Arnold Kling  

Paul Graham on Software Patents

PRINT
Will Wilkinson on Happiness Re... Paul Graham on Many Subjects...

A commenter on a previous post mentioned a couple of essays by Paul Graham. He is one of my favorite writers, and it's been too long since I visited his site. Here is an essay on software patents.


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.

...I think the problem is more with the patent office than the concept of software patents. Whenever software meets government, bad things happen, because software changes fast and government changes slow.

...Before patents, people protected ideas by keeping them secret. With patents, central governments said, in effect, if you tell everyone your idea, we'll protect it for you. There is a parallel here to the 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 for protecting them. (Magnates still have bodyguards, but no longer to protect them from other magnates.)


This reminds me of The Arbiter with the Golden Scepter.


Comments and Sharing





COMMENTS (8 to date)
Heather writes:

I completely disagree with the statement "if you're against software patents, you're against patents in general." Software is a language, and like any language, is covered under copyright protection. Patents were created to protect physical innovations, not language. They both have places where they are applicable, but the rules are, and should be different.

Josh writes:

Heather,

Software is a blueprint. And like all blueprints, it should be coverable by a patent.

John writes:

Heather:

Say I am Ron Rivest, one of the creators of the RSA cryptography, a clearly novel and unobvious invention. I create a hardware and a software implementation of RSA. Are you seriously suggesting that I be granted a patent for the hardware implementation but not the software one?

Neel Krishnaswami writes:

It is manifestly the case that software should not be patentable. From the pov of natural rights theory, patents are not justifiable, period, because ideas are not rival. My use of the theorems of public key cryptography does not prevent Ron Rivest from doing so, also.

From a economic/utilitarian perspective, patents only make sense when the cost of developing an invention is greater than the private benefit of developing it, and less than the public benefits. If the public benefit is smaller than the cost, then it's not worth developing. If the private benefit is greater than the cost, people will develop it regardless of whether or not patent protections exist -- in this case all patents do is transfer consumer surplus to a monopoly (with the usual associated deadweight losses). There are no cases where this condition unambiguously holds for software -- and quite a few cases of patent trolls shaking down genuinely innovative companies.

John writes:

It is manifestly the case that software should not be patentable. From the pov of natural rights theory, patents are not justifiable, period, because ideas are not rival.

This sounds like "it is intuitively obvious that ...", but nevertheless:

You can rail against patents all you want, but they are enshrined in the US Constitution, and seem to have legal backing in most of the world, so the question is not whether we should have patents, but what kind of patents should we have.

Which brings me to my original question -- we have an RSA implementation in hardware (say, an ASIC) and one using software and a general purpose CPU. Why, in your perfect world, does only the hardware implementation rate a patent?

Lex Spoon writes:

Patents on an area should be evaluated in the economic way Neel describes, and that evaluation will go differently for software and for drugs. Graham is just wrong when he says all patents are the same.

Still, software patents need to go. For every RSA or MPEG, there are thousands of patents on one-click shopping or the xor sprite hack. These patents are not supporting anything productive, and they are land mines for people doing truly inventive work.

The software industry is different from the drug industry, and software patents are holding back our potential.

Heather writes:

There are rare cases where software patents make sense, primarily in assembly code areas that are directly linked to hardware and architectural implementations. The problem is that a patent does not allow "fair use" which means that single lines of code can be used to litigate patents.

Josh, you are mistaken in the assertion that blueprints are covered by patents. They are covered by copyright.

John, patents are meant to cover methodologies, not code. The cryptography implementation is patentable, whether it is in hardware or software, however the individual lines of code are not. This is appropriate.

John writes:

John, patents are meant to cover methodologies, not code. The cryptography implementation is patentable, whether it is in hardware or software, however the individual lines of code are not.

I don't think individual lines of code are patentable. Having worked on a few software patents in the past, I've seen claims that describe how the system (CPU + software) works, but never a code listing. The fuction of the system is patented, but not the code base.

Perhaps this is a personal failing of mine, but as an Electrical Engineer/Computer Science type, I define the functionality first, then decide on the instantiation. The instantiation might be custom hardware if performance is a concern, or CPU+software if circumstances warrant. In either case, the functionality is identical, so claiming that a patent on the hardware version is good while a patent on the software instantiation is verboten seems to me to be a false dichotomy.

YMMV, of course.

Comments for this entry have been closed
Return to top