By	Casey Muratori

To be honest, I don’t just think software patents are bad. I think all patents are bad. Suppose you invent a flying car. You invent it all by yourself  —  literally. You did all the work. You invented every last piece and built the car yourself. You unveil it to the world, and everyone loves it. Then, lawyers come and take all your money and tell you you can’t sell flying cars anymore because it turns out someone had already filed a patent on a flying car a few years beforehand. That’s the reality of patents. Patents are not about protecting intellectual property from theft  —  that is an almost negligible part of the law. What patents actually are is a right for people who file them to sue other people for damages, regardless of whether or not any actual theft took place. Don’t believe me? Just take a look at how damages are awarded in patent trials: regular damages are assessed if the court concludes the patent was infringed, but triple damages are assessed if it turns out the infringer knew about the patent.

It has nothing to do with theft. The assumption is that you didn’t steal it at all, you came up with it yourself in the normal course of business. The only time the law even cares if you stole the idea is when it comes time to determine how much you’ll pay in damages! So it’s very important to understand that patent law is not about preventing theft. It’s about forcing people to pay for the use of patented ideas even when they came up with the idea themselves. So I am anti-patent in all fields  —  it’s bad law that has no sound research to support it even existing. It’s not like copyright law which is more sensibly designed to protect a product from theft, which is the kind of law that actually makes some sense in a digital economy. But enough with the preamble. On to software patents specifically. Back in 2014, a Stanford Law student asked me why I thought software patents were bad. Since it came up again recently on Twitter, I thought I’d post my response as a starting point. It’s just an e-mail I typed in off the top of my head, so you’ll forgive me the lack of references I might otherwise pepper throughout:

Both these notions run counter to the fundamental idea that laws exist for the betterment of society as a whole, not to make individuals rich. Even the original phrasing of our Constitution clearly states that the ability for Congress to grant exclusive rights to “inventors” is expressly for the purpose of advancing the arts and sciences. It says nothing about any other purpose. I believe the original wording, and I think it should be taken quite literally. In a world where people needed no economic incentives to come up with ideas, allowing all ideas to be used freely is always equal to or better than applying restrictions  —  hopefully that is self evident. So the question of whether or not to extend patent protection to a field should be exclusively concerned with whether or not it increases the number of and application of good ideas. When viewed in this light, patents for software have, to my knowledge, no support whatsoever. Here are some elaborations:

There are over 40,000 software patents issued yearly, and the only time you will be able to know if you have violated one of them is once you are contacted by a lawyer for the patent holder. This means that regardless of the intended benefit of patents  —  that people would license them, that people could prevent others from entering a market, or that people would invent new things to avoid infringing  —  none of them occur directly, because nobody creating a new piece of software knows a priori that they are infringing. Instead, almost the entire practical result from software patents is to create an uncertainty settlement market, whereby software developers pay patent holders after the fact to avoid entanglement in lawsuits. Most of the time, neither party actually knows whether a patent was infringed, or even whether the patent is valid, since technically a court of law is required to determine either of those things. The only ways to change this would be a) to develop some method by which a program could be automatically checked for patent violations, and this is a definitively unsolved problem in artificial intelligence, or b) reduce the total number of software patents to some sub-100 number that every programmer could know, and print them in a clear way that made it easy to understand what they meant. But even if you did that, it is unclear that this is valuable holistically, as you have wasted every programmer’s time with it, and you are “protecting” a scant double-digit number of algorithms at best.

I have personal experience with this, as the vast majority of the money that I’ve ever made in the software industry has come directly from licensing software that I did not patent (because I am against them on principle). Unlike simple machines, which cannot be copyrighted, software is automatically copyrighted by the creator and when it is licensed, the licensee cannot continue to use the software without obeying the terms of the license agreement. This allows for a flourishing software licensing market that works exactly as intended, no patents necessary. Since valuable software is very time-intensive to create and debug, and because technical support is a desired complimentary service, companies often have no incentive to clone software that they could license, and when they do wish to do so, it is often because the licensable software cannot do what they need it to do, and therefore they do not “clone” it anyway, but rather create a wholly different system for their own use.

