The Tricky Business of “Copylefting” Hardware
[Diesen Artikel gibt es auch auf Deutsch.]
It’s probably safe to say that the copyleft principle has been essential for the success of free software. Copyleft means that all versions of a software or document will remain free, preventing companies from creating “value-added” versions of free programs and selling them as proprietary, non-free software. The GNU General Public License (GPL)—the first and most well-known incorporation of the copyleft principle—is used for about 50–70% of all free programs, making it more popular than all other free software licenses together.
At first sight, the situation in the newly emerging field of free and open hardware might seem similar—here, copyleft licenses such as the GPL and the Creative Commons Attribution-ShareAlike License (BY-SA) are very popular too (see below for a more detailed analysis). But actually, the situation is very different for hardware design, since copyleft relies on copyright, and hardware is (in most cases) not protected by copyright law.
The Problem: Copyleft Depends on Copyright
In hardware design, the copyleft clauses of the GPL and the CC BY-SA license thus don’t work in the way you might expect them to work. Anybody who modifies and spreads the design description is bound by them, but people and companies using the design to build and distribute actual hardware are not. Vendors can modify the hardware and sell or distribute the modified hardware without having to publish their modified design, since licenses only govern the design descriptions (the instructions on how to build the hardware), not the hardware itself. Here’s how the FSF puts it in their GPL FAQ:
Can I use the GPL to license hardware?
Any material that can be copyrighted can be licensed under the GPL. GPLv3 can also be used to license materials covered by other copyright-like laws, such as semiconductor masks. So, as an example, you can release a drawing of a hardware design under the GPL. However, if someone used that information to create physical hardware, they would have no license obligations when distributing or selling that device: it falls outside the scope of copyright and thus the GPL itself.
The problem is that copyright governs only the distribution of information, but not its usage. If somebody has access to information, they can use them and act upon them in any way they like without violating copyright law.
If a bookseller publishes a book describing how to build a product X, the bookseller cannot prevent me from using this information to build and sell X. The book is protected by copyright laws, but the factual information contained in it is not, and my use of that factual information falls outside of copyright laws. The GPL relaxes the restrictions of copyright, giving me more freedom to use copyrighted information than I would usually have, but of course it cannot deny me the freedoms which I always have, even with full copyright in force.
A Solution: Patents?
I know of only one license that is specifically targeted at open hardware and tries to work around this issue by relying not only on copyright, but also on a mutual patent immunity clause: the TAPR Open Hardware License, published in May 2007 (Version 1.0). The idea is that the licensor grants everybody who complies with the terms of the license the right to use any relevant patents they control. So if you don’t comply (e.g. by not publishing the “source” for physical products you build based on the licensed information), you might get sued by the licensor for patent violation, but if you comply you’re safe.
TAPR has published two versions of the license: the Open Hardware License (OHL), which emulates standard copyleft (like the GPL or CC BY-SA), and the TAPR Noncommercial Hardware License (NCL), which additionally prohibits commercial usage (like the CC BY-NC-SA).
Relying on patent law instead of copyright law for hardware makes sense, since patent law is made for hardware, while copyright law is made for information. But it’s also a problem, since getting a patent is a difficult and costly process, while getting copyright is free and automatic.
The TAPR licenses will only be really effective if the licensor possesses relevant patents, or at least if people can’t be entirely sure that (s)he doesn’t have any such patents. That’s a pretty bad limitation, since patents are difficult and expensive to get. Most peer projects will be unable or unwilling to apply for a patent, so the TAPR licenses will hardly be suitable for them.
Another unrelated problem of the TAPR licenses is that they require you to distribute the original designs you’ve received from others in addition to your own modified version—you must distribute both “before” and “after” versions of any files you’ve modified (§ 4.2 (b)). This might be extremely impractical for large version histories (think of hundreds or thousands of versions), and it makes the creation of printed versions of modified hardware documentation practically impossible. It’s also unclear what’s the point of this requirement.
Pragmatic Solutions: Just Use Standard Copyleft—Or No Copyleft Altogether
Most open hardware projects seem to care little about the specific issues of hardware licensing. Most projects aiming for copyleft just apply a standard license such as the GNU GPL or the Creative Commons BY-SA license, apparently either not knowing or not caring that is won’t apply to building hardware (see list of projects below).
