Talk by Richard M. Stallman
The Danger of Software Patents
In the 1980's the Australian Government commissioned a study of the patent system. The patent system in general, not software patents. This study concluded that Australia would be better off abolishing the patent system because it did very little good for society and cause a lot of trouble. The only reason they didn't recommend is that international pressure. So one of the things they cited was that patents which was supposed to disclose information so that they would no longer be secret or in fact useless, for that purpose, engineers never looked at patents to try and learn anything because it's too hard to read them. In fact they quoted that an engineer saying "I can't recognize my own inventions in patents".
Few years ago an engineer in US named Paul Heckel was sueing Apple. He had a couple of software patents in the late 80's for a software package and then when he saw hypercards and looked at inside - this is nothing like my program. He didn't think any more of it. But later on his lawyer explained to him that if you read this patent carefully hypercards fell into the prohibitted area. So he sued Apple feeling this is an opportunity to get some money. Well once when I give a speech like this, he was in the audience, and he said "oh no that's not true. I just wasn't aware of the scope of my protection" and I said "yeah, that's what I said".
So you are going to have to spend a lot of time working with a lawyer and expaining to the lawyer what projects you are working on, so that the lawyer can explain you what are the patents employed. This is going to be expensive when you're done the lawyer will tell you something like this : "If you do something in this area you are almost sure to loose a lawsuit. If you do something in this area you are in a substantial danger and if you really want to be safe you better stay out of this area and of course there is a substantial element of chance in the outcome of any lawsuit." So now that you have a predictable terrain for doing business what are you going to do .
Well, you have three options to consider - you can try to avoid the patent, you can try to licence the patent, or you can try to challenge its validity in court. Any one these three is sometimes a viable alternative and sometimes not.
First let's consider avoiding the patent. Well in some cases that's is easy. You know Unisys was frightening people using the patent on LZW compression which is have to find another data compression algorithm or we can avoid that patent. Well, that was somewhat difficult because there are many other patents covering lots of other data compression algorithms. But eventually we found one that was not in the area that those others' patents cover, eventually we did. So that program was implemented. It actually gave better compression results and so we now have GZIP and a lot of people use GZIP. So in that one case it was, there was considerable work and we were able to do it to avoid that patent.
But in the 80's Compuserv defined an image format called GIF and used LCW compression algorithm in defining it. Well, of course once the uproar about the patent become known, people defined another image format using a different compression algorithm. They used GZIP algorithm and that format is called the PNG format which I suppose to mean that PNG is Not GIF.
But there was a problem: lots of people have already started using GIF format and there were many programs that could display GIF format and produce GIF format but they couldn't display PNG format. So the result was people felt it too hard to switch. You see when you are dealing with a data compression program used by somebody who says "I want to compress some data or you can give them a data compression program". If he can get sued for using this one and you give him another one he will switch.
But what you want to do is make images that can be displayed by Netscape, then he can't switch unless Netscape handles the other format, and it didn't . It took years, I think, before Netscape is started to handle PNG format. So people essentially said "I can't switch, I just have't" and so the result was the society had invested so much in this one format but the inertia was too great for a switch even now there was another superior format available.
So even when a patent is rather narrow, avoiding it can be very hard. Postscript specifications includes LZW compression which we and our implementation of a postscript cannot implement. We supported another kind of compression in some sense that's not correct even though it does the useful job. So even a narrow patent is always feasible to avoid.
Now, some times a feature get patented. In that case, you can avoid the patents by taking out that feature. In the late 80's the users of the word processor ZIRITE got a downgrade in mail. That word processor had a feature where you could define a short word or sequence as an abbreviation. Whenever you type in that short sequence and then a space it would turn into a longer expansion. You could define this anywhere you write. Then somebody patented this and ZIRITE decided to deal with the patent by removing the feature. They contacted me because in fact I had put a feature like that into the original Emacs editor back in the 70's - many years before that patent. So there was a chance that I could provide evidence would enable them to fight the patent.
Well, this at least showed me that I had at least one patentable idea in my life. I know of it because someone else patented it. Now, of course you can respond to these patented features by taking the features out. Once your program starts being missing several features that users want, it might be useless as a program.
Now you may have heard of Adobe Photoshop. We have a program called the GIMP which is more powerful and general than Photoshop. But there is a one important feature that it doesn't have which is pantom colour matching. Which is very important for people who want actually print the images on paper and get reliable results. This feature is omitted because it is patented. And as a result is that the program for one substantial class of users is crippled. If you look at programs today you see that they are often provide many features and what the users demand is features. If any important feature is missing, well, it is easy to leave it out. But the results may be very bad .
Of course, sometimes a patent is so broad that is impossible to avoid it. Public key encryption is essential for computer users to have privacy. The whole field was patented. That patent expired just four years ago, so they could dream now Free Software in the US for public key encryption. Until then many programs free and non-free were wiped out by the Patent Office. And in fact that the whole area of computing was held back for more than a decade despite strong interest. So, that is the possibility of avoiding the patents.
Another possibility that is sometimes available is to licence the patent. Now the patent holder is not required to offer you licence that's his. The patent holder can say "I am not licensing this, you are just out of business. Period".