This is the part that is perhaps the most enraging to serious software developers, although it is the least relevant, perhaps, to the overall argument for or against. But it is worth noting, since most people talking about software patents are either not programmers at all, or are not good programmers. It is difficult to explain to a non-programmer or average programmer what this means, but an analogy would be if someone were to tell you that movies had to be protected by patents to encourage innovation (whatever that means), but the movies that were being patented were “Police Academy 5”, “Tyler Perry’s Medea’s Family Reunion”, and Pauly Shore’s “Encino Man”, while the classics like “Gone with the Wind” and “Star Wars” were unpatented. When you say, “wait, why are we spending all this money on patent protection when all the great films don’t need it, and all that happens is that lousy films sue people for copying them?”, someone says, “No no, it’s necessary, we just need to fix the system so that only good films are patented!” Hopefully that gives you some insight on what it looks like. It makes no sense to say that you need to “fix the system” when none of the great stuff required it! Just get rid of the system, and you’re done. The best you could do by “fixing” the system is making it do nothing, and then you’d just be wasting money by having an ineffectual system in the first place.

Unless we wish to go down the Potter Stewart road of “I know it when I see it”, people talking about software patent reform producing “high quality patents”, or that “low quality patents” are the problem with the current system, have never attempted to give a definition of what these terms mean. What is a measure of software patent quality? How would a patent examiner apply it? If we are claiming to encourage people to spend money on research and development, the amount of manpower necessary to produce the patent would presumably have something to do with its “quality”. Yet currently, there are no software patents (to my knowledge) that involve anything other than very short research and development periods. Most software patents are ideas that were had in hours and developed in weeks. In fact, there is rarely any portion of a piece of software small enough to be patented that wasn’t developed in less than a few months. Is that really the kind of R&D size we need to be protecting with monopoly rights?

Pure software patents were not really used prior to the 1990s. If you split time in half from pre-1990 and post-1990, it is kind of amusing what you get. Pre-1990 saw the invention of FORTRAN, C and C++, UNIX, TCP/IP, the Windows-Icons-Menus-Pointers interface, the command line operating system, the graphical operating system, the word processor, the spreadsheet, the directory-based file system, the relational database, CAD, raytracing, rasterization, e-mail, hypertext, the client/server architecture, computer-aided mathematics, virtual painting, text-based search… the list goes on forever. Post-1990 saw… well, it saw basically a bunch of new languages like HTML/JavaScript/PHP (none of which were patented, BTW) and then a slow movement of all the things that had already been invented into those languages once the “web” became popular. Whenever you sit down at a computer today, it is highly likely you will interact exclusively with things that are modified versions of something that existed in 1975, and the difference you’d perceive between it and the 1975 version will largely be due to the increased power afforded by the hardware, not by any software innovation. About the only thing that really could be said to have drastically improved would be the “search engine”, and of course the company who pioneered all that (Google) is actually on record as saying they don’t like software patents. So the question is, what exactly are we claiming we are encouraging with these software patents? If everything we use was conceived before software patents were viable, what evidence do we actually have that patents are useful? If anything, our rate of innovation seems to have slowed dramatically, although I would assume that is more likely due to the fact that new things are easier to discover when there are less things already discovered. But even that is uncertain. It could be that patents have hurt more than we realize.

Unlike simple machines or chemicals, personal expression is both a possible and frequently employed capability of software, and more artists write art programs every day. Allowing patents on software is therefore much more similar to allowing patents on phrases or sentences in English, and raises all the same concerns. What would the Supreme Court say to an artist whose program infringes on some number of patents? There is no chance that patents pass strict scrutiny, if we are being reasonable. The government does not have a compelling interest in patents; giving out a monopoly to inventors is merely the method they have preferred, but as is readily apparent from the period of robust software innovation from the 1950s through the 1980s, the government does not need software patents to have a viable software industry. Software patents are in no way “narrowly tailored”  —  giving out a 20-year monopoly on an entire concept is nowhere near the narrowest possible thing the government could do to protect software IP, and this is proven by the fact that copyright exists and works well. It is much less restrictive and serves the same purpose.

This is just an off-the-top-of-my-head explanation of the main points involved. There’s actually a lot more behind each point, and I would encourage people to really dig in and see how the software patent system works and whether it ever accomplishes any of the positives its proponents often tout. Based on years of experience, ingesting a lot of research, and having read every US Supreme Court case on the subject, I can tell you that I have found literally no evidence to ever suggest that software patents are useful, and plenty to show that they are actively harmful. And I’ve yet to hear a coherent argument to the contrary that didn’t fall apart immediately under a moment’s scrutiny.