Another solution is to forgo copyleft altogether and just use a permissive license such as the (modified) BSD license. This allows everybody to use the provided information in any way they like, without having to keep modifications or improvements open.
There even is a hardware-specific non-copyleft license, the Balloon Licence. The Balloon Licence is a simple MIT-style license. Since it does not try to apply any restrictions on the manufacturing of hardware (no copyleft or non-commercial clause), it doesn’t have the problems of other open hardware licenses. However, since this license is very similar to the commonly used MIT and (modified) BSD licenses and doesn’t address any new issues, there seems to be little reason to choose this special license instead of one of the standard ones.
What To Do?
As can be seen from the list below, there is a clear tendency of projects to use standard copyleft (the GPL or the CC BY-SA), protecting the design information itself but not any physical hardware built on their basis. Does this mean that projects are willing to live with a limited copyleft, or are they just unaware of the problem? I’m not so sure…
In any case, no convincing solutions seem to exist. The TAPR license, the only license that tries the address the problem of extending copyleft to the hardware itself, never became very popular—almost nobody outside the TAPR project seems to be using it. And indeed, its reliance on patent law makes the license very impractical to use for all but the biggest projects.
On the other hand, just using permissive licenses (like the Apache project and the BSD family do) doesn’t seem to be an attractive option for the majority of open hardware projects—most prefer to get at least the partial copyleft protection that standard licenses such as the GPL and CC BY-SA can give them.
But, as described above, their copyleft protection is very incomplete in case of hardware—hardware builders aren’t required to give back their improvements. In the case of the RONJA project, this difference between hardware builders not giving back but the project maintainers expecting them to do so, has already caused tensions that contributed to the downfall of the project. Whether similar problems will appear in other projects and weaken the open hardware community, remains to be seen…
* * *
Appendix: Project Listing
The following projects use standard copyleft licenses:
- GPL:
- Bug Labs, modular open source computer hardware (cf. Bug Labs: License)
- Elphel, high performance cameras based on free software and hardware designs (cf. About Elphel)
- RepRap, an open source fabber (cf. RepRap GPL Licence)
- Sun’s OpenSPARC project, publishes free hardware designs for the UltraSPARC T1 and T2 processors
- Free Telephony Project
- CC Attribution-ShareAlike:
- Arduino (cf. Arduino Hardware) and Freeduino (cf. Freeduino PCB Design Files), two variants of a free microcontroller board. They are aware that the CC clauses don’t apply to physical production: “You don’t need to give attribution on the physical products that you might make with these files, as copyright in general, and CC2.5 in particular, do not apply to physical objects.”
- Beagle Board, a free low-cost computer board based on a Texas Instruments processor
- MultiMachine, a low-cost multi-purpose machine that can be built with common hand tools
- Openmoko, open source mobile phones (uses BY-SA for CAD Files and schematics)
- GNU Free Documentation License:
- Ronja, optical data transmission (cf. Ronja Copying)
Projects using a standard share-alike license that prohibits commercial usage (CC Attribution-Noncommercial-ShareAlike):
- Ronen Kadushin, free furniture designs
Projects supporting various Creative Commons license:
- Open Architecture Network, a repository of architectural designs and plans; users can choose any CC license, including public domain (cf. OAN Licensing)
- Thingiverse, a repository of digital designs for physical objects; users choose either standard copyright or a CC license, GPL and public domain are also supported (cf. Thingiverse Terms of Services)
- Ponoko, a repository and marketplace for digital fabrication; available choices are full copyright and Creative Commons licenses (list of CC-licensed designs)
Projects using the TAPR Open Hardware License, a share-alike license designed specifically for hardware which relies on patent rather than copyright law:
- TAPR, the community of radio amateurs which developed that license
Projects dual-licensing between TAPR License and standard copyleft:
- Open Graphics Project, open source graphics cards; three license alternatives for its hardware descriptions (schematics and artwork): GPL, TAPR License, or a commercial proprietary license, cf. OGP FAQ and Announcement from 7 April 2009. Of course, dual/triple-licensing with TAPR as one of several options makes the special protections of the TAPR License void (users can just choose the GPL instead), so this multi-licensing model seems somewhat pointless.
