March 2006(This essay is derived from a talk at Google.)A few weeks ago I found to my surprise that I'd been granted four patents. This was all the more surprisingbecause I'd only applied for three. The patents aren't mine, ofcourse. They were assigned to Viaweb, and became Yahoo's when theybought us. But the news set me thinking about the question ofsoftware patents generally.Patents are a hard problem. I've had to advise most of the startupswe've funded about them, and despite years of experience I'm stillnot always sure I'm giving the right advice.One thing I do feel pretty certain of is that if you're againstsoftware patents, you're against patents in general. Gradually ourmachines consist more and more of software. Things that used tobe done with levers and cams and gears are now done with loops andtrees and closures. There's nothing special about physical embodimentsof control systems that should make them patentable, and the softwareequivalent not.Unfortunately, patent law is inconsistent on this point. Patentlaw in most countries says that algorithms aren't patentable. Thisrule is left over from a time when "algorithm" meant something likethe Sieve of Eratosthenes. In 1800, people could not see as readilyas we can that a great many patents on mechanical objects werereally patents on the algorithms they embodied.Patent lawyers still have to pretend that's what they're doing whenthey patent algorithms. You must not use the word "algorithm" inthe title of a patent application, just as you must not use theword "essays" in the title of a book. If you want to patent analgorithm, you have to frame it as a computer system executing that algorithm.Then it's mechanical; phew. The default euphemism for algorithmis "system and method." Try a patent search for that phrase andsee how many results you get.Since software patents are no different from hardware patents,people who say "software patents are evil" are saying simply "patentsare evil." So why do so many people complain about software patentsspecifically?I think the problem is more with the patent office than the conceptof software patents. Whenever software meets government, bad thingshappen, because software changes fast and government changes slow.The patent office has been overwhelmed by both the volume and thenovelty of applications for software patents, and as a result they'vemade a lot of mistakes.The most common is to grant patents that shouldn't be granted. Tobe patentable, an invention has to be more than new. It also hasto be non-obvious. And this, especially, is where the USPTO hasbeen dropping the ball. Slashdot has an icon that expresses theproblem vividly: a knife and fork with the words "patent pending"superimposed.The scary thing is, this is the only icon they have forpatent stories. Slashdot readers now take it for granted that astory about a patent will be about a bogus patent.That's how bad the problem has become.The problem with Amazon's notorious one-click patent, for example,is not that it's a software patent, but that it's obvious. Anyonline store that kept people's shipping addresses would haveimplemented this. The reason Amazon did it first was not that theywere especially smart, but because they were one of the earliestsites with enough clout to force customers to log in before theycould buy something. [1]We, as hackers, know the USPTO is letting people patent the knivesand forks of our world. The problem is, the USPTO are not hackers.They're probably good at judging new inventions for casting steelor grinding lenses, but they don't understand software yet.At this point an optimist would be tempted to add "but they willeventually." Unfortunately that might not be true. The problemwith software patents is an instance of a more general one: thepatent office takes a while to understand new technology. If so,this problem will only get worse, because the rate of technologicalchange seems to be increasing. In thirty years, the patent officemay understand the sort of things we now patent as software, butthere will be other new types of inventions they understand evenless.Applying for a patent is a negotiation. You generally apply for abroader patent than you think you'll be granted, and the examinersreply by throwing out some of your claims and granting others. SoI don't really blame Amazon for applying for the one-click patent.The big mistake was the patent office's, for not insisting onsomething narrower, with real technical content. By granting suchan over-broad patent, the USPTO in effect slept with Amazon on thefirst date. Was Amazon supposed to say no?Where Amazon went over to the dark side was not in applying for thepatent, but in enforcing it. A lot of companies (Microsoft, forexample) have been granted large numbers of preposterously over-broadpatents, but they keep them mainly for defensive purposes. Likenuclear weapons, the main role of big companies' patent portfoliosis to threaten anyone who attacks them with a counter-suit. Amazon'ssuit against Barnes & Noble was thus the equivalent of a nuclearfirst strike.That suit probably hurt Amazon more than it helped them. Barnes &Noble was a lame site; Amazon would have crushed them anyway. Toattack a rival they could have ignored, Amazon put a lasting blackmark on their own reputation. Even