[ohf-licenses] [open-hw] Classification of openhardware
Greg London
email at greglondon.com
Sun May 11 13:03:26 EDT 2008
> Date: Sat, 10 May 2008 14:24:40 -0700
> From: Jonathan Gray <jonathan.gray at okfn.org>
> Subject: Re: [ohf-licenses] [open-hw] Classification of openhardware
> Frederic Renet wrote:
>> And more generally how do we define open hardware ?
>
> I've also been wondering about this.
I think open hardware would center around any works which fall under
copyright law, which describe and specify something which could be built
as a physical device from that spec.
In some cases, copyright law would apply to the spec and to the device.
The simplest example would be a design for some hardware core which then
got implemented in a Xilinx FPGA, and that FPGA was programmed via a
compiled bitstream.
The physical device operates directly off of the copyright work. The
bistream is a derivative of the spec, so if the spec were protected under
some copyleft license, then the bitstream would be protected the same way.
The problem is, in the above example, you can take the design and convert
it into an ASIC design instead of an FPGA. And that ASIC is nothing but
functional bits, semiconductors and wires. And that ASIC is no longer a
copyright derivative of the spec.
So someone could take the original work, improve it, keep the improvements
private, sell ASIC's containing those improvements, and a license such as
the GNU-GPL would not be able to protect the community because the
copyleft aspect of GNU-GPL is triggered only when the derivative is
distributed. And an ASIC is not a distribution as far as copyright is
concerned.
The solution seems to be to adapt a copyleft licene which applies to all
derivatives, distributed or not. This means that private derivatives are
not allowed. But it also means that the private derivative that comes from
someone distributing only an ASIC will still require that person to make
their derivative available to the community who provided the original.
It turns out there is a license that does this.
The Apple Open license.
It was designed for software and it was designed to close a specific
loophole in GNU-GPL. Websites could take GNU-GPL software, modify it, use
that modified software to host their website, and not actually distribute
the software. Because they didn't distribute, they didn't have to make
their changes public. So the Apple license closes this loophole by saying
all derivatives must be made public, whether you distribte those
derivatives or not.
The one other caveat, then, would be that the Apple license would apply to
the work, and to any thing into which it was compiled. Based on the way
hardware works, this would be problematic, since as soon as someone pulled
in a Free Hardware Fifo, the copyleft aspect of "derivative" would infect
the entire chip, which would mean that most manufacturers would likely
never use free hardware.
To address this, I proposed that Open Hardware would use the Apple license
but that the definition of "derivative" which would trigger the copyleft
aspect of the license and the making-public of those derivatives would be
limited to whatever hierarchical level of the design as it exists, and
down.
If you instantiate an Open Hardware fifo in your design, then your design
is not a derivative and does not need to be made public.
If you modify the fifo design and then instantiate it in your chip, then
you have to make the modificatiosn to the fifo public.
An interesting side effect of this license approach is that the actual
definition of what is and is not Open Hardware becomes less important.
The license protects copyrighted works, and if those copyrighted works are
modified, they must be made public. It becomes little different than using
the Apple license to protect software or fiction or any other written work
protectted by copyright law.
The Open Hardware license could conceivably be used to protect software.
Or to protect an ASIC design. Or to protect a design for a jet pack. Or to
protect a novel. Or to protect whatever that will fall under copyright
law.
It just happens to be a license that specifically will protect hardware,
but it does so without requiring any specific definition of what
"Hardware" is.
Greg
More information about the ohf-licenses
mailing list