Ben Klemens—the blog



 

































31 October 08. In regards to In re Bilski

[PDF version]

The ruling for In re Bilski is out. We had all hoped that it would resolve some long-standing questions on the patentability of software, but instead it is merely the prelude to a debate.

This was an en banc ruling, where the patent on a business method was struck down 9-3. That alone shows us that the Court of Appeals for the Federal Circuit is not a bunch of pro-patent fanatics--especially the majority that aren't former patent attorneys. The judges who do still believe that everything should be patentable are in a dwindling minority. They don't have support at the Supreme Court, they don't have support at the Patent Office, and they don't have majority support in their own building.

OK, enough politics; on to the content of the ruling. The primary contribution of the ruling is to sweep away much of the cruft that has grown around this topic over the last decade, including several ad hoc inventions of the very court that wrote this ruling. It arrives at a simple rule for determining whether something is patentable: is the patent claim for a machine or transformation of matter? If so, then it is patentable subject matter. The overall ruling seems to congratulate itself over sweeping away the cruft and getting to that simple, clear heart of the matter.

But it doesn't answer the question. All the cruft (which I'll discuss further below) evolved as many separate parties tried to answer the question of what is a machine. It's not an easy question. For example, a mental process, which is definitely not a machine, could easily be written down to a hard drive, which definitely is a machine. Is the data-plus-drive amalgam a machine or just a mental process with an off-the-shelf part added to get around a technicality? That's really the billion-dollar question: when does an abstract algorithm become a machine?

Here's the answer from the ruling: “We leave to future cases the elaboration of the precise contours of machine implementation, as well as the answers to particular questions, such as whether or when recitation of a computer suffices to tie a process claim to a particular machine.”

So the ruling explicitly states that it won't answer the key, central question.

On the positive side, the ruling does make progress. First, in throwing out all that cruft, it threw out many of the excuses that people give for software being patentable. Notably, it rejects the assertion that any process of any sort can be patented. The machine-or-transformation rule requires that the process either be fixed in the form of a certain machine, or make a transformation beyond just multiplying numbers together.

Despite claiming that all that matters is the machine-or-transformation test, the ruling also bears in mind many other necessary conditions for patentability, such as the rule that a patent may not “wholly pre-empt” a law of nature or principle or mathematical formula. Also, if you wholly pre-empt a mathematical algorithm within some narrow field of endeavor, the court rules that this is still a pre-emption. I'll have a little more on this below.

There is the State Street rule that an item is patentable if it produces a “useful, concrete, and tangible result,” which this ruling explicitly dismissed. I always hated this rule, because it is entirely vacuous, especially as the State Street court took pains to interpret concrete and tangible broadly--and it's hard to imagine a patent that is worth filing but is somehow not useful. If your patent draftsman can't write up your patent to be useful, concrete and tangible, it won't be hard to get a new one who can. But we now need to determine patent-eligible subject matter by the machine-or-transformation test presentend in this ruling, which is not vacuous, so State Street's blank check for any process patent has been revoked.

The ruling pointed out that U.S. law has never had a “technological arts” test for patentability. They did the right thing here, because the phrase is impossible to define precisely. We're having enough trouble defining the word machine; there's no benefit to shifting the debate to the more vague and subjective word technology. Similarly, there's no point in defining an exception to patentability like business methods, the ruling states, because it'd be a waste of time to try to pin down a definition of the excluded field.

On the key question of when information becomes a machine, the ruling does provide some clues: “First, [...] the use of a specific machine or transformation of an article must impose meaningful limits on the claim's scope to impart patent-eligibility. [...] Second, the involvement of the machine or transformation in the claimed process must not merely be insignificant extra-solution activity.” For more on extra-solution activity, have a look at my law review article (PDF) that focuses heavily on the idea.

In the earlier segment on pre-emption, we got what I take as a sample application of the first of these rules: in the algorithm in the Supreme Court's Benson case “... the limitations tying the process to a computer were not actually limiting because the fundamental principle at issue, a particular algorithm, had no utility other than operating on a digital computer. [...] Thus, the claim's tie to a digital computer did not reduce the pre-emptive footprint of the claim since all uses of the algorithm were still covered by the claim.” The ruling gives other examples, many taken from the Supreme Court, of cases where the computer does or does not limit the algorithm in a sufficient manner to restrict patentability (notably in Flook, where an alarm to measure hydrocarbon temperature is still not sufficiently tied to the machine to merit patentability).

The Benson example, which talks about limiting scope and pre-emption, and the machine-or-transformation rule regarding how “the use of a specific machine ... must impose meaningful limits” bear much in common. They indicate that if a computer is merely being used as a transparent medium for the algorithm, then the patent is an attempt to patent and wholly pre-empt an abstract algorithm, and is not a claim on a machine. If the computer is not merely a transparent medium and is a limited use of the algorithm, then the patent is for a machine.