now I think if you asked hackersto free-associate about Amazon, the one-click patent would turn upin the first ten topics.Google clearly doesn't feel that merely holding patents is evil.They've applied for a lot of them. Are they hypocrites? Are patentsevil?There are really two variants of that question, and people answeringit often aren't clear in their own minds which they're answering.There's a narrow variant: is it bad, given the current legal system,to apply for patents? and also a broader one: is it bad that thecurrent legal system allows patents?These are separate questions. For example, in preindustrial societieslike medieval Europe, when someone attacked you, you didn't callthe police. There were no police. When attacked, you were supposedto fight back, and there were conventions about how to do it. Wasthis wrong? That's two questions: was it wrong to take justiceinto your own hands, and was it wrong that you had to? We tend tosay yes to the second, but no to the first. If no one else willdefend you, you have to defend yourself. [2]The situation with patents is similar. Business is a kind ofritualized warfare. Indeed, it evolved from actual warfare: mostearly traders switched on the fly from merchants to pirates dependingon how strong you seemed. In business there are certain rulesdescribing how companies may and may not compete with one another,and someone deciding that they're going to play by their own rulesis missing the point. Saying "I'm not going to apply for patentsjust because everyone else does" is not like saying "I'm not goingto lie just because everyone else does." It's more like saying"I'm not going to use TCP/IP just because everyone else does." Ohyes you are.A closer comparison might be someone seeing a hockey game for thefirst time, realizing with shock that the players were deliberatelybumping into one another, and deciding that one would on no accountbe so rude when playing hockey oneself.Hockey allows checking. It's part of the game. If your team refusesto do it, you simply lose. So it is in business. Under the presentrules, patents are part of the game.What does that mean in practice? We tell the startups we fund notto worry about infringing patents, because startups rarely get suedfor patent infringement. There are only two reasons someone mightsue you: for money, or to prevent you from competing with them.Startups are too poor to be worth suing for money. And in practicethey don't seem to get sued much by competitors, either. They don'tget sued by other startups because (a) patent suits are an expensivedistraction, and (b) since the other startups are as young as theyare, their patents probably haven't issued yet. [3]Nor do startups,at least in the software business, seem to get sued much by establishedcompetitors. Despite all the patents Microsoft holds, I don't knowof an instance where they sued a startup for patent infringement.Companies like Microsoft and Oracle don't win by winning lawsuits.That's too uncertain. They win by locking competitors out of theirsales channels. If you do manage to threaten them, they're morelikely to buy you than sue you.When you read of big companies filing patent suits against smallerones, it's usually a big company on the way down, grasping atstraws. For example, Unisys's attempts to enforce their patent onLZW compression. When you see a big company threatening patentsuits, sell. When a company starts fighting over IP, it's a signthey've lost the real battle, for users.A company that sues competitors for patent infringement is likea defender who has been beaten so thoroughly that he turns to pleadwith the referee. You don't do that if you can still reach theball, even if you genuinely believe you've been fouled. So a companythreatening patent suits is a company in trouble.When we were working on Viaweb, a bigger company in the e-commercebusiness was granted a patent on online ordering, or something likethat. I got a call from a VP there asking if we'd like to licenseit. I replied that I thought the patent was completely bogus, andwould never hold up in court. "Ok," he replied. "So, are you guyshiring?"If your startup grows big enough, however, you'll start to get sued,no matter what you do. If you go public, for example, you'll besued by multiple patent trolls who hope you'll pay them off to goaway. More on them later.In other words, no one will sue you for patent infringement tillyou have money, and once you have money, people will sue you whetherthey have grounds to or not. So I advise fatalism. Don't wasteyour time worrying about patent infringement. You're probablyviolating a patent every time you tie your shoelaces. At the start,at least, just worry about making something great and getting lotsof users. If you grow to the point where anyone considers you worthattacking, you're doing well.We do advise the companies we fund to apply for patents, but notso they can sue competitors. Successful startups either get boughtor grow into big companies. If a startup wants to grow into a bigcompany, they should apply for patents to build up the patentportfolio they'll need to maintain an armed truce with other bigcompanies. If they want to get bought, they should apply for patentsbecause patents are part of the mating dance with acquirers.Most startups that