Letter to the Patent Office from Professor Donald Knuth (circa 1994) | Hacker News Reader

Letter to the Patent Office from Professor Donald Knuth (circa 1994)

pluto.it

125 pointsfogus31 comments

"Therefore the idea of passing laws that say some kinds of algorithms belong to mathematics and some do not strikes me as absurd as the 19th century attempts of the Indiana legislature to pass a law that the ratio of a circle's circumference to its diameter is exactly 3, not approximately 3.1416."

now that is a real gem

Seems like the Indiana bill had a few more options than pi = 3. :)

http://www.straightdope.com/columns/read/805/did-a-state-leg...

> If present trends continue, the only recourse available to the majority of America's brilliant software developers will be to give up software or to emigrate.

There is one other alternative which is available to anyone not selling their software... ignore patents. I'm not sure if it's a legal option, but it certainly seems to be the common case. I know I don't bother to stop and figure out if an idea is patented before I implement it for my own uses.

It's possible that the changes from software patents haven't trickled down far enough yet. Currently it's just the big players, but once a patent troll finds a way to extract money from local software shops the landscape will change dramatically. Ignoring software patents works as long as everyone does it; it just takes one organization to bring it all down.

Many companies function in this way and problems do not typically arise in a day-to-day sense in most cases.

However, in any due diligence situation (e.g., funding or acquisition), patent problems are likely to be unearthed and can then be very serious unless a ready workaround is at hand.

Software patent disputes are uncommon in the little league, but unfortunately not uncommon in the big boys' league. It wasn't more than a few weeks ago that Microsoft was ordered to stop selling Word due to software patents.

I'm afraid patents are too strong for that. It's illegal to infringe on patents even for your own use.

I'm aware that it may not be legal. My argument was that it's also unenforceable.

Even when a patent claim tries to tie a mathematical algorithm to a machine, the legal tests get all convoluted.

For example, in Ex Parte Gutta (BPAI 2009) (decided August, made precedential December, 2009), the current USPTO test for patentability in such a case is set out:

"(1) Is the claim limited to a tangible practical application, in which the mathematical algorithm is applied, that results in a real-world use (e.g., 'not a mere field-of-use label having no significance')?

(2) Is the claim limited so as to not encompass substantially all practical applications of the mathematical algorithm either 'in all fields' of use of the algorithm or even in 'only one field?'"

If the answer to either question is "no," the claim fails.

In that case, the claim indicated that a processor (working with memory) is configured to identify a "mean item" whose symbolic value minimizes the variance of a set. (See http://www.patentlyo.com/patent/2009/12/patentable-subject-m... for a visual depiction)

The BPAI found that the claim failed the first prong by not limiting itself to any "tangible practical application in which the mathematical algorithm is applied that result in a real-world use." Likewise, the claim failed the second prong because it "encompasses substantially all practical applications" of the algorithm. "That is we are unable to identify any other practical application [of the algorithm] outside of the broadly defined claim."

Any time someone tries to grab a monopoly on an algorithm, there is a real danger to innovation. Sometimes (as here) common sense prevails but often it does not.

I'm not really sure how this has existed for 16 years and I haven't read it.

This is the best argument I've ever heard against software patents.

Indeed! Although, with the recent rise of Open Manufacturing, Knuth's distinction between physical and abstract patents is looking a bit dated though?

Thanks for the find fogus!

Imagine that IT was a more mature field, where basic techniques formed a body of knowledge, a "state of the art" (perhaps in a "text-book" for IT), that is known to all "qualified" practitioners. Then, the few innovations that do occur would be truly new (and it'd be immediately clear whether they are or not). A problem with this maturation is the complexity of software, and the many ways to implement the "same" thing. Some fp academics (Smullyan) have gone some substantial way to categorizing different techniques, such as a list of "the" combinators, and ways to determine whether some code really implements that combinator or not. A theory of computation.

As it is, we keep "reinventing" the same techniques, because they aren't well-known - and many of us don't want to learn them because it's much more fun to reinvent [mea culpa].

But this is really only half the idea of patenting. Really, the point of patents is not to own new techniques, but to own new ways of solving practical problems. It's not the technique in isolation; the patent should only cover that specific way of solving that specific problem.

e.g. the patent on the telegraph should be on a particular solution to transmitting information over long distances - not a patent on connecting wires in a certain way. It should be specific to the problem.

Probably the most persuasive essay I've read in favor of patents is by pg.

http://www.paulgraham.com/softwarepatents.html

I'm still not in agreement, but he he makes a good case.

His defense that software patents are the same as all patents is a bit weak, I feel.

Good quote from the article:

One thing I do feel pretty certain of is that if you're against software patents, you're against patents in general. Gradually our machines consist more and more of software. Things that used to be done with levers and cams and gears are now done with loops and trees and closures. There's nothing special about physical embodiments of control systems that should make them patentable, and the software equivalent not.