Did my rubric in the last paragraph sound vague? As stated, it is, and it's all this ruling leaves us. We have moved a little forward in the debate, because the ruling has thrown out things like the “useful, concrete and tangible” rule and done what it can to shut down the argument about a technical arts test. It thus allows us to focus a little more closely on the question of when an abstract work becomes sufficiently fixed to a physical environment that we can call it a new machine. It provides a few hints about how that measurement should be made, in the form of a repetition and not-overly-narrow interpretation of pre-emption rule and an emphatic repetition of the rule from two Supreme Court rulings that extra-solution activity does not make something patentable. But the key question itself remains unanswered and still open for debate.


[link][a comment]


15 July 08. How strict constructionism can be judicial activism

[PDF version]

Today's episode is a guest blog by Mr. BK of Baltimore, MD

This is a note on the term `judicial activism', which is misused in subtle ways among pundits and politicians.

The key to how it is misused is the ambiguity of the terms liberal and conservative. I count five (5) distinct uses of these terms.

The first three are familiar to everybody. There's the liberal/conservative scale regarding change in general, where the L team is forward-looking and the C team seeks stability. There's the social scale, where liberals believe people should be left to do what they want, and conservatives seek a social order reminiscent of Norman Rockwell paintings. There's the economic scale, where liberals believe some social services are necessary, and conservatives seek smaller government.

These three scales are only tenuously related. It is easy to find futurist social conservatives, social liberals for smaller government (aka libertarians), and any other combination of the above. But, with only the words liberal and conservative used for all three axes, there's a strong--and clearly false--implication that one who is liberal on one axis is liberal on the others.

That said, let us move on to judges. Judges are often described as constructionists or activists, as if there is a single axis along which we measure judges. But as with liberal/conservative, it confounds a couple of concepts and just creates confusion. So, let's make some definitions.

Constructionism
There are two components to a law: the statute in the Constitution or as passed by Congress, and the interpretation of the statute by courts who had to contend with the law. One school of thought, strict constructionism, contends that one should focus as much as possible only on the statute as written, rather than subsequent interpretation. Congress wrote what it darn well intended the law to be, so why should later judges and pundits modify that intent?

The constructionist view bears much in common with the neoclassical economist's viewpoint, that people are very rational and very capable of forseeing the future. To the extent that this is correct, the constructionist claim (that Congress wrote what it intended the future to look like) works.

I work with patents, and patents are an excellent example of how constructionism and the hyperrational assumption can go horribly wrong. Thomas Jefferson wrote what is now 35 USC §101 (inventions patentable), and it hasn't even been looked at since 1952. So: did the 1952 Congress, or Thomas Jefferson, intend that web page designs should be patentable? Even the Psychic Hotline would have difficulty with such a question, yet a strict constructionist has a simple answer: yes, they did intend so, because they would have said so otherwise.

The alternative is to look at more recent rulings and try to conform with those. My impression is that this is the modal type of judge: they try to rule in conformance with the law, but that includes equal measures constitution/statute and recent rulings. Let us call this the developmentalist approach; some call it the activist approach. The language typically used sets constructionist = conservative and activist = liberal.

As an aside, the constructionist view toward the U.S. Constitution is often characterized as interpreting the constitution the way the Founding Fathers intended it. But this is an incorrect phrasing. Jefferson again: “No society can make a perpetual constitution, or even a perpetual law. The earth belongs always to the living generation.” [Letter to James Madison, 6 Sept 1789] The statement `I am a strict constructionist, because I interpret the law the way a set of developmentalists did in 1776' is somewhere between incoherent and ridiculous. Rather, the sane strict constitutional constructionist generally shoots for a direct reading of the words as written, outside of the context of colonial times.

Stare decisis
This is legal Latin for “to stand by things decided”. That is, if judges past have decided that the law of the land is X, then ya don't change it to Y unless there's a darn good reason.

Different judges interpret the phrase darn good reason differently. Some overturn past rulings at the drop of a legal hat; others steadfastly stand by the past rulings, and just mumble something about `it's a bad law, but it's Congress's job to change it' in rulings that they aren't happy writing.

There are two pairs of terms used to describe a judge's attitude toward stare decisis. The first is liberal/conservative and the other is constructionist/activist, and once again, both pairs of terms don't correspond to any of the above uses of these terms.

There are four possible combinations of liberal/conservative in the context of statute and stare decisis, and it's worth going over all four, because they reveal an important asymmetry.

