Richard Stallman on Software Patents - UNSW October 2004
But my intention here is to summarise some of the key points of the talk so that I don't forget them and perhaps so others may benefit. The good ideas presented are Stallman's (I've paraphrased mercilessly, and embellished in spots - but I'm sure he'd still recognize his own ideas here) - errors are probably mine.
Stallman opened by attacking the term "Intellectual Property". There's an obvious bias in the term as it implies that an idea is like a croissant - something you can own, and that if you own, noone else can have the use of without taking something from you. But he did not attack it on these grounds - instead he pointed out that it's a somewhat dishonest term as it is generally used to blur (at least) two extremely different things - copyright and patents. [I attended a talk he gave on copyright last week, but will say no more on it here].
He said that copyright and patents were in fact so different that if you assumed something that was true about one was false about the other, then you would usually be right. Perhaps the most important difference is illustrated as follows:
Thus, the patent is seen to protect the little backyard inventors. [And who cannot sympathise with them. Just as most of us imagine we have a great novel in us, so we imagine we probably have a great invention in us as well, and the last thing we'd want is those bastards from Giant Pin Corp dudding us out of it. In reality great works pretty much always require lots and lots and lots of mediocre and even poor works to precede them - just like a great piano performance. Writing and Inventing merely look easy - in fact they're as hard or harder than most other jobs].
Stallman painted a picture very different from the fantasy above. In part:
But let's imagine he is a good businessman and is successful. Giant Pin Corp sees his pin and decides to copy it. So he knocks on their door and says "Hey, I've got this patent!". Giant Pin Corp then says "Yes, but we have patents covering 'curly springs', 'bendy clips', and 'the sharpening of a single end of a contorted wire to enhance its fabric piercing ability', all of which you've infringed with your invention. So here's what we'll do - we'll cross-license our technologies to each other, so that we both may continue to build and sell these new pins, or we'll put you out of business."
In an alternative scenario, Giant Pin Corp has no interest in manufacturing the new pin, but comes knocking on the inventor's door with their patents demanding license fees.
Stallman likened the patent system to a lottery. In particular, both have a very high possible payout, and it's this that is used to entice people, but the expected payout from either is extremely low (a negative return on investment).
Another chance element arises - why did the inventor not know about Giant Pin Corp's patents? The problem here is that the inventor must do the almost impossible - find which of the tens or hundreds of thousands of patents his invention might impinge upon. Some may be easy to find, but many are quite obscure - perhaps the 'curly spring' patent is to be found in 'a method for opening a dolls eyelid'. How would you know?
This is exacerbated by the fact that patents are deliberately written as broadly as possible and in an almost impenetrable legalese. Stallman gave an example of a patent holder who sued for an infringement he hadn't even realised his patent covered until his lawyer told him. If a patent holder doesn't know what his patent covers, how is anyone else supposed to?
Part of the problem for the new pin inventor is that his invention was the expression of more than a single idea, and some of the ideas did not actually belong to him, even if he happened to think of them independently.
Stallman put forward a continuum of number of ideas expressed in a product. Pharmaceuticals fit mostly at the "1" end of the continuum - each one stands on its own, and at least until recently each was covered by a single patent. Thus if you invented a drug that hadn't been invented before, there was essentially zero chance that you had infringed any patents.
Mechanical goods fall a little way further up the spectrum, as the safety pin example attempts to demonstrate. A large part of the problem in creating a physical product is that you need to deal with all the problems of real world materials, and you also need to determine how it will be manufactured. Often the problem of designing the machinery to produce the product dwarfs the problem of designing the product in the first place.
Software has neither of these problems. The statements that make up the software behave absolutely reliably - they do not warp, crack, suffer from heat - Stallman gave a number of amusing examples - and the manufacturing effort is zero - you simply make a copy of the program using a 'copy' command.
Since software engineers are roughly as clever as other kinds of engineers, and spend roughly the same sorts of time on their products, this freedom from manufacturing and physical constraints translates into a lot more ideas going into the product. Thus, a software product falls at the opposite end of the continuum from pharmaceuticals - a software product may contain hundreds or thousands of patentable ideas.
The problem is of course that others have probably patented most of them.
As an example, consider that a patent exists for the feature in a word processor that replaces an abbreviation with the full word as you type. And now consider how many other features your word processor contains, each patentable. It's rapidly clear that there's not much room for competition in the field when obvious ideas are unusable.
Big companies like IBM and Microsoft and Sun each hold vast numbers of patents. Since it's essentially impossible to build a piece of software without infringing on some of these patents, they can easily be used against any small company who attempts to compete. They have been used in the past to obstruct or prevent free software projects, and will certainly be used again.
And yet there is some competition (not much when it comes to word processors I'd have to say though). This is because the big companies, knowing it's impossible not to infringe each other's patents, cross-license their patents. Thus Sun allows IBM to use it's patents and vice versa.
There's a wildcard though - the patent parasites. These are companies that have got hold of a patent and have no wish to make use of it, only to sue. Only recently Sun settled such a case with Kodak out of court for $100 million. Large companies generally consider this simply a cost of business, but for smaller companies it can be sink or swim.
Stallman referred to an IBM article [Think magazine, issue #5, 1990] that stated that while the revenue IBM get from licensing their patents is beneficial, the real benefit ("by an order of magnitude") comes from the protection the patents give them from the patent suits of others - any software producer that might true to sue IBM for infringement will almost certainly be infringing some of IBM's patents, so a cross licensing deal can be established.
This, Stallman asserts, shows that the cost of patents is ten times (an order of magnitude) the benefit, but that large companies with a sufficiently broad portfolio can avoid the cost and simply reap the benefit, while the little guy almost certainly pays the cost for no benefit.
The winners of course are the lawyers - they get paid to prepare the patents, and they get paid in the patent battles regardless of which side wins. Stallman pointed out that although patent lawyers encourage patents in other fields, they've managed to avoid patents in their own field - you can't patent a line of reasoning or a way of addressing the judge, or a method for charging clients....
The purpose of the patent system is to encourage progress. That it fails utterly in its purpose when applied to software, if not already obvious, should be clearer with Stallman's following example:
Now imagine Beethoven sits down to write his first symphony in 1800. Many useful musical ideas will be unavailable to him. Should Haydn turn to Beethoven and say "can you not come up with your own ideas man?"
To lock up ideas is to halt progress.
LinksPhilosophy of the GNU project - Richard Stallman is the founder of the GNU project, and this page links to a number of articles on Free Software, Copyright, Software Patents and more. You can find transcripts and audio recordings of speeches given by Stallman, including the one below.
The Danger of Software Patents - a transcript of a speech given in India in 2001, but essentially the same as the one given at UNSW.
Research On Innovation - Richard Stallman referred to this website in his speech - they have a number of research papers on (amongst other things) the effects of software patents on innovation.
Patently Absurd (Forbes article) - gives a real-life anecdote showing how patents are used as a cash-generating weapon. "Stick 'em up, or I'll fire a patent at you".
(*)"New useful and unobvious" - this is supposedly the criteria for deciding whether a patent should be granted. "Unobvious" is somewhat controversial however, as the criteria for deciding whether an ideas ia unobvious result in "unobvious to someone with an IQ of 80" - hence many patents are granted which engineers would consider exceedingly obvious.