I agree. I just think this is a strong argument against all patents rather than an argument for patents. The march of information technology makes boundaries fuzzier and fuzzier and we have to think what will happen at the limit of this process. The boundaries between algorithms and information expressions becomes fuzzy within a world of pure information. And ultimately, either we will be in a position where all information, all expression, is a priori owned by one or another entities or we'll be in a position where all information is free (sorry about the cliche being true..). I'd rather be in the latter world.

> There's nothing special about physical embodiments of control systems that should make them patentable, and the software equivalent not.

I think he's wrong, there is something special, the rate of innovation, and the cost of both change and the the initial barrier to entry. Physical machine are generally much more expensive to create, change, and startup a business with so granting the patent holder some limited monopoly on the invention makes sense. With software, not so much.

Uh,

This may be true currently but point I agree with PG on is that things are moving towards a situation where this will no longer be true. And the movement is getting faster and faster. If the law is formulated merely on how things are now, it will become obsolete just when it can do the most damage.

This isn't so clear cut for me. In many endeavors, it is far easier to copy than to invent. Patents and the entire field of IP law exists to give innovators and creators more incentives to take on risky or costly projects that ultimately provide value.

Do people have a problem with:

1.) The concept of owning IP (and collecting the rewards)

2.) That there is human subjectivity in the process

3.) That the legal system for resolving disputes is broken and hackable via expensive lawyers?

For me I think I'm only angry at (3) and some of (2), and maybe there is a way to fix the system while keeping most of the current spirit of the law intact.

Software, and for that matter, movies and books are already covered quite effectively by copyright. If we already have IP protection that provides incentives to take on risky and costly projects, why do we need yet another protection? Especially one which has been demonstrated to be bad for companies and consumers.

While I agree that we are better off without software patents, copyright does not fill the same role. Copyright protects against direct duplication of a work, while patents protect against a range of possible implementations.

For example, if you install a copy of Microsoft Word on your computer off of a friend's CD, you are violating Microsoft's copyright. If you instead write your own word processor from scratch, then you are not violating; Microsoft's copyright is independent from the copyright you hold on your word processor. But if Microsoft has a patent on a fundamental aspect of word processors, then your new word processor violates the patent, and Microsoft can limit what you can do with it.

The question is: Is the "risky and costly project" the act of implementing the word processor, or the act of inventing word processors in the first place? (And: Do patents provide incentives for these projects?)

> Patents and the entire field of IP law exists to give innovators and creators more incentives to take on risky or costly projects that ultimately provide value.

This may be true, but has anyone ever verified that IP law actually brings about more innovation? It certainly provides a legal framework to build a business around a new idea, but it's unclear to me whether it actually causes people to come up with more new ideas.

There are no areas of computer science that seem to rely on patents to make investment worthwhile. The closest would be compression I guess, but unpatented video compression techniques seem to be advancing just fine. The inherently low cost of computing research means there is a massive amount of invention without the help of patents, and in those circumstances patents are just getting in the way.

I'd also like to add the Intellectual Property rights were originally supposed to expire a few (7, I think) years. The idea was to give innovators a chance to profit, before letting society follow and help improve the situation. Now, Intellectual Property is practically permanent, which I think prevents innovation.

> Now, Intellectual Property is practically permanent, which I think prevents innovation.

Huh? Patents are 20 years.

Copyrights keep getting extended, but they're not the only form of IP.

Trademarks and trade secrets are permanent though.

> Patents and the entire field of IP law exists to give innovators and creators more incentives to take on risky or costly projects that ultimately provide value.

I don't think so, the purpose of the system was to make that knowledge open so society as a whole could benefit from it. Given innovators a monopoly was the price of buying that knowledge, otherwise they wouldn't give up their trade secrets.

"Nor is it possible to distinguish between 'numerical' and 'nonnumerical' algorithms [...]"

But I guess it is possible to distinguish "seminumerical" algorithms (vol 2)!

Here's a slashdot thread on the subject of software patents that I find horribly interesting.

http://yro.slashdot.org/comments.pl?sid=624631&cid=24319...

The title should perhaps say "... sent in 1994". (I don't think any of it is any less applicable now than it was then, though.)

Wait so what you are trying to say is that in 1994 we already knew that patents on software are going to kill us, and someone spoke up against it, and nothing was done?

Yes because enough people were not pissed off. This is changing now (sort-of-kind-of-maybe-is) because a company like Red-Hat was ticked off well enough. I give the current movement a 30% chance of success and a 80% chance of a success with no real good impact that was intended by something like this letter.

Heh... It was pretty obvious what would happen, well before 1994. From a 1991 Bill Gates memo:

If people had understood how patents would be granted when most of today's ideas were invented, and had taken out patents, the industry would be at a complete standstill today. I feel certain that some large company will patent some obvious thing related to interface, object orientation, algorithm, application extension or other crucial technique. If we assume this company has no need of any of our patents then they have a 17-year right to take as much of our profits as they want. The solution to this is patent exchanges with large companies and patenting as much as we can.

http://arstechnica.com/business/news/2007/03/analysis-micros...

It should also say the US Patent [and Trademark] Office (USPTO). There is at least one Patent Office in the majority of countries.

Reply on news.ycombinator.com