In the second part of Richard Stallman's speech at the Cambridge Computer Lab, the free software guru explains how software developers can avoid software patents.
Tackling software patents I: Avoidance
Stallman: In some cases, a feature is patented. Then you avoid the patent by not implementing that feature. Then it just matters how important is that feature.
In some cases, you can live without it. A while ago, the users of the word processor XyWrite got a downgrade in the mail. The downgrade removed a feature which allowed you to pre-define abbreviations. That [is,] when you typed an abbreviation followed by a punctuation character, it would immediately replace itself with by some expansion, so that you could define the abbreviation for some long phrase, type the abbreviation then the phrase will be in your document. They wrote to me about this because they knew the Emacs editor has a similar feature. In fact, it had it since the 70's. This was interesting as it showed me that I had at least one patentable idea in my life. I knew it was patentable because somebody else patented it afterward!
Actually they tried these various approaches. First they tried negotiating with the patent holder, who turned out not to negotiate in good faith. Then they looked at whether they could have a chance at overturning the patent. What they decided to do was to take out the feature. You can live without this feature. If the word processor lacks only this feature, maybe people will still use it. But as various features start getting hit, eventually you end up with a program people think is not very good and they are likely to reject it.
That is a rather narrow patent on a very specific feature. What do you do with the British Telecom patent on traversing hyperlinks together with dial-up access? Traversing hyperlinks is absolutely essential to a major use of computers these days. Dial up access is also essential. How do you do without this feature, which, by the way, isn't even one feature, it is really a combination of two just arbitrarily juxtaposed? It is rather like having a patent on a sofa and television in the same room.
Sometimes the idea that's patented will be so broad and basic that it basically rules out an entire field. For instance, the idea of Public Key Encryption which was patented in the US The patent expired in 1997. Until then, it largely blocked the use of Public Key Encryption in the US A number of programs which people started to develop got crushed -- they were never really available because the patent holders threatened them. Then, one program got away, the program PGP which initially was released as free software. Apparently, the patent holders, by the time they got around to attacking, realised they might get too much bad publicity. So they imposed restrictions, making it for non-commercial use only, which meant it couldn't catch on too much. They greatly limited the use of Public Key Encryption for a decade or more. There was no way around that patent. There was nothing else you could do like that.
Sometimes a specific algorithm gets patented. For instance, there is a patent on an optimised version of the Fast Fourier Transform. It runs about twice as fast. You can avoid that by using an ordinary FFT in your program. That part of the program will take twice as long. Maybe that doesn't matter, maybe that is a small part of the program's running time. Maybe if it is twice as slow, you won't really notice. Or maybe your program won't run at all as it will take twice real time to do it's job. The effects vary.
In some cases, you can find a better algorithm. This may or may not do you any good. Because we couldn't use compress, we started looking for an alternative compression algorithm. Somebody wrote to us saying he had one; he had written a program and decided to contribute it to us. We were going to release it. Just by chance, I happened to see a copy of the New York Times, it happened to have the weekly patent column in it. I didn't see a copy of the Times more than once every few months. So I looked at it and it said someone had got a patent for "Inventing a new method of compressing data". I figured I better take a look at this patent. I got a copy and it turned out to cover the program that we were just a week away from releasing. That program died before it was born.
Later on we did find another algorithm which was unpatented. That became the program gzip, which is now effectively the de-facto standard for data compression. As an algorithm to use in a program for data compression, it was fine. Anyone who wanted to do data compression could use gzip instead of compress.
The same LZW compression algorithm was also used in image formats such as the GIF format. But there, because the job people wanted to do was not to just compress data but to make an image that people could display with their software, it turned out extremely hard to switch over to a different algorithm. We have not been able to do it in 10 years! Yes, people used the gzip algorithm to define another image format, once people started to be threatened with law suits for using GIF files. When we started saying to people stop using GIF files, switch over to this, people said "We can't switch. The browsers don't support the new format yet". The browser developers said "We're not in a hurry about this. After all, nobody is using this new file format".
In effect, society had so much inertia in the use of the GIF format, we have not been able to get people to switch. Essentially, the community's use of the GIF format is still pushing sites into using GIF format with the result that they are vulnerable to these threats.