It may happen that the law as currently interpreted differs significantly from the law as written. This is common for a law written decades ago, due to simple drift in conditions and legal understanding. The first option in this case is to be a statute liberal and a stare decisis conservative. That is, a judge could be a conservative in the sense of maintaining the status quo.

When there is a difference between the status quo and the original statute from times past, it is impossible to simultaneously be conservative with regards to both. Of course, this doesn't keep many judges from trying.

The next possibility is for a judge to be liberal with regards to both statute and stare decisis. Such a judge really is just making up the law. You won't find a judge anywhere who claims such a position, though there's endless debate as to whether some judges act like this. Therein lies the asymmetry between liberal and conservative: conservative on both scales is OK but usually impossible, but liberal on both scales is an abuse of judicial power.

The next option is to be a statute liberal and a stare decisis conservative. There are people like this in many contexts: folks who insist that the U.S.A. once had a decidedly Christian government (a claim that is itself up for dispute) and therefore the present government should be devoutly Christian as well; folks who insist that the only good music is the kind they heard in high school; folks who insist that all families must consist of a mother, father, and at least two children because that's how it had to be on the farm. Such people are radically liberal, in the sense that they oppose the status quo in favor of something different, which happens to have been the status quo at one point in the past.

Judges of this type are often called judicially conservative. Yup, a judge who rules for changing the status quo when faced with a conflict between statute and rulings is called judicially conservative, and a judge who prefers to maintain the status quo is typically called judicially liberal. It's things like this that make people learning English as a second language hate it so much.

To revert to the means of judging in times past would be a radical, jarring change from the present. (Painting credit: Raphael: {<EM>The Judgement of Solomon</EM>}, c. 1518-19)
Figure One: To revert to the means of judging in times past would be a radical, jarring change from the present. (Painting credit: Raphael: The Judgement of Solomon, c. 1518-19)

Patent law is a good example of judicial conservatism/status quo liberalism. The circuit judge who decided that software and business methods should be patentable (Judge Giles Rich) was very vehemently constructionist in citing statute and reading it as literally as possible. As such, he was massively activist, because he overturned a century's worth of stare decisis, including several rulings from the Supreme Court.

You know I am not happy with Judge Rich's ruling, but there are other cases of activist constructionists, the most salient being those who ruled in Brown v Board of Education, whom we all love to death. So even after we have acknowledged that the scales of liberal/conservative with respect to statute and liberal/conservative with respect to stare decisis are entirely different scales, and after we've pegged a judge on both, we still won't know whether their rulings are liberal/conservative with respect to the social and economic scales that people actually care about.

Gay marriage
The term activist judge has been bandied about by certain individuals, invariably as a derogatory term, but without clarifying to which of the above two sometimes contradictory definitions the speaker is referring. But the confusion is typically deliberate, and implies that any activism in the sense of interpreting laws based on judicial understanding must be of the radical form of arbitrarily revising law.

The activism question often comes up in judicial hearings as well, where judges often attempt to characterize themselves as strict constructionists, implying that this is a good thing. But it seems preeminently clear that a good judge makes an effort to balance statute and recent rulings in every situation. The Constitution just doesn't say anything about computer-generated pedopæliac images, so for a judge to claim that he considers only the constitution in deciding such an issue is to say that the judge feels at liberty to just make stuff up.

Not to accuse President Bush of simplistic thinking, but to say that any judge that does not strictly follow statute is rewriting the law is simplistic. Such a claim only works when the law as written is entirely and perfectly appropriate to all situations, even decades later--and remember that a case appears before a higher court only when there is some sort of open question, ambiguity, or controversy about the law as written. Thus, any high court judge that isn't braindead is an activist in the first sense of re-interpreting statute as written; if we insist that that means activist in the sense of inventing law, then we can only conclude that all judges are activist in the derogatory `hijacking the law' sense used by folks such as the guy linked above.

But why be abstract when we have an easy example? The term `activist judge' is the term preferred by people complaining about gay marriage. The big ruling in the gay marriage issue was Goodridge v Department of Health, which was the ruling in Massachusetts that allowed same-sex civil unions--and did so via an allegedly strict constructionist reading of the equal protection clause in the MA constitution, no less.

With regards to statute, there is clearly ambiguity because nobody ever bothered to strictly define the meaning of marriage, just as Jefferson didn't specify whether web pages should be patentable. You could ask what the word marriage meant in 1776, in which case you'd probably find gay marriage was not intended by the law--and neither was marriage between whites and nonwhites.

As for overturning stare decisis, I'm no expert on MA marriage precedent, but giving a casual look `round, I am unable to find claims that Goodridge was in contradiction of past court rulings. For so many claims that this is an activist court, you'd think somebody would find the ruling that they were supposedly contradicting. Rather, marriage law in the U.S.A. has been a slow slide toward disaggregating marriage into a series of social services (especially given the strict interpretation of the “no establishment of religion” clause in the Constitution), and Goodridge fell right into that by interpreting the civil union as such a bundle of social services.