In the League for Programming Freedom we heard in the early 90's from somebody whose family business was making casino games, computerised of course, and he had then threatened by somebody who have a patent on a very broad category of computerised casino games. The patent covered a network where there is more than one machine and each machine support more than one type and they can display more than one game in progress at time. Now one thing you should realise is the patent office thinks that is really brilliant. If you see that other people implemented doing one thing and you decided to support two or more. You know if they made a system that plays one game and if you make it able to play more than one game that's an invention. If you can display one game and you decided to setup so that display two games at once that's an invention .
If he get with one computer and you do it with network having notebook computers thats an invention for them. They think that these steps are really brilliant. Of course we in Computer Science know this is just a rule that we can generalise anything from one to more than one. So most obvious principle there is. So every time you write a subroutine that's what you're doing. So this is one of the systematic reasons why the patent system produces and then oppose patents that we would all say are ridiculously obvious. You can presume just because it's ridiculously obvious that they wouldn't be upheld by a court. They may be legally valid despite the fact that are stupid. So he was faced with this patent and the patent holder was not even offering him the chance to get a licence. "Shutdown!" - that's what the patent holder said, and that's what he eventually did. He couldn't afford to fight it.
However many patent holders will offer you chance of a licence. But it cost you dearly. The owners of the natural order recalculation patent would be demanding five percent of the gross sales of every spreadsheet and that I was told was the cheap pre-lawsuit price. If you insisted on fighting over the matter they would be charging more. Now you could, I suppose, sign a licence like that for one patent, you could do it for two, you could do it for three. For what I think twenty different patents for your program and each patent holder wants five percent of the gross sales. What I have said twenty - one of them disagrees, then you are pretty badly screwed.
But actually business people tell you that two or three such patents would be such a big burden that they would make the company fail. In practice, even if in theory, it might have the chance. So a licence for a patent is not necessarily a feasible thing to do and for us, the Free Software developers were in the worst position because we can't even count the copies and most licences domain in the field for copying so that's absolutely impossible for us to use one of these licences. You know, that if a licence charged one millionth part of a rupee for each copy, we would be unable to comply because we can't count the copies. The total amount of money that I might have in my pocket, but I can't count it so I can't pay it. So we suffers some special burden occasionally.
But there are one kind of organisations for which licensing patents works very well and that is the large multinational corporations and the reason is that they own many patents themselves and they use them towards cross licensing. What does that mean? Well, essentially the only defence against patents is deterrence, you have to have patents of your own. Then you hopes that somebody points patented you, you will be able point patent back and say "don't sue me, because I'll sue you".
However, deterrence doesn't work as well for patents as it does with nuclear weapons and the reason is that each patent is pointed in a fixed direction. It prohibits certain specified activities. So the result is that most of the companies that are trying to get some patents to defend themselves with. They have more chance of making a success. They might get a few patents that might get a patent that points there and they might get a patent that points there. And then if somebody over here threatens this company what are they going to do? They have a patent pointing over there, so they have no defence.
Meanwhile, sooner or later, somebody else who wander over there and the executive of the company will think "gee, we're not as profitable as I would like, why don't I like to squeeze some money out of them." So they say first to getting this patents for defensive purposes. But they often change their minds later when a victim wants to buy. And this by the way the reason why the fallacy in the myth that the patent system "protect" the "small inventor".
Let me tell the message in the myth of the starving genius. If somebody who's been working in isolation for years and starving and he has a brilliant new idea for how to do something or other. And so, now, he's starting a company and he is afraid some big companies like IBM will compete with him and so he gets a patent and this patent will "protect him". Well, of course, this is not the way of things work out in fields. People won't make this kind of progress in isolation is where you working with other people and talking with other people and developing software usually. And so the whole scenario doesn't make sense and besides, if there's such a good computer scientist there is no need for him to starve. He could have got a job at any time if he wanted.
But let's suppose this happened, and suppose he has this patent, and he says "IBM, you can't compete with me because I have got this patent ". But here is what IBM says: "well, gee, let's look at your product, hmm, I have this patent, this patent and this patent and this patent and this patent that your patent is violating. So how about a quick cross-licence?". And the starving genius says "hmm, I haven't had enough food in my belly to fight these things, so I better give in" and so they sign a cross-licence and now guess what .
IBM can compete with him - he wasn't protected at all. Now IBM can do this because they have a lot of patents. They have patents pointing here, here, here, everywhere. So that anybody from almost anywhere that attacks IBM is facing a stand-off. A small company can't do it but a big company can.
So IBM brought an artcle it was in Stink magazine, I believe issue number five of 1990, that's IBM's own magazine - an article about IBM's patent portfolio. IBM said that they have two kinds of benefits form it's 9000 active U.S patents. One benefit was collecting royalties from licences. But the other benefit, the bigger benefit, was access to things patented by others. From mission to not to be attacked by others but with their patents through cross licensing. And the article said that the second benefit was an order of magnitude greater than the first.
In other words, the benefit of IBM is to make it things freely, not being
sued, was ten times the benefit of collecting money from all their patents.
Now the patent system is lot like a lottery, in that the what happens with
any given patents is largely around them and most of them don't get bring
benefits of their owners. But IBM is so big that these things average out
over the scale of IBM. So you can say IBM is measuring what the average
is like. What we see is, and this is a little bit sudden, the benefit of
IBM of being able to make use of ideas patented by others is equal to the
harm that the patent system would have done to IBM if there were no cross