Subject: Re: [ppc-mobo] Re: GPL-like hardware design license?
From: "John Metzger" <john_metzger@worldnet.att.net>
Date: Mon, 04 Oct 1999 13:58:43 -0700

> On Mon, 4 Oct 1999, Richard Stallman wrote:
> 
>> I think this is basically impossible; I wrote an article
>> explaining the reasons, which was published in Linux Today.
...
> ... it's available at
> http://features.linuxtoday.com/stories/6993.html
>
> Google also found this link, which contains the core points:
> http://www.openresources.com/news/June_23_Free_Hardware.html
>
> He does make a good point about the difficulty of protecting
> the openness of the design though a copyright on a circuit representation.
> I'd been thinking that would actually be sufficient, combined with a
> copyleft-like license requiring derived works to be protected as well:
> A copylefted schematic would require that any layout generated from that
> schematic also be copylefted, and likewise, distibutors of hardware based
> on a copylefted design would be required to make available the design
> files used in its production.
>
> This last bit needs some fleshing out. Does it make sense to require that,
> e.g. the gerbers be made available as well as the schematic? (The GPL
> doesn't say anything about the status of intermediate assembly generated
> by the compiler.) I don't see the same broadly-understood distinction
> between source and executable upon which to hang the requirements for
> distribution. I suspect such a license will look fairly different from the
> GPL in detail. Still, I'd like us to develop a license for hardware
> designs in the same spirit as the GPL.

I hope not. Certainly IBM's POP won't follow the GPL requirement of making
public all changes made to the design. IBM's purpose is to sell more chips,
not enlighten the world.

Let's take the different elements of the design, the schematics, the layout
(Gerber files) that result and the chips (with or without embedded
firmware).

First the schematics. I doubt it will be possible to copyleft the
schematics. There are only so many ways to hook up the chips and we should
not restrict people from how they use the design. If someone modifies the
design and keeps it private so what? For instance, suppose we do have our
own design which is not derived from IBM's POP. Suppose it uses all the same
chips as POP. The schematic will for all practical purposes be very much the
same as POP. Data Bit 0 from the CPU is connected to Data Bit 0 of the north
bridge, etc. Now further suppose someone decides to upgrade the design by
adding a USB and Firewire controller and changing the north bridge. Are they
required to publish their schematics? If they want to keep them private it's
a _small_ job to simply create a "new" schematic from "scratch."

Next consider the Gerber files or layout that results from such a changed
schematic. Of what real use is this to anyone? It is only useful if you
intend to build the PCBs. That's an expensive proposition even if your uncle
owns the PCB fab. Assembly isn't easy either. You'll go blind trying to hand
place and solder all those little itty, bitty tiny resistors and capacitors.
This step of actually making the hardware from the design is expensive no
matter how you cut it. Making a single unit is fairly expensive, making 10s
of units isn't much less expensive (per unit), and making 1000s of units
begins to bring the per unit costs down dramatically. So what good do the
Gerber files do for a modified design unless you intend to make lots of
them?

Lastly the silicon. Certainly in the short term you won't get an "open" CPU
chip. We are unlikely to get an "open" bridge chip either. If we are using
off the shelf components nobody will be making "changes" to them. The most
we can get is complete documentation on how the "black boxes" work.

If we do our own chips then again if someone modifies the design what good
does it do to have the "mask" set unless you are going to build the silicon,
which is the same problem as the PCB, -- it's expensive to set up.

If someone is willing to invest the money and time it takes to modify the
chips and produce better chips, but is unwilling to publish the design,
would you rather not have the improved chips? If they sell the chips at what
ever price they think is fair isn't that enough? I'd certainly like to have
the improved designs published but not at the expense of not being able to
buy an improved design, be it a chip or motherboard.