So what we get here is that the court in Goodridge wasn't actually activist at all in either the statute or stare decisis sense. It read directly from the MA constitution using a plain English understanding of the language about equality under the law, and did not seem to disagree with past court rulings. So we conclude what many of you have probably been thinking all along: that the term `activist judge' in this context really is just a polysyllabic way of saying `socially liberal'.

Now, there's a specific reason for conservative rhetoricians to confound all five axes and claim that liberal on one means liberal on the other four, which returns to the asymmetry discussed above: if a judge is liberal on both judicial axes at once, then that judge is just making up law.

This is the crux of the implicit argument in the term `activist judge'. If we start with the false premise that a judge who is liberal on one axis is also liberal on all four other axes, then we get the false conclusion that all socially liberal judges are just making up the law, and so the only good judge is a socially conservative judge.



[link][a comment]


21 April 08. The legality of absinthe

[PDF version]

I finally got some absinthe last week.

There are many types. Like many liqueurs, it is basically just an infusion of a few pleasantly-flavored plants in grain alcohol. Being grain alcohol, it is about 50% alcohol by volume, but you water it down. Now, when watering it down, it goes from grain alcohol clarity to the opacity of milk. This is fun, and there is a little ritual built around the process, typically involving a sugar cube perched on top of the glass. Having a ritual associated with a social beverage is a big perk, and probably has had some hand in making absinthe so popular. I also have a full tea set, and get great enjoyment from the process of pouring water from the kettle to the teapot to the ocean of tea to the sniffer cup to the tea cup.

But you know the real reason absinthe is famous: wormwood, which has a chemical known as thujone. I have no idea what the effect of thujone is on the human brain; I get the impression that nobody else does either. Some commenters have said that it's something of a stimulant, so absinthe is a bit like having a black Russian (i.e., vodka + coffee), but I'm not sure if even this much effect has verifiable evidence behind it. As above, it's 50% alcohol straight, is watered down, and is heavily spiced, which all adds up to being able to drink a lot of alcohol without tasting it. Having such an easy means of drinking quite a bit of booze without knowing it is already enough of a recipe for loopy behavior, without recourse to obscure chemicals.

Absinthe was a scapegoat of the temperance movement, and was banned in both the USA and Europe. I don't have the exact history--the USA obviously spent the `20s banning a lot more than just absinthe--but as we rolled in to the new millennium, thujone was banned (sort of) in both the USA and EU.

You may have read that absinthe is now legal in the USA. What changed in the law?

Nothing. Nothing at all.

My favorite textbook on labor theory.
My favorite textbook on labor theory.

The most credible source I've found to this point has been an egghead-oriented site on mind-altering substances named Erowid, and this interview with a lawyer involved with an effort to import absinthe into the USA.

From what I gather, thujone is banned, but the test used for checking for thujone is not very sensitive, which opens the question of what to do with beverages with low thujone levels; the decision was eventually made that they are OK.

Beyond the thujone issue, the Alcohol and Tobacco Tax and Trade Bureau (TTB) felt that the word absinthe was a drug term, and therefore could not be used to sell a product. This was basically an arbitrary opinion of the bureau, and was eventually arbitrarily reversed.

With those roadblocks eliminated, absinthe was once again legal for import and sale.

The “law”
The anecdote is an interesting demonstration of how often law is entirely unclear. We like to think that there's a list of regulations out there, and they tell the world's bureaucrats what to do or not do, and it's all very simple. In most cases, it really is that simple. But then there are those other cases where it's hard to tell exactly where to draw the line, or what authority the individual bureaucrat has.

After all, Congress (or the head of any organization) can't possibly dictate everything that every bureaucrat all the way down the chain will do with his or her hands. In any organization, the top sets broad rules, and grants limited authority to the individuals at the lower levels who will make up whatever needs to be made up along the way.

The patent office is especially prone to the conflict between top-down law and bottom-up rulemaking, because patents are a technical field and most of the distinctions to be made are fine line-drawing rather than night-and-day. Under the clearest of rules, we would expect that some applications would still be considered novel and non-obvious by one examiner and rejected by another.

Further, patent law is, more than other fields of law, decided by judicial rulings, not by the U.S. Code. Congress hasn't bothered to touch the law for patents in any significant way since 1952, and you can see why from the massive, ineffective effort that has been the Patent Reform Act of 2008. Instead, we have centuries' worth of rulings, each about a single test patent (sometimes several), and a resulting rule based on the test patent regarding how to draw the line between the patentable and unpatentable. Judges try to be careful, but it's common enough that two rulings will contradict each other.

