (bochs's pcidev kernel module and gelato project at http://lwn.net/Articles/66829/ and http://www.gelato.unsw.edu.au/IA64wiki/UserLevelDrivers). Apparently PCI drivers in user-space are fast and have many other benefits.

The problem is, nVidia's drivers (and probably ATI's) dig quite deeply into the kernel, accesses the BIOS and system-management mode, and other horrors.

Reply Parent Score: 1