mernst@theory.lcs.mit.edu (Michael Ernst) (07/29/90)
brnstnd@kramden.acf.nyu.edu (Dan Bernstein), who is correct in conclusion but incorrect in detail, says: > By the way, the 1981 decision that software patenters keep gibbering > about didn't say that software patents were valid. This is Diamond v. Diehr, decided on March 3, 1981 by the United States Supreme Court. I quote from the Opinion of the Court, page 191: We view respondents' claims as nothing more than a process for molding rubber products and not as an attempt to patent a mathematical formula. Here "mathematical formula" = "computer program" = "algorithm" > What happened was > that company A used a computer in one part of an innovative method for > doing process X. Now it has been established that using a computer to > perform an algorithm doesn't make the algorithm patentable. Nothing had been established then, of course (and I'm not convinced that this has been firmly established even now, though it should be), but this case was supposed to decide a separate issue. Diehr wasn't claiming a patent on the Arrhenius equation, and the court emphasized that had it thought he was trying to do so, it would have thrown the patent out. From page 185: We defined "algorithm" as a "procedure for solving a given type of mathematical problem," and we concluded that such an algorithm, or mathematical formula, is like a law of nature, which cannot be the subject of a patent. > Company B > took the opportunity to challenge A's patent, saying that the use of a > computer rendered the entire process unpatentable. Actually, "Company B" is Commissioner of Patents and Trademarks Diamond, who was supporting the judgement of a patent examiner (who'd rejected the patent) and of the Patent and Trademark Office Board of Appeals, which had been overturned by the Court of Customs and Patent Appeals. > This is, of course, > backwards: using a computer doesn't give you a patent, but it doesn't > take away a patent either. And that's all the court said. This is absolutely correct. The case has been widely misunderstood, of course, and the result has been an about-face in the attitude of the Patent Office, which has fallen over itself in its eagerness to grant patents, even on algorithms. It has no idea of how many such patents it has granted. Twenty more years of expensive lawsuits is the wrong way to decide this issue. Computer professionals need to air the issues and then make their opinions known. Then perhaps we can propose thoughtful legislation which protects intellectual property without strangling the industry via overprotection. What the law currently says shouldn't matter to us in the debate over what it should say. Rather, we should be acting to change those laws which are inadequate. At present this includes both the copyright and patent statues, at least as they're being interpreted by the courts. -Michael Ernst mernst@theory.lcs.mit.edu
escher@Apple.COM (Michael Crawford) (07/31/90)
In article <MERNST.90Jul28233614@emu.lcs.mit.edu> mernst@theory.lcs.mit.edu (Michael Ernst) writes: (the supreme court sez... > We defined "algorithm" as a "procedure for solving a given type of > mathematical problem," and we concluded that such an algorithm, or > mathematical formula, is like a law of nature, which cannot be the > subject of a patent. > Let me first say that I oppose software patents, and I have no legal qualifications... However, I have read a lot about them, and this is my understanding and opinion: I don't feel that an algorithm is a law of nature. I think an algorithm is either a process, or a device, just like a manufacturing process, or a gear train. It seems to me that if one argues that an algorithm is a law of nature, one could also argue that a gear train is also a law of nature. A gear train, or manufacturing process, actually, is an expression of how the laws of nature may be obeyed and exploited to achive a particular goal. The fact that an algorithm strikes a judge as being an abstract law seems to me to be a mistake on the judge's part. About existing patents... the basic criterion used by the Patent Office is that the patent must be written in such a way as to show the relationship of the algorithm to the control of a physical device. In this way it would be some kind of control mechanism, like a governor. However, one could go to some length, after inventing an algorithm, to devise a machine that the algorithm can control, and then patent it. I think this may be happening. The reason I feel that algorithms should not be patented goes back to the very reasons patents exist: to encourage technological creativity. Patents aren't really there to guarantee anyone a profit. They exist to encourage someone to actually market a device, and _publish_its_secret_workings_for _posterity_. After 17 years (or less, if the patent is not maintained), the patent goes into the public domain. Without patents, we would have both fewer mechanical inventions, and what there were would be kept secret. If the secret died with the inventor, it would be lost to the world. Consider, for example, the document disclosure program, which gives you two years of patent-like protection for a small fee, and not much hassle. If you don't follow through with the patent, it is placed in the public domain, and the government has the specs for all to see. The objective here is to get the inventions into use, and if they are not used, to give them to the world. Now, all these reasons exist for algorithms, but I think the algorithm is a special case: 1 The cost of manufacturing an algorithm is nearly zero. (I mean copying software, not writing it in the first place). 