INTRODUCTION I'm finally going to write down some of my evolving views on why software patents are fraught with dangers. Yes, many of you are no doubt already convinced that software patents (XOR cursor, RSA, digicash, etc.) are a Bad Idea, but I have a different angle on this which I'd like to present. SUMMARY Thesis: Software patents are a bad idea because they freeze the evolution too early and payment metering schemes are too difficult to arrange, which also helps to freeze evolution. Software patents are bad because customers cannot freely and without entanglements incorporate the ideas into their own products. The situation has become much worse with software, because there is no physical object which can be used to meter usage of a patent. HOW PATENTS WORK WITH PHYSICAL OBJECTS Before explaining this thesis, let's look at how patents work with _physical_ inventions, even if those inventions embody abstract process or software inventions. The microprocessor is a good example, which I'll use here for most of my examples. (Note that Intel did not try to patent the "basic idea" of the microprocessor; Gilbert Hyatt claims he filed before Intel produced the first 4004, but this claim and the swirl of issues around the Patent Office's granting of a patent to him are a separate topic.) A microprocessor chip incorporates numerous items that are patented, copyrighted, or that have trade secret status. Some of these patents, etc., even involve design tools used to design the chip. And some of the patents involve sophisticated production methods. And yet a customer can simply *buy* the chips and use them as he *wishes*, in hamburger cookers or in digital money schemes, without any further regard for the patents. No entanglements. You go down to Fry's Electronics, buy some Pentiums or PPCs and use them as you wish. No permissions needed from Intel or Motorola, no complex license agreements saying you won't use the chips to compete with Intel's board business or with Motorola's cellular phone business, no restrictions (save for government laws about munitions, etc.) on who you can resell the chips or systems to. That is, the whole set of patent and copyright issues is _encapsulated_ or _reified_ into the instance being sold. The physical object carries the embodiment of the patents (including the process patents used to fabricate the chip: the user of such a chip can be oblivious to these patents and need not worry that a design will infringe on these patents). [Sidenote: The astute observer will note some conceptual similarities to "on-line clearing," to a "cash-and-carry" business. Once you've paid your money and taken delivery of your 486 or your laser or your widget, how you incorporate it into future products is generally your business and yours alone. I say "generally," and legal beagles will point out that some hardware sales may have restrictions placed on what can be done with the hardware. Contracts are always possible. But for most objects, there is no such additional contract. The widget seller has made his money by the sale of his widget and cannot really try too hard to make more "downstream" money, except by modifying his future prices to reflect what he perceives demand to be.] SOFTWARE IS REPLICABLE Hardware object cannot be easily replicated, and hence patents can be reified into the objects. Software objects of course _can_ be easily replicated, which is why convoluted licensing and complex payment contracts are involved. (This is an incredibly important distinction, and one which also relates closely to why "software ICs" and "object widgets" are not built, sold, improved upon, etc. by the software industry. Ted Kaehler of Apple once told me that an H-P analyst/thinker came up with this sort of analysis of why there is no "software IC" industry to parallel the actual IC industry. He argues that there is no "learning curve" for software, in the way there is for physical objects. Sorry, I don't recall his name. If I were pursuing this essay here as an actual, researched paper, I would did up this guy's analysis. Also, Brad Cox has done a lot of work on "software ICs"--he coined the term, in fact--and his papers are worth looking at.) HOW THIS APPLIES TO THE CRYPTO COMMUNITY Contrast this with the following "software patent" situations, the ones directly relating to our crypto community: 1. The public key and RSA patents. RSADSI (and I'm ignoring for now the implications of the PKP/Cylink complication) wants to see intended applications and to work out arrangements for license payment based on profits, volume, effects on their other licensees, etc. This limits the ways in which the RSADSI patents have been incorporated into evolutionary and revolutionary products. 2. The Chaum Digicash patents. Digicash wants $150K upfront, plus 10% of profits. (Cf. Adam Back's posting on this today.) Digicash has their own particular version they are pushing, but others are effectively shut out of developing experimental applications. Imagine the situation if the garage shop developers of personal computers had been forced to pay Intel up front for the right to design in an 8008 or 8080 microprocessor, and then promise to pay 10% of their profits to Intel (and 5% to the power supply vendor, 8% to the keyboard supplier, 4% to the capacitor makers, 9% to the memory chip companies, and so on....). In both of these cases, the confusing, complicating, and almost insurmountable issue is that these conditions are very hard to meet. It's not just the issue of sharing a business plan with RSADSI or Digicash, it's the fact that many aspects of one's business