succeed do it by getting bought, and most acquirerscare about patents. Startup acquisitions are usually a build-vs-buydecision for the acquirer. Should we buy this little startup orbuild our own? And two things, especially, make them decide notto build their own: if you already have a large and rapidly growinguser base, and if you have a fairly solid patent application oncritical parts of your software.There's a third reason big companies should prefer buying to building:that if they built their own, they'd screw it up. But few bigcompanies are smart enough yet to admit this to themselves. It'susually the acquirer's engineers who are asked how hard it wouldbe for the company to build their own, and they overestimate theirabilities. [4]A patent seems to change the balance. It gives theacquirer an excuse to admit they couldn't copy what you're doing.It may also help them to grasp what's special about your technology.Frankly, it surprises me how small a role patents play in thesoftware business. It's kind of ironic, considering all the direthings experts say about software patents stifling innovation, butwhen one looks closely at the software business, the most strikingthing is how little patents seem to matter.In other fields, companies regularly sue competitors for patentinfringement. For example, the airport baggage scanning businesswas for many years a cozy duopoly shared between two companies,InVision and L-3. In 2002 a startup called Reveal appeared, withnew technology that let them build scanners a third the size. Theywere sued for patent infringement before they'd even released aproduct.You rarely hear that kind of story in our world. The one exampleI've found is, embarrassingly enough, Yahoo, which filed a patentsuit against a gaming startup called Xfire in 2005. Xfire doesn'tseem to be a very big deal, and it's hard to say why Yahoo feltthreatened. Xfire's VP of engineering had worked at Yahoo on similarstuff-- in fact, he was listed as an inventor on the patent Yahoosued over-- so perhaps there was something personal about it. Myguess is that someone at Yahoo goofed. At any rate they didn'tpursue the suit very vigorously.Why do patents play so small a role in software? I can think ofthree possible reasons.One is that software is so complicated that patents by themselvesare not worth very much. I may be maligning other fields here, butit seems that in most types of engineering you can hand the detailsof some new technique to a group of medium-high quality people andget the desired result. For example, if someone develops a newprocess for smelting ore that gets a better yield, and you assemblea team of qualified experts and tell them about it, they'll be ableto get the same yield. This doesn't seem to work in software.Software is so subtle and unpredictable that "qualified experts"don't get you very far.That's why we rarely hear phrases like "qualified expert" in thesoftware business. What that level of ability can get you is, say,to make your software compatible with some other piece of software--in eight months, at enormous cost. To do anything harder you needindividual brilliance. If you assemble a team of qualified expertsand tell them to make a new web-based email program, they'll gettheir asses kicked by a team of inspired nineteen year olds.Experts can implement, but they can't design.Or rather, expertise in implementation is the only kind most people,including the experts themselves, can measure. [5]But design is a definite skill. It's not just an airy intangible.Things always seem intangible when you don't understand them.Electricity seemed an airy intangible to most people in 1800. Whoknew there was so much to know about it? So it is with design.Some people are good at it and some people are bad at it, and there'ssomething very tangible they're good or bad at.The reason design counts so much in software is probably that thereare fewer constraints than on physical things. Building physicalthings is expensive and dangerous. The space of possible choicesis smaller; you tend to have to work as part of a larger group; andyou're subject to a lot of regulations. You don't have any of thatif you and a couple friends decide to create a new web-basedapplication.Because there's so much scope for design in software, a successfulapplication tends to be way more than the sum of its patents. Whatprotects little companies from being copied by bigger competitorsis not just their patents, but the thousand little things the bigcompany will get wrong if they try.The second reason patents don't count for much in our world is thatstartups rarely attack big companies head-on, the way Reveal did.In the software business, startups beat established companies bytranscending them. Startups don't build desktop word processingprograms to compete with Microsoft Word. [6]They build Writely.If this paradigm is crowded, just wait for the next one; they runpretty frequently on this route.Fortunately for startups, big companies are extremely good at denial.If you take the trouble to attack them from an oblique angle, they'llmeet you half-way and maneuver to keep you in their blind spot. Tosue a startup would mean admitting it was dangerous, and that oftenmeans seeing something the big company doesn't want to see. IBMused to sue its mainframe