Projects using a non-share-alike license specifically developed for hardware (the Balloon Licence):
- Balloon Project, another circuit board (only parts of the design are free, some design files are kept proprietary)
Projects using a standard non-copyleft license (the BSD license):
- Fab@Home, another project for building an open source fabber (cf. Fab@Home: General disclaimer)
Mixed:
- OpenCores, develops computer components for CPUs, memory controllers, peripherals, motherboards etc.; uses a mix of GPL or (preferably) LGPL and modified BSD license, depending on the preferences of the contributors (OpenCores FAQ)
If I have forgotten important projects, add them in a comment and I’ll update the article. I didn’t try to cover projects that are small or in early development. Long lists of open hardware projects can be found in the Wikipedia article on open source hardware, the P2P Foundation’s Product Hacking page, the Open Innovation Projects site, the GOSH List of Open Hardware Projects and the GOSH List of Open Hardware Organizations.
Thanks, great research!
What I miss is the role of „prior art“ as a different way to deal with the patent problematique.The good thing about hardware is that you dont „flood“ and „block“ the market if you build a proprietory instance, as long as the knowledge to build it is free. The whole reason for the GPL to exist is in my view that information is different than hardware. Since software is reproduceable, any commerecial product could be easily uncommercialised by copying. So copying is prohibited and creativity is killed. We simply dont have this problem with hardware as long as patents dont apply. You can sell things but by selling them you dont necessarily endanger someone elses business by creating a potential myriad of the object itself. The best protection for free hardware is simply a well-maintained documentation pool, that would allow anyone to build. Thats pretty much the same as the Public Domain.
The social process that encourages hardware builders to give back their improvements is at the core of things, and endavours like fairware by Thomas Diener are starting to address this issue.
And of course, eventually the business logic will be transformed itself. I think currently we have to feel comfortable with unleashing a business logic that spurs unlimited cooperation – and we will see the ripening of conditions that will at some point convince everybody that abandoning the business logic itself is the way to go.
There is an interesting answer from Graham to this article on Oekonux mailinglist, which I repost here:
Really interesting post. And thanks for starting the debate about Open Hardware..
Regards from the Ponoko.com team
@Franz: I don’t really understand the difference between hardware and software you’re making–if you create and sell a proprietary „value-added“ software, you don’t „flood“ and „block“ the market either. Any if somebody builds proprietary hardware, it’s usually not trivial to reserve-engineer and duplicate their work (even if not protected by patents or similar). Hence it certainly makes sense to ask for „giving back“, for sharing any improvements made, just as copyleft does.
But in general, I tend to agree with Franz’s and Graham’s points in favor of going for public domain (or BSD-style licenses, which are roughly the same). Maybe share-alike is indeed less important for hardware than for software…
Dear fellows,
recently I had been at the piksel festival in Bergen, Norway, to continue with the work we had started at http://www.gosh2009.ca about licensing Open Source Hardware devices.
As I can tell from my own observations and contacts the topic Open Source Hardware is developing more and more momentum. On the other hand until today there is no applicable way to connect the physical outcome of free open source designs (the devices and objects) to a copyleft license, as mentioned here already.
At GOSH! and there after we had many talks with various lawyers in Canada, the US and Europe, and we finally seem to have found a valid way to work around this problem. The „trick“ is a trademark. Please find attached our latest results on this idea and the process of it’s application. [1]
As you can imagine, we would all still be more than happy if we would not have to form another legal entity for OHANDA, but rather work together with FSF and/or Creative Commons. We would be very pleased – and so would be most of the open source hardware developers – if the trademark and process were hosted and held by the FSF and/or CC instead.
Therefore I wanted to keep you updated on our latest activities and please you to support OHANDA to get FSF and/or CC to support/adopt our approach. We would all be more than happy to work together with either or both of them to get a copyleft like licensing process off the ground for open hardware designs, devices and products.
Curious about your comments!
Very kindly,
Juergen
[1] http://www.gosh2009.ca/wiki/images/Ohanda_piksel.pdf
@Juergen: that sounds like a very interesting idea, but I’m not quite sure how it is meant to work. So, using the OHANDA trademark is only allowed if you ensure that your hardware is open hardware with the 4 freedoms guaranteed. Others who download and use the hardware designs must agree with the same conditions, if they want to use the trademark too, leading to a kind of copyleft. Right?
