Drivers for hardware *in* the device

[motorollin]

Why are we not seeing hardware which contains its own driver? This seems entirely feasible to me. For example, an AGP graphics card could have an area of flash memory which contains the drivers required. If the card is supported by multiple platforms the a jumper can be used to select Windows/MacOS/Linux/None (where none bypasses the built-in driver). If the developer releases an updated driver, then the flash memory on the card is updated with the new version with an update utility (instead of installing the driver on the computer). I see no reason why this couldn't work with USB devices as well.

Somebody is bound to ask what the point it. Well imagine how much easier it would be to install new hardware if you only had to plug it in and never worry about installing drivers. It truely would be plug and play.

--
moto

[Speelgoedmannetje]

It can work, and it should have been done a long time ago with USB. But hardware manufacturers figured out it'd be too expensive for a 'too insignificant feature'. User friendliness doesn't sell as well as megaherzmyths (and alike).

[Karlos]

Drivers conforming to which API, though? Consider windows alone. What you commonly consider to be the "driver" for a given graphics card usually ends up being a range of drivers that allow it to work with various APIs (basic 2D, video, 3D etc). You've then got different versions of those drivers for different versions of the API's they provide the services for.

If you look at the DirectX drivers for a typical modern graphics card alone, they can be several tens of MB in size and that's before you throw in support for DirectX 7,8,9,10 interfaces that could reasonably be expected to exist.

OSX also has multiple driver layers to support various APIs. There might be a single unified driver at some layer which provides all the services required by Quartz, Core Graphics etc. but even then I doubt it, to be honest.

Providing a hardware level detection and capability reporting really is the most pragmatic choice and funnily enough, that's generally what the cards have already.

[motorollin]

Are you saying that this diversity of standards makes embedded drivers impossible, or just more complex?

--
moto

[Karlos]

I'm saying it makes it impractical to the point of being worthless.

-edit-

That is, unless, a completely unviversal API is developed for each class of device that all manufacturers can work to.

[motorollin]

Shame the standards are so non-standard :roll:

--
moto

[Karlos]

Those standards that are employed are pretty well defined. But they only cover the basic configuration of the device and it's connectivity into the system.

[Oliver]

motorollin wrote:
Shame the standards are so non-standard :roll:

--
moto

No,no,no.  The great thing about industry standards, is there are so many to choose from.