competitors regularly, but they didn'tbother much about the microcomputer industry because they didn'twant to see the threat it posed. Companies building web based appsare similarly protected from Microsoft, which even now doesn't wantto imagine a world in which Windows is irrelevant.The third reason patents don't seem to matter very much in softwareis public opinion-- or rather, hacker opinion. In a recent interview,Steve Ballmer coyly left open the possibility of attacking Linuxon patent grounds. But I doubt Microsoft would ever be so stupid.They'd face the mother of all boycotts. And not just from thetechnical community in general; a lot of their own people wouldrebel.Good hackers care a lot about matters of principle, and they arehighly mobile. If a company starts misbehaving, smart people won'twork there. For some reason this seems to be more true in softwarethan other businesses. I don't think it's because hackers haveintrinsically higher principles so much as that their skills areeasily transferrable. Perhaps we can split the difference and saythat mobility gives hackers the luxury of being principled.Google's "don't be evil" policy may for this reason be the mostvaluable thing they've discovered. It's very constraining in someways. If Google does do something evil, they get doubly whackedfor it: once for whatever they did, and again for hypocrisy. ButI think it's worth it. It helps them to hire the best people, andit's better, even from a purely selfish point of view, to beconstrained by principles than by stupidity.(I wish someone would get this point across to the presentadministration.)I'm not sure what the proportions are of the preceding threeingredients, but the custom among the big companies seems to be notto sue the small ones, and the startups are mostly too busy and toopoor to sue one another. So despite the huge number of softwarepatents there's not a lot of suing going on. With one exception:patent trolls.Patent trolls are companies consisting mainly of lawyers whose wholebusiness is to accumulate patents and threaten to sue companies whoactually make things. Patent trolls, it seems safe to say, areevil. I feel a bit stupid saying that, because when you're sayingsomething that Richard Stallman and Bill Gates would both agreewith, you must be perilously close to tautologies.The CEO of Forgent, one of the most notorious patent trolls, saysthat what his company does is "the American way." Actually that'snot true. The American way is to make money by creating wealth, not by suing people. [7]What companies like Forgent do is actually the proto-industrialway. In the period just before the industrial revolution, some ofthe greatest fortunes in countries like England and France weremade by courtiers who extracted some lucrative right from the crown--like the right to collect taxes on the import of silk-- and thenused this to squeeze money from the merchants in that business. Sowhen people compare patent trolls to the mafia, they're more rightthan they know, because the mafia too are not merely bad, but badspecifically in the sense of being an obsolete business model.Patent trolls seem to have caught big companies by surprise. Inthe last couple years they've extracted hundreds of millions ofdollars from them. Patent trolls are hard to fight precisely becausethey create nothing. Big companies are safe from being sued byother big companies because they can threaten a counter-suit. Butbecause patent trolls don't make anything, there's nothing they canbe sued for. I predict this loophole will get closed fairly quickly,at least by legal standards. It's clearly an abuse of the system,and the victims are powerful.[8]But evil as patent trolls are, I don't think they hamper innovationmuch. They don't sue till a startup has made money, and by thatpoint the innovation that generated it has already happened. Ican't think of a startup that avoided working on some problem becauseof patent trolls.So much for hockey as the game is played now. What about the moretheoretical question of whether hockey would be a better game withoutchecking? Do patents encourage or discourage innovation?This is a very hard question to answer in the general case. Peoplewrite whole books on the topic. One of my main hobbies is thehistory of technology, and even though I've studied the subject foryears, it would take me several weeks of research to be able to saywhether patents have in general been a net win.One thing I can say is that 99.9% of the people who express opinionson the subject do it not based on such research, but out of a kindof religious conviction. At least, that's the polite way of puttingit; the colloquial version involves speech coming out of organs notdesigned for that purpose.Whether they encourage innovation or not, patents were at leastintended to. You don't get a patent for nothing. In return forthe exclusive right to use an idea, you have to publish it,and it was largely to encourage such openness that patents wereestablished.Before patents, people protected ideas by keeping them secret. Withpatents, central governments said, in effect, if you tell everyoneyour idea, we'll protect it for you. There is a parallel here tothe rise of civil order, which happened at roughly the same time.Before central governments were powerful