But then, won’t people just be able to download and use the hardware designs, without using the trademark and thus without being required to share their own improvements? They couldn’t call their variant „OHANDA-certified open hardware“, but they could still built and sell their hardware, without having to comply with the licensing terms. So it wouldn’t be really copyleft-like, would it?
Or what’s your trick to make everyone comply with the copyleft-like freedoms?
Hi Christian,
thanks for your feedback and your questions. I think that of course we should add a passage addressing the re-user and obliging him/her to re-apply the trademark to any physical device deriving from the original source. We would bind that to the freedoms we are granting, I guess. In other words: I give you the 4 freedoms and it is my precondition (or: my strong wish?) that you re-apply the trademark to any device derinving from my work.
Is that a legal contract? I am not sure, but I am not a lawyer either and therefore I cannot say if there is a way to make that a bullet-proof legal agreement. We should try to find that out, though!
It is of course very good to have law as a backing, but I think that if we are not going for patents (which we won’t do, for good reasons – many of which discussed here already) we might always end up with some kind of blind spot.
On the other hand, law is just one constraint to put pressure on companies. And as we can tell from all exisiting product piracy, it is not perfect either.
In my belief we are talking about a field of products and markets which include a lot of ideological load. There are other questions which are at least as relevant as law itself. Having a quick look at Porter’s Five Forces helps to estimate their relevance, I think.
If companies/manufacturers will use the trademark, they will do it not only due to ideological reasons. They will of course also do it, because they think it will add one or more advantages to their products and to their organisation and/or brand.
We can try to collect all those, like e.g. better access to good developers, better selling, less time to market, less costs for R&D, better products, ideological glue towards the employees and customers, ethical bonus for the brand, and so on. Still this may remain a question of faith – but looking at FOSS we can have some hope, I guess 😉 (Besides digital hardware there is the whole issue of drugs and medical patents and other fields in the physical world where ethics may even have a much stronger impact.)
So turning all these points into a negative list is exactly what will happen to those companies who will fraud the trademark and copyleft idea – a lesson we have also learned from FOSS.
So even if there may not be a legal way to stop one from doing something we do not want them to do, there may at least be the possibility to harm them with the public voice and to blacklist them, if we think this is truely relevant.
The aim of the open hardware trademark is to shift the copyleft license as far towards hardware as we can get, but also to make the copyleft license and the idea behind it more visible on the devices itself and towards its markets.
We need to re-check the legal boundaries and see if we can implement something to impose an obligation to stick with the trademark for all derivations, but in my opinion we should not give up the trademark, even if some legal obsticals may remain.
To move on with opening up the world, the trademark is at least a good next step which I would like to take until we may have found a better solution. I’d be happy to get your feedback and to learn more about what you think we should do next!
Very kindly,
JuergeN
PS: If you produce organic[TM] eatables you can never know if they will also end up on a whole organic compost pile. Still there are many good reasons to produce them … plus: they sell!
@Juergen:
Hm, that seems to boil down again to the same old problem: a legal obligation must be based on copyright or some other law, and copyright doesn’t not apply to the user’s behavior which you want to influence. I guess you could try to make it legally binding through some kind of „shrink-wrap license“ instead („you are only allowed to download this, if you agree to the following conditions“), but „shrink-wrap licenses“ after problematic in various ways (e.g. they are probably ineffective in German law, and they would make free redistribution of the protected information difficult).
True, maybe it’s better to rely both on ethical pressure (this is what you SHOULD do) and practical motivations („I WANT to call my hardware <insert trademark name>“) instead of legal obligations.
Hm, I would think the obvious thing would be to try to resolve the open issues (such as: how it is meant to work exactly?) as good as possible and get the whole idea into a form where it can actually be used. Do you need help? Is there a public mailing list or similar where you coordinate your activities?
An interesting reading here. I think a „clickwrap“ license would do the job??? There is lot of talking about the licensing of open hardware projects, but isnt it actually a very trivial problem? If you want A, you have to accept B. As simple as EULA, which is legally binding, right?