2 Algorithms are useful for many purposes. An algorithm conceived for a particular use may be readily applied to a completely different problem. Since a general purpose computer can use any algorithm, the entirety of the set of algorithms is available to the designer of a particular application. This is not so with patents on physical devices; a designer of an aircraft engine cannot make much use of a patented hybrid plant! These two facts sum up what I feel is the very beauty of software. It is the ultimate mass-producible product, and there is no cost (inherent) in including previously invented mechanisms in a new application. I wonder if anyone has made the argument of what free software could do for the third world? Consider that when GNU is released, a third world country could be set up with a multi-user Unix system, with complete source, for about $2000 (a 386 SX, hard disk, and some terminals). Consider that one of the great strengths of Soviet and mainland Chinese science is in theoretical studies, because they do not have the cash to buy much hardware -- what could they do in software? All they need is a relatively cheap computer system. How about agricultural forecasting in tropical countries? If all these users had to pay a per-user fee for their OS's and applications, they could not get tenth the amount of work done that they could with free software. -- Michael D. Crawford Oddball Enterprises Consulting for Apple Computer Inc. 606 Modesto Avenue escher@apple.com Santa Cruz, CA 95060 Applelink: escher@apple.com@INTERNET# oddball!mike@ucscc.ucsc.edu The opinions expressed here are solely my own. alias make '/bin/make & rn'
brad@looking.on.ca (Brad Templeton) (08/01/90)
I agree that there should be no difference between a truly devised algorithm and a mechanical device. They are the same. If a patent applies to one, then it should apply to the other. And while it is true that *one* of the reasons for patent law is to encourage disclosure, and that some could argue that society merely trades a temporary monopoly (and the possible profits) for disclosure and eventual release to the public domain, I don't think that's the whole story. Property ownership is a fundamental principle in our society, though some may disagree with it. Many, however, feel it is right that you own, control and profit from the creations of your own mind. Patent law exists not only to encourage disclosure of inventions, but because it's the right thing to do, under this ethic. If this were not the case there would be no reason to offer patent protection for ideas that are impossible to hide. One can invent a specially shaped tool but the moment you let it out the cat is out of the bag. It's impossible to sell it and hide it. In some rare small-market cases this can be done with trade secret contracts, but that's not always what happens. So I think patents, like copyrights, are granted not only for the utilitarian reason of encouraging dissemination of the idea or work, but for the propertarian ethic as well. Whether this *should* be the case is the subject of an ongoing debate all over the world. I am not sure we can settle it here. But if you *do* believe in property, then I argue that intellectual property is the truest form of property there can be. The relationship between myself and my real estate is a virtual one created by society and law. The relationship between myself and my software is a real one that exists outside of society. Society could crumble and my land would not be my land, but my thoughts would still be mine. I actually suspect things might be best served by a special kind of software/algorithm patent. This patent would be granted after examination by experts in the field, and might last for a shorter time than the current mechanical patent scheme. -- Brad Templeton, ClariNet Communications Corp. -- Waterloo, Ontario 519/884-7473
kolender@ics.uci.edu (Kurt Olender) (08/04/90)
I too at first thought that if a patent could be applied to a physical process (like a process to manufacture a chemical product) it should be applicable to an algorithm. But thinking about it further, an algorithm is a process that can (theoretically at least) be applied by the human mind that has results that can (theoretically at least again) exist only in the human mind. Suppose that Gauss had patented his method for solving simultaneous equations. You legally could not use that method to solve a 2x2 system of equations in your head in a high school algebra class without someone paying Gauss a fee. This seems on the face of it a ridiculous concept. The fact that a computer can be used to solve this problem is essentially irrelevant. You don't need a computer to use the algorithm. And that in my opinion is the fundamental difference between an abstract algorithm and a physical "algorithm". A physical process REQUIRES a realization outside the human mind for its application. A computational algorithm does not, and for that reason should not be patentable.
brad@looking.on.ca (Brad Templeton) (08/04/90)
I fail to see the difference between in the human mind and outside the human mind. I can envison many patented physical processes and devices in my mind, and even perform gedankenexperimenten with them. Likewise I can implement computer algorithms in real world hardware. If Gauss could patent gaussian elmination (and he couldn't if you could really do it in your head without having been told how in advance) so what? The patent would have expired long ago. Many other things have been patented with exactly the same constraints. No, a process, mechanism and algorithm are the same, and they should be treated the same. Patent them all or patent none. -- Brad Templeton, ClariNet Communications Corp. -- Waterloo, Ontario 519/884-7473
kolender@ics.uci.edu (Kurt Olender) (08/08/90)
In article <1990Aug04.031721.23238@looking.on.ca> brad@looking.on.ca (Brad Templeton) writes:
I fail to see the difference between in the human mind and outside the
human mind. I can envison many patented physical processes and devices
in my mind, and even perform gedankenexperimenten with them.
Say you perform a gedanken experiment with the chemical process used to
produce Teflon. In the end you still don't have any Teflon. You haven't
violated the patent because you don't have a useful result and that in the end
is the basis of a patent. You must get a useful result. If you use a
patented process in some way other than the patented useful result, you don't
violate the patent.