In fact, the situation is even more bizarre. There are in fact two patents covering the LZW compression algorithm. The patent office couldn't even tell they were issuing two patents on the same thing, they couldn't keep track. There is a reason for this: it takes a while of study of these two patents to see that they really cover the same thing.
If they were patents on some chemical process, it would be much easier. You could see what substances were being used, what the inputs were, what the outputs were, which physical actions are being taken. No matter how they are described, you'd see what they were and then you would see they are similar. If something is purely mathematical, there are many ways of describing it, which are a lot more different. They are not superficially similar. You have to really understand them to see they are really talking about the same thing. The patent office doesn't have time. The U.S patent office as of a few years ago was spending on average 17 hours per patent. This is not long enough to think carefully about them, so of course they make mistakes like that. In fact, I told you about the program that died before it was born. That algorithm also had two programs issued for it in the US Apparently, it is not that unusual.
Avoiding the patents may be easy, may be impossible. It may be easy but it makes your program useless. It varies depending on the situation.
Here is another point I should mention: Sometimes a company or consortium can make a format or protocol the de-facto standard. Then, if that format or protocol is patented, that is a real disaster for you. There are even official standards that are restricted by patents.
There was a big political uproar last September when the World Wide Web consortium was proposing to start adopting standards that were covered by patents. The community objected so they reversed themselves. They went back to insisting that any patents had to be freely implementable by anyone and that the standards had to be free for anyone to implement. That is an interesting victory. I think that was the first time any standards body has made that decision. It is normal for standards bodies to be willing to put something in a standard which is restricted by patents and people are not allowed to go ahead and implement it freely. We need to go to other standards bodies and call on them to change their rules.
Tackling software patents II: Licensing
Licensing the patent. The second possibility instead of avoiding the patent is to get a license for the patent. This is not necessarily an option. The patent holder does not have to offer you a license, it is not required. 10 years ago, the League for Programming Freedom got a letter asking for help from somebody whose family business was making gambling machinery for casinos, and they used computers back then. He received a threat from another company that said, "We have a patent. You are not allowed to make these things. Shut down!"
I looked at this patent. It covered having a number of computers on a network for playing games such that each computer supported more than one game and allowed you to play more than one game at a time.
You will find patent office really thinks there is something brilliant about doing more than one of anything. They don't realise that in computer science, that's the most obvious way to generalise anything. You did it once, so now you can do it any number of times, you can make a subroutine. They think that if you do anything more than once, that somehow means you are brilliant and that nobody can possibly argue with you and that you have the right to boss them around.
Anyway, he was not offered a licence. He had to shut down. He couldn't even afford really to go to court. I would say that particular patent was an obvious idea. It is possible that a judge might have agreed, but we will never know because he could not afford to go to court.
However, a lot of patent holders do offer licenses. They often charge a lot of money for that though. The company licensing the natural order recalculation patent was demanding 5 percent of the gross sales of every spreadsheet in the US I am told that was the cheap pre-lawsuit price -- if you actually made them sue you and they won, they'd demand more. You might be able to afford that 5 percent for licensing this one patent, but what if you need to license 20 different patents to make the program? Then all the money you take in goes on patents. What if you need to license 21 patents?
People in business told me that practically speaking, two or three of them would make any business unfeasible.
There is a situation where licensing patents is a very good solution. That is if you are a multinational mega-corporation. Because these companies own a lot of patents, and they cross-license with each other. That way, they escape most of the harm that the patent system does and they only get the good.
IBM published an article in Think magazine -- I believe it was issue No. 5 of 1990 -- on IBM's patent portfolio, which said that IBM got two kinds of benefit from it's 9000 US patents. (I believe the number is larger today.) These were, first, collecting royalties and second, getting "access to the patents of others." They said that the latter benefit is an order of magnitude greater. So the benefit that IBM got from being allowed to use the ideas that were patented by others was 10 times the direct benefit IBM could get from licensing patents.
What does this really mean? What is the benefit that IBM gets from this "access to the patents of others"? It is basically the benefit of being excused from the trouble that the patent system can cause you. The patent system is like a lottery. What happens with any given patent could be nothing, could be a windfall for some patent holder or a disaster for someone else. But IBM being so big [that], for them, it averages out. They get to measure the average harm and good of the patent system. For them, the trouble of the patent system would have been 10 times the good.