Naturally, there are many efforts and mechanisms to create standardization, and there is largely one patent law. For example, there is a Manual of Patent Examination Procedure (not written by anybody in Congress or the courts) that examiners are expected to follow with perfect uniformity. But to a great extent, the concept of one patent law is just a convenient fiction, and every examiner has a slightly different concept of what's going on. As with the change in opinions regarding absinthe, things can go from illegal to legal with just a change of opinion by a bureaucrat somewhere along the chain. That's how it is with a law built from the ground up on judgment calls like what is novel and non-obvious, and how it always will be.


[link][no comments]


7 April 08. Independent invention

[PDF version]

Hi. I took a break to handle a few important matters, like filing a brief in In re Bilski, catching up on numerous important pieces of reading, and drinking lots of absinthe. More on that last item next time.

Independent invention is not a valid defense against claims of patent infringement.

And that, dear reader, is the Achilles's Heel of patent law.

This is not to say that all of patent law will inevitably suffer a downfall to the Paris Convention?, but that the independent invention problem is an important consideration that advises when patent law is or is not appropriate.

By independent invention, I mean that a second party invents whatever a first party had patented, without knowledge of the first party's activity. The second party, without knowing that it had done anything wrong, is liable for patent infringement.

From this NYT article: “It's rare that you've got a major breakthrough that wasn't developed by multiple people at about the same time,” said Mark Lemley, professor of intellectual property at Stanford Law School.

On both an ethical and economic level, liability for independent invention is a problem. Imagine that you have put great effort into developing a new technology, start marketing it, and then get a letter in the mail telling you that another party has already obtained a patent on your work, and you now owe them money for putting out the product you've developed. There you were, minding your own business, not copying anybody, and now you suddenly owe a stranger perhaps millions of dollars. Ethically, it is hard to justify the justice of the situation.

In terms of promoting progress, if you know there's some chance that you could be doing extensive research that winds up in a product that you not only can't use, but which is a legal liability, then that's an incentive to avoid exploring new technologies.1

There's an incentive to patent inventions that are likely to be re-invented, because then you can sue the latecomer for $$$. But why is the government providing ex post research grants for inventions that are going to be re-invented by other parties without any government help? Patents can promote progress, but not via liability due to independent re-invention.

There are common counterarguments to why the problem of independent invention isn't really a problem. The first is that such a problem should be marginal: one could think of the still-extant restrictions on what is patentable (that it be novel and non-obvious) as basically requiring that the invention be unlikely to be independently re-invented. The obviousness standard isn't actually keyed to minimizing independent invention, and I haven't seen anybody talk about making that happen, so that doesn't get us far: examples of independent re-invention abound.

The other main argument is that the independent inventor shoulda known. We assume that patents are in the public record, and therefore everybody in the field should be aware of the state of technology. If you are a researcher, it is basically mandated by law that you check in with progress at the Patent Office on a regular basis. Corporations in patent-relevant fields are well aware of this and accept it as a cost of business.

The presumption of public knowledge means that there is no opt-out from patent law. If you are in a field that has patents, you can't just say you don't care about patents, won't apply for your own, and won't imitate products with a patent number stamped to their underside. That's not enough to get you clear of the USPTO's influence, because you could still independently invent a patented item and thus face liability.

Now, in a prior episode, we saw how little actual practitioners look at existing patents. They're legal documents written for other lawyers, not practitioners. But in most fields, the fiction that every technologist is following the patent literature is evidently not particularly painful.

Oh, and patents are kept secret in the USA for 18 months. That way, if an applicant wants to change its mind and keep its invention a trade secret, it can. There's a never-ending debate about this one. On the one hand, this option makes for a hard choice if you don't get an initial reply from the patent office until after the 18 months have passed (which is perfectly possible given the current backlog). On the other, some think that it's silly that we have to wait even 18 months, because in 17 months, a company could invent something and begin marketing it widely, and then get hit with liability (or at least major business hassle and costs) even though the company followed 100% of the rules. That is, even this relatively short secret period exacerbates the problem of independent invention.

By the way, the disclosure rule is an attempt to harmonize with other countries' laws, so if you promise to not file for patents in other countries, you can have the USPTO keep your invention secret right until the final patent granting perhaps several years in the future.

So liability from independent re-invention is an unpleasant side-effect of the system, that is hard to justify both ethically and economically. If you could design our patent system all over again, you'd probably want to take care to ensure that the likelihood of independent re-invention is minimized.

