A (Final) Few Words on FRAND Licensing
Published 15:26, 20 October 10
The issue of Fair, Reasonable and Non-Discriminatory (FRAND) licensing has cropped up quite a few times in these pages. The last time I wrote about the subject was just last week, when I noted that the Business Software Alliance was worried that the imminent European Interoperability Framework (EIF) might actually require truly open standards, and so was pushing for FRAND instead.
Since then, the FSFE has obtained a copy of the BSA's letter [.pdf], and it is certainly a corker. For example, there is the following passage:
if the EU adopts a preference for royalty/patent-free specifications, this undermines the incentives that firms have to contribute leading-edge innovations to standardization - resulting in less innovative specifications, and less competitive European products.
First, note the clever blurring of the distinction between “royalty-free” and “patent-free”. No one is talking about patent-free, since that's plainly not possible with a benighted patent system that gives patents for trivial and obvious ideas at the drop of a digital hat. What is being discussed is “royalty-free” (strictly speaking, restriction-free, because there are other ways in which restrictions can be placed on the use of technologies for which patents are claimed.)
As for the idea that no one will ever contribute to a standard that requires RF (royalty/restriction free) licensing, the experience of the World Wide Web Consortium contradicts that, as I've pointed out before.
That's just a sample of the BSA approach, and how it doesn't stand up to detailed analysis. If you want more of the latter, the FSFE have carried out a full fisking, complete with detailed references, so I won't duplicate their work here.
What I do want to explore a little further is an issue that has come up in online discussions on Twitter with Florian Mueller. He's written a post that offers a good summary of what was said, although I beg to differ on the conclusions to be drawn from that conversation.
One of the key issues was whether concentrating on the Royalty-Free part of Restriction-Free was reasonable. One very pragmatic reason for talking about money is that the BSA used that as an argument against RF licences, so it seems only fair that supporters of RF should be able to rebut that. Here's what the BSA wrote in both their letter to the EC and a press release about it (which seems to have disappeared - luckily I kept a copy, which you can read in the comments of my BSA post):
FRAND licensing allows inventors to charge a reasonable fee when their technologies are incorporated into specifications provided the technology is made available on FRAND terms to any person who wants to implement the standard.
Again, as I and many others have pointed out, the BSA solution - “a reasonable fee” - is simply not compatible with leading free software licences. And let me be clear: by that I do not mean legally incompatible, I mean incompatible in practice, because it is not possible to account for the uncountable because unknown downstream users.
Which brings us on to the second point to arise out of the discussion with Florian Mueller. Here's what he writes in his blog post on the matter:
One example that was mentioned is that one can't impose patent royalties on downstream users. So if an open source developer publishes a piece of software under a FOSS license, the developer can't keep track of every copy of the software that gets distributed (since anyone downloading or in some other way receiving the software can pass it on to others, and so forth). That creates a problem for per-unit patent royalties.
But that doesn't mean that no royalties can be paid at all. Per-unit royalties are a question of responsibility, accountability, and collection/refinancing. For example, MPEG LA has an annual royalty cap for its AVS/H.264 video codec license. A company paying that fixed amount never has to worry about the number of downloads. Chances are that Red Hat also negotiated fixed amounts with the patent holders from which it has obtained licenses. And there must have been ways -- maybe other ways -- in which this problem was solved for all the other companies in the industry who pay patent royalties on GPLv2-based software.