@Kozuch: I mentioned in my comment to Juergen that a shrink-wrap (or click-wrap) license might work, but I think it would be very difficult to get something that is internationally watertight. For example, in German law, click-wrap licenses which the user accepts when s/he installs the software (rather then when they bought it) are not valid at all, and even when the user accepts at buy, there are many things that cannot be agreed on in a EULA at all (the respective clauses will be invalid and can be ignored – cf. Endbenutzer-Lizenzvertrag). How does this translate to free hardware where there is no “buying” at all? And would it be possible to define a copyleft-like click-wrap agreement that was valid in German law? I don’t know…
And that’s just one jurisdiction—imagine the difficulties of writing an agreement that holds up in the major North American and European jurisdictions at least. Even the big companies with their huge legal divisions fail at that tasks (their EULAs are regularly invalid in some parts)…
Another issue is the passing on of information—whoever spreads such “click-wrapped” free information, would have the burden of making the other party accept the click-wrap agreement as well. What if I spread the information ignoring this requirement? Clearly I have violated the license terms, but what about the person I gave it too? They don’t have violated anything, but they have got the information and can probably use it in any way they like (without being bound by the license which they didn’t accept).
I think there must be a way how to solve the problem. I can imagine there are thousands similar cases in business out there where lawyers are not supported by current laws and they have to deal with it – create special types of contracts and special ways of dealing with problems… finally, what they are doing is they are just playing with probabilities that certain things happen. If one thing happens more often than the other, they take a special care of that case and try to „protect“ it better in their works. There must be a workaround here too.
What about including a notice in every file so that it is necessarily visible to the user of that file? I mean include it within the drawings themselves and a kind of „footnote“, not as a text in a binary file, but rather as a vector text within the drawing itself… I mean not in some „hidden“ code comment but in an immediatelly visible way so that the „user“ (developer, contrubutor, coincidental viewer) of the file just sees it and cant ignore it. Texturing the surface of a 3D model with such a notice would be another way. I think this could have at least some legal force, couldnt it?
Finally, we could end up using multiple ways of protection at the same time. Do usual licences (GPL etc. for copyrightable stuff), do a click-wrap thing, do notices in files in case someone omits the click-wrap, do trademark at the same time etc. I think mixing the most possible protections together could enhance the force of all these „protections“ (I mean to protect some freedoms here of course, like GPL does for instance).
@Kozuch: Well, we’ll see if the GOSH initiative (cf. Juergen’s comments) will achieve something. I also know that the c,mm,n car project tried to work out a license based on contract law (license draft), but their efforts seem to have got stuck…
A trademark could be used to support open hardware by giving consumers a choice. If some well known „open hardware“ label would exist, people could choose to prefer bying this products.
But the definition of open hardware should be broadened.
Products which are „hacker friendly“ (for example well supported with documentation, provided with real open source drivers, support of home brewed firmware etc) but do not have an open design, should also get a (different) OH-logo attached to them.
Thanks for publishing your thoughts on this topic. I have been struggling with understanding how to approach this topic with my own designs/projects.
On one hand I have learned most of „hacking“ from other projects, dissecting designs, implementing circuits, „stealing“ modules and pieces – so I want to add to this body of knowledge. On the other hand, I don’t want someone else (especially a big company) to get rich of my work when I still need to pay off my student loans.
This is definitely a grey area… Ladyada and others make an interesting argument for removing the Non-Commercial restriction from your license. http://www.ladyada.net/library/openhardware/license.html
Some other interesting readings
http://www.instructables.com/community/Open-source-Hardware-License—creative-commons-li/
http://antipastohw.blogspot.com/2010/01/grand-unifying-theory-of-open-source.html
Michel Bauwens has published an interesting series discussing obstacles to and controversies about open source hardware development:
For what it’s worth, Hexayurt Project (http://hexayurt.com) is an Free housing project of substantial scale (we had over 500 units built at Burning Man last year, for a total spend of about $250,000 from people building our design.)
We’ve always gone public domain because none of the licensing models offered us any substantial protection. Even patent is, in our area (shelter) very unreliable because of the mass of unsearchable prior art from the 1960s shelter design explosion, and submarine patents.
Hope that’s helpful.