Millions of independent inventors
What if the `industry' is everywhere? Literature is written by lots of people all over the place. There's no literature convention where everybody has to meet up. It's downright silly to expect that every novel author is abreast with the works of every other, even within a narrowly-defined genre. If storylines were patentable, we would certainly see a long stream of lawsuits against defendant authors who had never heard of or read the work of the plaintiff author. There simply aren't enough patent attorneys to handle the amount of due diligence that would be required to verify that all of the stories being written don't infringe an unknown patent.

What if there were only twenty storyline patents, and they were all really awesome? Every one of them is stupendously inventive. Well, that wouldn't change much, because every author would still have to check in with a patent attorney to see if their new work happens to have hit upon something patented. The attorneys would have an easier job, and could stamp 90% of new storylines as non-infringing pretty quickly, but the costs of a quick consultation with an attorney for every story published in the USA would still be pretty astronomical.

The result, if storylines were patentable, would probably be that most people just wouldn't bother to clear their work, and would just hope the risk goes away. Some number of them would eventually independently infringe even the most creative storyline, and the patent-holders would have an easy time of finding infringing stories at the local bookstore. Lawsuits would ensue. Lots of people would question why it is that we have literature patents anyway. Patent attorneys would repeatedly state that patents foster innovation, period, and refuse to hear discussion otherwise.

That is that state of software patents.

The software industry is indeed massively decentralized: just about every corporation in the country has an information technology department that is producing software that is sufficiently innovative that it could be infringing a patent. How do we know? Because there have been software patent infringement suits against the Green Bay Packers, OfficeMax, Walgreen's, Linens 'n' Things, the Washington Post, and a host of others; see the End Software Patents project's litany of lawsuits.

There are a number of potential resolutions, none of which point to promoting the progress. The `shoulda known' approach would be to say that if you have a web site, then it's your obligation to hire a patent attorney to clear the site. Sure, there are a hundred million web sites out there, but that just means that we need to have more patent attorneys.

Another resolution would be to centralize all production. The trouble with both the literature and the software market is that the only qualifications you need for trying to write something is a cheap PC and a half-decent facility with language. Here's a father telling the story of how he helped his 6-year old write a new computer game. We could fix the independent invention problem by dictating that all work must be done by a centralized organization, maybe IBM, and everybody else may have scope only to use the software in limited, non-innovative ways. IBM offers indemnification and has the legal staff to take care of things, and as long as users everywhere else don't get a sudden urge to experiment, everything is fine.

This may sound like a joke, but it is to some extent the logical conclusion, and where we could wind up. As above, not everybody is going to get sued. Instead, some get sued, and everybody else gets scared, and eventually fires their IT department and re-hires it as an IBM-backed consulting company that can offer indemnification. This outcome certainly serves the legal industry and IBM well, but I'm not convinced that it promotes progress or is what many programmers actually want.

I really don't know how to resolve the problem of independent invention with respect to software, literature, or other activities that really just require a brain and a writing implement. But it shows is that the mantra “patents promote innovation” needs qualification. For example, there are few people who truly believe that patents would promote innovation in storylines or artistic works, and there are many contexts where the trouble of independent re-invention swamps the research benefits provided by a limited monopoly. It is much more reasonable to say that patents promote progress in some contexts, but not all.


Footnotes

... technologies.1
There is the related idea of the patent race, that if you think others might develop a technology, then you have a strong incentive to work harder and develop it first. The hypothesis has many variables that are hard to measure, meaning that I know of no empirical studies that find real-world evidence that patent races help [add `em in the comments if you do]. The theoretical literature is split, because having two parties duplicating effort on the same problem is not necessarily socially efficient.


[link][a comment]


10 March 08. A comparison of copyrights and patents

[PDF version]

The average person on the street will tell you that copyright is for printed creative works like text, paintings, or musical compositions, while patents are for functional physical devices and processes.

But here in the modern day, that basic intuition just doesn't match up with the reality of U.S. law. The two regimes have been growing closer over the last few decades, which means that the things that some think of as the key distinction between the two are not much of a distinction anymore, while other features are of increasing importance.

Subject matter
There are an awful lot of types of work that could be covered by either system. In the absurdist plays section of my bookshelf, I've got a copy of Endgdame by Beckett. Although it is plain text, it is a series of instructions for actors, comparable to a series of instructions for people on an assembly line. If you'd like to excise words entirely, the publishers added a short bonus play, entitled Act without words: A mime for one player. The stage directions are text, but are entirely about how props or the actor should move, to achieve a final goal which I take to be demonstrating learned helplessness to the audience, but I'm sure Beckett had something much broader in mind.