Now execute Gaussian elimination in your head on a system of simultaneous
equations that describes the constraints under which your business can make a
profit. You have a useful result.
Likewise I can implement computer algorithms in real world hardware.
True. But then the physical device is patented, not the algorithm. No one
else can build a similar chip for 17 years. But they can use the
algorithm in other ways.
If Gauss could patent gaussian elmination (and he couldn't if you could
really do it in your head without having been told how in advance) so what?
The patent would have expired long ago. Many other things have been
patented with exactly the same constraints.
That's not the point either. Substitute some other useful computational
algorithm invented sometime in the past 5 years instead if you prefer.
Suppose Frobnob's computational algorithm (since you object to my use of
Gauss) is patented and the patent is still in effect. You can't teach it
without executing it. You can't execute it (even in your own mind) without
paying a license fee since if you do you are producing a useful result from a
patented process.brad@looking.on.ca (Brad Templeton) (08/08/90)
So what's the problem. The point of patent law is to protect inventions (and to also thus encourage their publication). If Frobnob invents a really good algorithm, and it's not at all obvious to those who haven't seen it before, then why not have Frobnob own it. We're assuming that without Frobnob's work, *nobody* would know the algorithm -- it wouldn't even exist. Algorithms are not laws of nature like F=MA. At least not the kind of algorithms I'm talking about. They don't exist until somebody creates them. This is the philosophy. We then have a couple more pragmatic issues to deal with. a) What about the kind of invention that is "waiting to happen", such that one person is merely first, and others independently discover it. b) What about inventions that are truly obvious -- where it's the problem that's new, and not the solution. Well, these are tough questions, but they are the same for algorithms and for mechanical inventions. The decision, in patent law, to give the rights to the first filer, screwing subsequent independent developers, is a tough, and often unfair one. But what, other than the abandonment of patent, is the alternative. In the case of B, that's where we need judges and experts in the field of research. But again, no difference between algorithms and mechanisms. A mechanism *is* an algorithm to me. The thing that's valuable about a patented mechanism is "how it works" not how it is made. The difference is engineering. If the only difference between idea and realization is competent engineering, then as far as intellectual property is concerned, the idea and its realization are the same. If truly creative engineering is required, then it's more complex. (But truly creative engineering is also invention, unlike more mundane engineering of the "here's blueprints, make this for me" sort.) -- Brad Templeton, ClariNet Communications Corp. -- Waterloo, Ontario 519/884-7473
kolender@ics.uci.edu (Kurt Olender) (08/09/90)
Brad Templeton writes:
... <other text removed> ...
A mechanism *is* an algorithm to me. The thing that's valuable about
a patented mechanism is "how it works" not how it is made.
-----
I disagree. The thing that is VALUABLE about a mechanism is what it makes or
does. One of the points that must be satisfied to obtain a patent is to show
that the result of the mechanism or process is useful in some way. And in the
end, it's those results (or the capability to obtain them) that you are
selling.
Rube Goldberg used to conceive of lots of devices with interesting "how's"
that didn't do much that was useful.
Now what is the result of the new algorithm?
If it's a mathematical or mental entity, then I believe that it should not be
patentable, for the same reasons that mathematics has never been
patentable. The results can exist only in someone's mind, or as
visual/audio/electronic/etc representations of that thought. A compressed
data stream is simply an electronic representation of a mathematical
operation.
Let's look at it from a more practical viewpoint. How can you enforce a
patent on the use of a method to produce some result if someone can
effectively use the method and keep the result solely in their mind?
If it's a physical entity, then it should be patentable under the conditions
currently existing.
a) What about the kind of invention that is "waiting to happen", such that
one person is merely first, and others independently discover it.
b) What about inventions that are truly obvious -- where it's the problem
that's new, and not the solution.
I am not disagreeing with you about these points, but they are
irrelevant from the standpoint of what I am trying to state, which is:
There IS a fundamental difference in algorithms. Some produce results which
are merely mathematical objects or their representations. Others produce
results which are solely physical objects. The former should not be
patentable.brnstnd@kramden.acf.nyu.edu (Dan Bernstein) (09/02/90)
In article <1990Aug27.222359.4398@sco.COM> seanf@sco.COM (Sean Fagan) writes: > In article <67638@sgi.sgi.com> karsh@trifolium.sgi.com (Bruce Karsh) writes: [ basically claims that if you don't have a patent or a ] [ user-interface copyright, you'll find it difficult to ] [ get a loan for software development and marketing ] > Software companies (off the top of my head) that were founded without > requiring a software patent or user-interface copyright: Microsoft, Lotus, > the Dbase people, Borland Int'l, SCO (yep), ISC, Locus, the FSF (yep), > Cygnus Corp. Not to mention the many hardware companies that offer to market your software---hence reducing your real costs nearly to zero. It's entirely safe for the author, who gets royalties that are often more than the profits he'd get if he marketed without support. Patents and user-interface copyrights have always been irrelevant to such deals. ---Dan