Indeed, Red Hat's agreement with FireStar, analysed at length on Groklaw, is a good example of how patent royalties were paid (in some unspecified manner) on GPLv2 software. Groklaw's PJ calls this “historic”, and Florian Mueller sees it as proof that FRAND licensing can be made to work with free software licences like the GNU GPL, but I see two major problems with it, both mentioned in this part of his blog post on the issue:
A few weeks ago Glyn Moody wrote that FRAND licensing isn't non-discriminatory because, as he said, it discriminates against FOSS. This was also part of today's Twittersation. After the point on per-unit royalties was made, I asked him why he then (instead of continuing to oppose FRAND as a whole) doesn't just demand that there must be a possibility for "FRAND lump sums". Glyn replied:
“because that would allow companies to demand huge lump sums that were impossible to pay, citing millions of downstream users”
He immediately added:
“it would also be impossible for projects that had no company behind them, leading to a two-class system”
This is a perfect example of calling something "discriminatory" that actually isn't. Glyn is right that licensing patents is much more difficult for smaller companies, let alone projects without companies behind them, than for larger ones. But that isn't open-source-specific. A small company writing closed-source proprietary software and distributing it free of charge (as part of a "free" or "freemium" business strategy) would face the same issue. Discrimination as a term only applies if a group is uniquely disadvantaged. The advantages of large, deep-pocket companies over smaller, sparsely-funded ones, or of any kind of company over non-company software development projects, are manifold. Patent licensing is just one area in which large organizations benefit from size, and again, you don't have to be open source to face those challenges. It's just that you're small, or that you're not a company.;
So, to expand a little on my objections, imagine that FRAND were permissible in the new version of the EIF. First, there is no guarantee that those claiming patents in standards would be prepared to waive per-unit pricing in favour of a lump sum: if the EIF says FRAND, then they might insist on the latter. Suppose now that the EIF specified that a lump-sum option had to be offered; sure, the companies will say, that will 100 million Euros, thank you very much.
Yes, you can start inserting “reasonable” and the rest, but that doesn't address the problem - it's just going to lead to even more lawsuits, and I think we have enough of those in the world of patent law. And going beyond that to a detailed spelling out of what is and what isn't "reasonable" is simply not the level of detail that the EIF works at: it operates with very straightforward basic principles: FRAND or RF, not some complicated legalistic mixture of the two.
So, on the issue of discrimination, we are talking about vanilla FRAND, not some hand-crafted version, and that clearly does discriminate against free software. But for the sake of argument, let us assume that lump sums were available, in which case I return to my comments about a “two-class system” quoted above. That is, it is easy to imagine some small, perhaps unknown free software project that would like to implement a standard for which lump-sum FRAND is available, but without a bigger company willing to stump up the money, it is still locked out.
But, once more, leaving aside that problem, I do fully accept the point that
advantages of large, deep-pocket companies over smaller, sparsely-funded ones, or of any kind of company over non-company software development projects, are manifold
Smaller companies are indeed at a disadvantage in this situation when they cannot afford to pay licences as easily as big companies. And yet it is precisely such smaller companies that the European Commission wishes to prioritise:
Adopted in June 2008, the "Small Business Act" for Europe reflects the Commission's political will to recognise the central role of SMEs in the EU economy and for the first time puts into place a comprehensive SME policy framework for the EU and its Member States.
It aims to improve the overall approach to entrepreneurship, to irreversibly anchor the "Think Small first" principle in policy making from regulation to public service, and to promote SMEs' growth by helping them tackle the remaining problems which hamper their development.
So putting this group at a disadvantage, which Florian Mueller himself admits the FRAND policy would indeed bring about, is surely the last thing the European Commission will want to do in its EIF. Indeed, one of the key reasons for introducing the EIF is to create a level playing field so that more companies can bid for contracts - including smaller ones ("Think Small First"). Discriminating in favour of larger ones, as Florian Mueller himself admits would be the case, is entirely at odds with the intent of the whole document. So it turns out that this line of argument gives yet another reason to avoid FRAND licensing.
To summarise: general FRAND is not compatible with the practical realities of free software. Hand-crafted FRAND variants are unlikely to be adopted in a generic interoperability framework, and if they were, they would discriminate not just against free software, but against smaller companies too - promoting whose growth is one of the European Commission's avowed priorities. Once again, then, it is clear that the only fair and practical solution for open standards is RF: Restriction-Free licensing.