In fact, for any plain information, there is some sort of physical medium upon which the information is recorded, be it a piece of paper, a computer's hard drive, or your brain's neurons. In the last decade or so, the U.S. courts have ruled that that physical medium makes the information patentable. The key ruling on this one is In re Lowry, which is for a data structure on a computer. “More than mere abstraction, the data structures are specific electrical or magnetic structural elements in a memory.” There used to be a printed matter doctrine that states that you can't patent a surface (paper, a ribbon) just because the information written thereon is useful; Knight (2004) argues that under current law, the printed matter doctrine has been eliminated, and I agree.

Right now, the courts are scrambling to draw some sort of line to say that tax loopholes, plays, and musical compositions are not patentable, and I'm sure that six months from now such a line will exist. But in the mean time, for things that could be expressed as text on some sort of medium, the difference between copyrights and patents largely lies elswhere.

Duration
Patents provide a limited monopoly for twenty years from the date of filing. The exact duration of copyright has many conditions and exceptions, so here I'll just say that it's in the ballpark of a hundred years. As we'll see below, it somewhat makes sense that patents should expire much more quickly, because they offer greater power.

The economic life of most works, fictional or functional, is under twenty years anyway, so for most works the duration of both types of protection is the entire economic life of the product. Many works do shine for much longer than twenty years, but for most products we must again look elsewhere for the big difference between patents and copyright.

Scope
I've seen many people state that copyright covers only the literal text of a work, while patents cover the broad idea. Both halves of that statement aren't quite correct. When something could be covered by both copyright and patent, the scope of the two forms of coverage are increasingly convergent.

On the copyright side, consider fan fiction. Many people enjoy writing stories based on their favorite fictional characters, usually involving some number of them fornicating with each other. They use the name and characteristics of characters from a published work, but every other word in the text is entirely their own. There is minimal literal copying, if any. The lawyers call this sort of thing a derivative work, and it violates copyright as much as would a direct copy of the original work. See the Chilling Effecs Fanfic FAQ for examples and details.

So the scope of copyright is a little more patent-like than it is often made out to be. In a software context, if party B sees party A's code written in FORTRAN, and then produces a similar program in Java, then party A has decent odds of claiming the Java code is a derivative work and therefore infringing copyright. This sort of dispute sometimes crops up when employees jump from one company to another.

Meanwhile, patents don't cover pure ideas, but their implementation. This is good because it allows future designers to re-implement the same idea using new methods; this is known as inventing around a patent, and is generally credited as a positive side-effect of patents, because future designers are forced to think hard about a problem where they otherwise would have just copied the other guys. So patents are not for ideas, but the current implementation, which sounds roughly like the scope of copyright.

For software, the implementation is very broadly construed, because courts have ruled that the actual process of writing code is a “mere clerical function.” It's not often that a ruling insults an entire industry of people; in this case doing so even creates problems, because calling working programmers a bunch of code monkeys and denying any creativity in code design leaves little latitude to work around a software patent. But in theory, if a patent should “wholly pre-empt” a mathematical algorithm, then the courts say that it shouldn't be granted.

So for a product that is primarily expressed in text, the differences in breadth of coverage between copyright and patent are not nearly as large as many make it out to be. Neither is intended to cover solely the literal text as typed by the original author, and neither is intended to cover the underlying ideas.

Registration
Copyrights are automatic: the text you are reading is copyrighted by the simple act of my writing it down. The Library of Congress offers a registration system, but works aren't examined in the process, but just catalogged and possibly shelved.

The Patent Office, however, tears patent applications apart. The examiner is expected to find the prior literature, determine exactly how the application differs from the prior literature, determine whether the application's improvement is obvious to a person having ordinary skill in the art, and otherwise run the application through a long series of tests.

Finally, we have a night-and-day distinction between the two systems. The distinction means that copyrights are going to be a lightweight system, in the sense that the Library of Congress has one not-very-large department devoted to administration, gaining a copyright is literally costless (and full registration is on the order of fifty bucks), and no lawyers are involved. Patents are at the other extreme: the number of patent examiners is constantly growing, patent applications go through several layers of examination, and you can maybe do it for cheap, but don't expect to get anything useful and enforceable out of the process without the services of an attorney.

The difference in registration processes can partly be explained by the different powers granted by a patent or a copyright, which brings us to one more major difference...

Independent invention
You can only be held liable for copyright infringement if you have seen a work and copied it. For most copyrighted works, it's pretty easy to identify direct copying anyway--if you're writing fiction about Superman and Wonder Woman going on a date, then you've definitely been directly exposed to both characters as drawn by DC Comics. If you worked at a company that writes programs for dentist's office administration, and then quit and start up a business that sells a program for administering dentist's offices, then your old employer would have an easy time demonstrating your access to the original progrm if the need arises.