are unknown: the viability, the volume, the future products, etc. The "garage shop" folks usually can't even get in the door to talk to these companies. THE REAL ISSUE: TRANSACTION COSTS ARE THWARTING EVOLUTONARY DEVELOPMENT To me, the issue of concern is not that "software ideas should not be patented." After all, why should the _idea_ of a safety pin or a windshield wiper be any more patentable than the idea for a blinded transaction? Rather, the issue of concern is that the patents on the software ideas and concepts mean that experimenters, developers, and hackers cannot buy a license for digicash the way they would buy some ICs and then experiment, develop, and hack. Another way of looking at this is that _transaction costs_ are too high and are thwarting normal evolutionary development. The guy in the garage trying to develop a "digital postage stamp," for example, can't use the Chaumian blinding protocols without hiring lawyers, paying Chaum his up-front fee, and laying out his designs and business plans (which he very probably doesn't even have!). (Clarification: I'm not saying one can't buy fairly cheap RSA versions, such as the code in RSAREF. What I'm saying is that one can't get a "core module" for digicash, for example, and then test out the market with various implementations, going into volume production with the ones that are most successful.) (And you can perhaps tell from my views here that I am not villifying either RSADSI or Digicash here for their policies: they are trying to make a reasonable profit in the face of a situation very much unlike the situation faced for physical objects like microprocessors. David Chaum points out that "there is no digital coin." Likewise, there is no "software coin," and so the developers of novel ideas cannot "reify" the ideas in things they can sell...rather, we see the current approach of convoluted and restrictive licensing contracts. They try to make up for the lack of a "software coin" by grilling potential customers about their expected markets and by setting up complicated contracts to ensure--they think--enough profits.) EVOLUTION What do I mean my "freezing" the evolution of products? Take the case of Digicash and their test release. As I pointed out in a recent essay ("Crypto + Economics + AI = Digital Money Economies"), we are _impoverished_ with regard to the basic building blocks we have for a digital money system. Where are the equivalents of what we find necessary in the existing financial world? Does anybody think that a particular instance of digital cash is the end state of the evolution of digital cash? It is likely just the beginning. But by the licensing of particular systems, and by having convoluted and restrictive arrangements for use of the patents, the evolution of digital money and crypto systems grinds to a halt. Instead of having dozens of implementations of digital cash systems (I mean real digital cash systems, embodying the Chaumian or Brandsian blinding schemes, not the cheesy "smartcard" systems that are carelessly called "digital cash), we have only a very few: a road toll system here, a play money experiment there, etc. Evolution works by _differential reproduction_ (often misleadingly called "survival of the fittest"). Several dozen variants of personal computers are introduced (Sphere, Altair, Imsai, Sol, Exidy, Pet, Apple I, Apple II, IBM PC, Macintosh, etc.), and customers reward the ones they like with increased sales, causing the "genes" (or memes) of the winning products and companies and designers to propagate. The ability to incorporate PGP into tools and objects without complicated entanglements is one reason PGP did so well. It was a "building block" that anybody could build into other tools and objects, without regard for how big the market might be for their tools, without regard to what other users were doing, etc. (I'm ignoring for the moment any claims RSADSI may have had that PGP infringed.) CONCLUSIONS Both the public key and digital cash situations are being affected by this inability of people to use the core ideas as building blocks for more complicated--or more revolutionary--systems. It's as if Intel had, in 1971, insisted that all designs be approved by them, and that chips could not be sold to competitors who had already licensed the chips, and that 10% of all profits be ceded to Intel. This would likely have had a big effect on the explosion of applications that came about in the years since the micro's introduction. Do I have a solution for RSADSI, Cylink, Digicash, and others? Given that there is no "software coin," my hunch is that no simple solution exists. To take David Chaum's case in particular, I think he'll end up making a lot more money by being the guru and consultant on digital money systems (the inventor of a field and all) than by having licensing schemes which are probably unenforceable and which are mostly slowing down the evolution of the bits and pieces needed for true digital money economies. I've had these ideas swirling in my head for a long time, and thought it was time to share them with you folks. Comments are welcome. --Tim May Views here are not the views of my Internet Service Provider or Government. ---------:---------:---------:---------:---------:---------:---------:---- Timothy C. May | Crypto Anarchy: encryption, digital money, tcmay@got.net 408-728-0152 | anonymous networks, digital pseudonyms, zero Corralitos, CA | knowledge, reputations, information markets, Higher Power: 2^756839 | black markets, collapse of governments. "National borders are just speed bumps on the information superhighway."