enough to enforce order,rich people had private armies. As governments got more powerful,they gradually compelled magnates to cede most responsibility forprotecting them. (Magnates still have bodyguards, but no longerto protect them from other magnates.)Patents, like police, are involved in many abuses. But in bothcases the default is something worse. The choice is not "patentsor freedom?" any more than it is "police or freedom?" The actualquestions are respectively "patents or secrecy?" and "police organgs?"As with gangs, we have some idea what secrecy would be like, becausethat's how things used to be. The economy of medieval Europe wasdivided up into little tribes, each jealously guarding theirprivileges and secrets. In Shakespeare's time, "mystery" wassynonymous with "craft." Even today we can see an echo of thesecrecy of medieval guilds, in the now pointless secrecy of theMasons.The most memorable example of medieval industrial secrecy is probablyVenice, which forbade glassblowers to leave the city, and sentassassins after those who tried. We might like to think we wouldn'tgo so far, but the movie industry has already tried to pass lawsprescribing three year prison terms just for putting movies onpublic networks. Want to try a frightening thought experiment? Ifthe movie industry could have any law they wanted, where would theystop? Short of the death penalty, one assumes, but how close wouldthey get?Even worse than the spectacular abuses might be the overall decreasein efficiency that would accompany increased secrecy. As anyonewho has dealt with organizations that operate on a "need to know"basis can attest, dividing information up into little cells isterribly inefficient. The flaw in the "need to know" principle isthat you don't know who needs to know something. An ideafrom one area might spark a great discovery in another. But thediscoverer doesn't know he needs to know it.If secrecy were the only protection for ideas, companies wouldn'tjust have to be secretive with other companies; they'd have to besecretive internally. This would encourage what is already theworst trait of big companies.I'm not saying secrecy would be worse than patents, just that wecouldn't discard patents for free. Businesses would become moresecretive to compensate, and in some fields this might get ugly.Nor am I defending the current patent system. There is clearly alot that's broken about it. But the breakage seems to affectsoftware less than most other fields.In the software business I know from experience whether patentsencourage or discourage innovation, and the answer is the type thatpeople who like to argue about public policy least like to hear:they don't affect innovation much, one way or the other. Mostinnovation in the software business happens in startups, and startupsshould simply ignore other companies' patents. At least, that'swhat we advise, and we bet money on that advice.The only real role of patents, for most startups, is as an elementof the mating dance with acquirers. There patents do help a little.And so they do encourage innovation indirectly, in that they givemore power to startups, which is where, pound for pound, the mostinnovation happens. But even in the mating dance, patents are ofsecondary importance. It matters more to make something great andget a lot of users.Notes[1]You have to be careful here, because a great discovery oftenseems obvious in retrospect. One-click ordering, however, is notsuch a discovery.[2]"Turn the other cheek" skirts the issue; the critical questionis not how to deal with slaps, but sword thrusts.[3]Applying for a patent is now very slow, but it might actuallybe bad if that got fixed. At the moment the time it takes to geta patent is conveniently just longer than the time it takes a startupto succeed or fail.[4]Instead of the canonical "could you build this?" maybe the corpdev guys should be asking "will you build this?" or even "why haven'tyou already built this?"[5]Design ability is so hard to measure that you can't even trustthe design world's internal standards. You can't assume that someonewith a degree in design is any good at design, or that an eminentdesigner is any better than his peers. If that worked, any companycould build products as good as Apple's just by hiring sufficiently qualified designers.[6]If anyone wanted to try, we'd be interested to hear from them.I suspect it's one of those things that's not as hard as everyoneassumes.[7]Patent trolls can't even claim, like speculators, that they"create" liquidity.[8]If big companies don't want to wait for the government to takeaction, there is a way to fight back themselves. For a long timeI thought there wasn't, because there was nothing to grab onto.But there is one resource patent trolls need: lawyers. Big technologycompanies between them generate a lot of legal business. If theyagreed among themselves never to do business with any firm employinganyone who had worked for a patent troll, either as an employee oras outside counsel, they could probably starve the trolls of thelawyers they need.Thanks to Dan Bloomberg, Paul Buchheit, Sarah Harlin, Jessica Livingston, and Peter Norvigfor reading drafts of this, to Joel Lehrer and Peter Eng for answeringmy questions about patents, and to Ankur Pansari for inviting meto speak.