Conversely, independent invention is not a valid defense against claims of patent infringement. If your work matches a patent that you've never seen or heard of, then you can be sued. It's that simple.

So copyright infringement requires direct imitation, while a patent holder can sue people who have never seen the patent or heard of the patent holder. This has major effects, which I'll pick up on next time.

Please note: full references are given in the PDF version


[link][no comments]


27 February 08. Disclosure

[PDF version]

One of the oft-cited stories behind patents is that they are a contract for disclosure. The inventor could have kept her1 invention a secret, but the government offers her a deal that if she reveals her technology so others can learn from it, then the government will give her a limited monopoly for twenty years (from the date of application).

This is not a fundamental reason for patent law. It's icing. The fundamental reason is the ex-post research grant I'd mentioned earlier.

Which is a good thing, because the disclosure story really only works in theory.

Those of you in technical fields can verify this through introspection: when you last worked on a project, did you first check the journals or the Patent Office? Did it even occur to you to learn from patents? You can search sites of technical working papers, like arxiv.org, for patent references, and you'll find a handful, but a handful of references out of hundreds of thousands of patents is not very impressive. If the patent record were a journal, the journal's impact factor would be right around zero.2

Firms in all industries tend not to search patent databases for technological instruction. Arora et al.(2003) state that “patent disclosures appeared to have no measurable impact on information flows from other firms, and therefore no measurable effect on R&D productivity.” [p 17] Arundel finds that “a consistent result in survey research on the use of patent databases is that they are among the least important external information sources available to firms.”

Software: even less disclosure
Campbell-Kelly (2005) wrote a paper whose title typifies pro-software patent research: “Not All Bad: An Historical Perspective on Software Patents.” His primary argument for why software patents aren't all bad is that patents set a disclosure requirement.

But disclosure is even less functional in software than for general industries.

For example, Campbell-Kelly cites the algorithm for LZW (Lempel-Ziv-Welch) encoding as a success story for patents. However, as Campbell-Kelly notes, Terry Welch had published the algorithm in a peer-reviewed journal within a few months of applying for the patent, but years before receiving it Welch (1984). If Mr Welch had truly wanted to keep the LZW algorithm a trade secret should patent protection be unavailable, then he would have waited for the ink on his patent application to dry before publishing the algorithm.

In Northern Telecom v. Datapoint,3the CAFC ruled that writing source code--in fact, authoring anything more detailed than the broad flowchart describing the overall logic of the design--is a “mere clerical function.” So the court has stated that the requirements for patents (what they call the enablement requirement) are not much more than you can find by poking at a copy of the program for a while.

Mann and Sager (2005) interviewed software technologists and found that the apathy toward learning from patents revealed by the general surveys is evident in software as well: “...none of the startup firms to which I spoke suggested a practice of doing [patent] searches before beginning development of their products.” [p 1004, italics in original.]

And gosh, have you ever looked at a patent? It is a terrible way to teach others. I've read many a debate where some people having ordinary skill in the art look at a patent and declare it obvious and generally idiotic. Eventually, a patent examiner pipes up that the document is written in a technical language that laypeople outside the patent field are not trained to read. Looking at any patent, you can see that this is definitely true. It'd be difficult indeed to have a document that is both a legal declaration of a limited monopoly and a technical document for teaching fellow engineers. The teaching part takes a definite backseat to the legal declaration.4

So the disclosure story is an interesting theoretical anomaly: it makes intuitive sense on a broad scale, but we know both from looking around and from careful analysis of survey data that the broad intuition just doesn't play out here in the real world. And yet people keep referring to it, because the broad concept makes sense.

So I won't mention disclosure further, but will stick to the other, actually functioning, stories for why patents exist.

Please note: full references are given in the PDF version


Footnotes

... her1
Following a suggestion by Thompson (2001), I have chosen the gender of representative agents in this book by flipping a coin.
... zero.2
Now, patents themselves cite prior patents all the time. Patent attorneys, as opposed to inventors, are no doubt learning up a storm from the patents they are reading all day long. But to say that patents are intended to educate attorneys writing other patents is ingrown logic, to say the least. There's still that disconnect between the people writing and learning from patents and the people who are at work developing new technology.
... Datapoint,3
908 F.2d 931, 940-41 (1990).
... declaration.4
It is possible that after a patent is granted, the author then proceeds to produce a journal article. But you need a story beyond economics for why the journal article is getting written. No patent is a perfect monopoly, so the more information you give your competitor, the more power they have to invent around your work. So there are stories where disclosure plus journal publication work, but they partially advocate against the entire concept of patents (because the typical inventor really wants to show off and score publications) and I've seen no evidence that they dwarf the lack of evidence of benefits from disclosure up there in the main text.


[link][2 comments]