I say "would have been" because IBM through cross-licensing avoids experiencing that trouble. That trouble is only potential, it doesn't really happen to them. But when they measure the benefits of avoiding that trouble, they estimate it as 10 times the value of the money they collect from their patents.
This phenomenon of cross-licensing refutes a common myth, the myth of the "starving genius", the myth that patents "protect" the "small inventor". (Those terms are propaganda terms. You shouldn't use them.)
The scenario is like this: Suppose there is a "brilliant" designer of whatever. Suppose he has spent "years" starving in the attic designing a new wonderful kind of whatever, and now wants to manufacture it, and isn't it a shame the big companies are going to go into competition with him, take away all the business, and he'll "starve".
I have to point out that people in high tech fields are not generally working on their own and that ideas don't come in a vacuum, they are based on ideas of others; and these people have pretty good chances of getting a job if they need to these days. So this scenario, the idea that a brilliant idea came from this brilliant [person] working alone is unrealistic, and the idea that he is in danger of starving is unrealistic. But it is conceivable that somebody could have an idea and this idea along with 100 or 200 other ideas can be the basis of making some kind of product, and that big companies might want to compete with him.
So let's see what happens if he tries to use a patent to stop them. He says "Oh No, IBM. You cannot compete with me. I've got this patent." IBM says, "Let's see. Let's look at your product. Hmmm. I've got this patent and this one and this one and this one and this one and this one, which parts of your product infringe. If you think you can fight against all of them in court, I will just go back and find some more. So, why don't you cross license with me?" And then this brilliant small inventor says "Well, OK, I'll cross license". So he can go back and make these wonderful whatever it is, but so can IBM.
IBM gets access to his patent and gets the right to compete with him, which means this patent didn't "protect" him at all. The patent system doesn't really do that.
The mega-corporations avoid the harm of the patent system; they see mainly the good side. That is why they want software patents: they are the ones who will benefit from it. But if you are a small inventor or work for a small company, the small company will not be able to do this. Small companies cannot get enough patents to do this.
Any given patent is pointing in a certain direction. So if a small company has patents pointing there, there and there and somebody over there points a patent at them and says give me your money, they are helpless. IBM can do it [retaliate] because with 9000 patents, they are pointing everywhere; no matter where you are, there is probably an IBM patent pointing at you. So IBM can almost always make you cross license. Small companies can only occasionally make someone cross-license. They will say they want patents for defensive purposes but they won't get enough to be able to defend themselves.
There are cases where even IBM cannot make someone cross-license. That is when there is a company whose sole business is taking a patent and squeezing money out of people. The company that had the natural order recalculation patent was exactly such a company. Their sole business was to threaten to sue people and collect money from people who were really developing something.
There are no patents on legal procedures. I guess the lawyers understand what a pain it would be to have to deal with the patent system themselves. The result is that there is no way to get a patent to make that company cross license with you. So they go around squeezing everyone. But I guess companies like IBM figure that is part of the price of doing business so they can live with it.
So that is the possibility of licensing a patent, which may or may not be possible, and you may or may not be able to afford it
Tackling software patents III: Litigation
Supposedly, in order to be patented, something has to be new, useful and unobvious. That is the language used in the US; I think other countries have other language which is pretty much equivalent to it. Of course, when the patent office gets into the game, they start interpreting "new" and "unobvious". "New" turns out to mean "We don't have it in our files", and "unobvious" tends to mean "unobvious to someone with an I.Q of 50".
Somebody who studies most of the software patents issued in the US, or at least he used to, I don't know if he can still keep up with them, said 90 percent of them wouldn't have passed the "Crystal City test", which meant if the people in the patent office went outside to the news stand and got some computer magazines, they would see that these ideas are already known.
