Do we need an open hardware license? (Linux.com)

Frankly, the first ones to make such a thing for a video card that has some really good horsepower in the 2D realm but especially in the 3D realm, I for one will be the first to buy one.

On the motherboard front, I would love to see someone do the same and use the LinuxBIOS. But then if the existing mobo manufacturers were to use LinuxBIOS that would be an equally good thing.

In the end if such things truly existed, it is where my money would go.

http://www.tapr.org/OHL

Although there are major differences, open specifications could be very useful. You see a number of processor vendors implementing about the same processor specification. In other area's, e.g. graphics processors, you see a lack on open specifications, resulting in hardware supplied with only drivers for one specific OS. Often even access to the circuit specifications is not available.

So that if you buy a piece of hardware, you know that the hardware *can* be driven by Open Source drivers if it contains the "Open Hardware" stamp.

It is frustrating to go to a shop to write down what hardware they have available, to go back home to look up whether the hardware in fact will work properly, then go back to the shop and buy the hardware, only to come home and find out the second revision suddenly uses a different chipset that does not work. Then bring back the hardware and try another piece or switch shop and do it all over again.

If you want to enlighten customers, they have to know what to look for in the shop without being a technology expert.

My students at Portland State University also produce open hardware in the form of schematics and PCB layouts for circuits. As I understand the law, while the schematics and PCB layouts themselves are protected by copyright, hardware produced from them is not. We've been slapping the GPL on them anyway, but it probably has little legal force.

True hardware, i.e. devices, is typically protected by patent and trademark. The clever trick that makes the GPL work thus is difficult to apply to them.
It's all quite complicated. As open hardware grows, there will be some sorting out to be done.

A hardware design (video board, disk controller, A-D/D-A, motor controller, etc.) could be described in terms of its design and interface. Depending on the level of design, a reference implementation might be an FPGA that implements the design, or a circuit composed of many parts (board level or subsystem level). An open HW design may contain non-open parts (such as memory, CPU, op-amps, etc.), but the design itself could be used, modified, and returned to the community, assuming that the parts were off-the-shelf.

A commercial venture might take an open design, implement it as an ASIC or incorporate it in a larger product. The Open license should require that any changes to the open parts of these products remain open, and that changes within the design also be open, but like the LGPL, should not require that those portions of the end product that are not directly based on the part. Actual ASIC "code" or templates would not need to be released, but anyone else with sufficient resources could implement the same design without fear of infringing.

This would give designers a set of Open Source building blocks from various sources that could be assembled during the design phase to solve a particular problem. Details such as what constitutes a derived design need to be explicitly stated in the license. A BSD style attribution clause stating that you always have to give credit where it's due even if the overall product design is not open source is not a bad idea.

This is separate from a certification of Open Source Compatible Hardware, which would require an organization such as OSI to certify that a manufacturer provide sufficient documentation for a product to write a driver that takes advantage of all of its features, and that the documentation is available on a non-restricted basis that allows a Free and Open driver implementation. In this case, the hardware itself need not be "Open Source", just available to it.

I think we actually need not one certification for Open Hardware, but *2*.

First, we need a certification for hardware that is, by itself, open. Where you get the sourcecode for the actual hardware, and have the 4 classical freedoms to tinker, improve, make derived products, etc. We won't have much hardware under such a certificate anytime soon.

Second, and perhaps more important, we need a certification that says: "The API for this piece of hardware is completely open. We publish complete documentation as to how to talk to the device."

I think the second matters more than the first at the moment, and it's much more realistic -- a lot of devices conform already, by virtue of communicating over an already open, well-defined protocol. All IDE-devices lacking proprietary "extensions" to the IDE-command-set, All devices correctly implementing "Usb-storage", all web-cameras correctly implementing "usb-pointer" and so on.

It would be very nice to be able to see at a glance in the store which ones of the available web-cameras (for example) are open, in the sense that docs on how to talk to them are available, and which ones use proprietary crap.