The patent office does things that are so obviously foolish, you wouldn't even have to know the state of the art to see they are foolish. This is not limited to software. I once saw the famous mouse patent which was obtained after Harvard genetically engineered a mouse with a cancer-causing gene. The cancer-causing gene was already known, and was inserted using known techniques into an already existing strain of mouse. The patent they got covered inserting any cancer-causing gene into any kind of mammal using any method whatsoever. You don't have to know anything about genetic engineering to realise that is ridiculous. I am told that this "overclaiming" is normal practice, and that the US patent office sometimes invites patent applicants to make their claims broader -- basically, make the claims broader until you think they are running into something else which that's unambiguous prior art. See how much land grab in mental space you can get away with.
When programmers look at a lot of software patents, they say "this is ridiculously obvious!" Patent bureaucrats have all sorts of excuses to justify ignoring what programmers think. They say "Oh! but you have to consider it in terms of the way things were 10 or 20 years ago." Then they discovered that if they talk something to death then you can eventually lose your bearings. Anything can look unobvious if you tear it apart enough, analyse it enough. You simply lose all standard of obviousness, or at least lose the ability to justify any standard of obvious or unobvious. Then, of course, they describe the patent holders as brilliant inventors, all of them, therefore we can't question their entitlement to power over what we do. If you go to court, the judges are likely to be a little more stringent about what is obvious or not. But the problem is that it costs millions of dollars to do that.
I heard of one patent case, the defendant I remember was Qualcomm, and I believe the ruling was ultimately $13m of which most went to pay the lawyers on both sides. There were a few million dollars left over for the plaintiff (because they [Qualcomm] lost).
To a large extent, the question of the validity of a patent will depend on historical accidents. Lots of historical accidents such as precisely what was published when, and which of those things somebody manages to find. Which of them didn't get lost, precise dates and so-on. Many historical accidents determine whether a patent is valid.
In fact, it is a weird thing that the British Telecom following hyperlinks together with telephone access patent was applied for in 1975. I think it was in 1974 that I developed the Info package for the first time. The Info package allows you to traverse hyperlinks, and people did use telephones to dial up and access the system. So in fact, I did produce a piece of prior art for this patent. This is the second piece of prior art I have produced in my life.
I didn't think this was interesting enough to publish it. After all, the idea of following hyperlinks I got from the demo of Englebart's editor. He is the one who had an idea which was interesting to publish. What I done I called "poor man's hypertext", as I had to implement it in the context of TECO. It was not as powerful as his hypertext, but it was at least useful for browsing documentation, which it all it was meant for, and as for there being dial-up access to the system, well, there was, but it didn't occur to me that the one had anything particular to do with the other. I wasn't going to publish a paper saying, "Oh! I implemented this poor man's hypertext, and guess what! There are dial-up lines on the computer too!" I suspect there is no way to tell precisely on what date I implemented this.
And was it published in any sense? Well, we invited guests to come in across the ARPAnet, and log in on our machine, so they could have browsed documentation using Info and seen the thing. If they had asked us, they would have found we had dial-up access. As you can see, historical accident determines whether you have prior art.
Now of course, there is a publication made by Englebart about hypertext, which they [the defendants] are going to show. I don't think it says anything about having dial-ups on the computer however, so whether it will suffice is not clear.
So, this is an option, the possibility of going to court to overturn the patent. Because of the expense, it is often out of the question even if you can find solid prior art which ought to be sufficient to overturn the patent. As a result, an invalid patent, a patent which nominally shouldn't have existed (but in fact lots and lots of them do) is a dangerous weapon. If someone attacks you with an invalid patent, that can really cause a lot of trouble for you. You might be able to bluff them away by showing them the prior art. It depends on whether they can get scared off that way. They might think, "Well, you are just bluffing, we figure you can't really go to court, you can't afford it so we'll sue you anyway".
All of these three possibilities are things that sometimes you can manage to use, but often you can't. So you have to face patent after patent after patent. Each time you may be able to find one of these three possibilities you can use, then there is another patent, then another and another. It gets like crossing a minefield. Each step you take, each design decision, probably won't step on a patent, so you can take a few steps and probably there won't be an explosion. But the chance you can get all the way through the minefield and develop the program you want to develop without ever stepping on a patent gets less and less as the program gets bigger.
Copyright 2002 Richard Stallman
Verbatim copying and distribution are permitted without royalty in any medium provided this notice and the copyright notice are preserved.