Patents, Copyrights, and Software

Reader Dale B. Halling left the following comment that articulates two major misconceptions that one frequently encounters in the software patent debate:

There is no doubt that the patent system should be more accessible, less complicated, and less expensive. Inventing is creative, this is not unique to developers of software. Writers would be and have been historically disgusted by people that can pirate their writings. In the 19th century copyrights were national (like patents are today), so foreign publishers stole the works of famous authors and reproduced them in their country without paying the author. Mark Twain suffered this fate particularly at the hands of English publishers. I am sure the English publishers did not see why they should have to fuss with copyrights of American writers and vice versa. The moral point is the same with software developers who do not want to bother with those complicated patents – stealing is stealing.

Some people believe that software should not be patentable. The arguments against software patents have a fundamental flaw. As any electrical engineer knows and software developer should know, solutions to problems implemented in software can also be realized in hardware, i.e., electronic circuits. The main reason for choosing a software solution is the ease in implementing changes, the main reason for choosing a hardware solution is speed of processing. Therefore, a time critical solution is more likely to be implemented in hardware. While a solution that requires the ability to add features easily will be implemented in software. Software is just a method of converting a general purpose electronic circuit (computer) into a application specific electronic circuit. As a result, to be intellectually consistent those people against software patents also have to be against patents for electronic circuits.

Reader Rhayader offers a good response to the first point: stealing is always deliberate, but patent infringement is frequently accidental. If you survey recent software patent litigation, you’ll find many cases where the defendant had never heard of the plaintiff or its technology until after it had developed its own product. “Stealing is stealing” just doesn’t apply in these cases.

I think the second argument about the equivalence of software and hardware is based on a misunderstanding of how the patent system works. It’s common to talk about patents as though they cover particular products or devices. But it’s more precise to say that patents cover particular characteristics of devices or processes. So if you’ve created an innovative new widget, your patent application can’t just describe how your widget works. It also needs to identify the specific characteristics of the widget that you believe are novel and non-obvious, and therefore entitled to patent protection.

So a “no software patents” rule simply says that the software aspects of an invention cannot be patented. Mathematical algorithms—which is all software is—ought not to be patentable whether they’re implemented as software or on silicon. But that doesn’t mean I’m against semiconductor patents in general. Rather, I think patents should cover aspects of a semiconductor device that doesn’t reduce to mathematics. For example, I wouldn’t have allowed C. A. R. Hoare to patent the quicksort algorithm. And I wouldn’t have allowed someone to patent the concept of hardware-accelerated quicksort. However, someone could patent a particular, non-obvious design for a hardware quicksort implementation. That’s not a patent on the algorithm, but on a specific machine for implementing the algorithm.

In other words, a software-related patent must be tied to a specific machine or physical process. That’s similar to the standard the Federal Circuit articulated in In re Bilski. One of the great virtues of this approach is that it frees computer programmers from worrying about inadvertent patent infringement. Because by definition, a “pure” software implementation of an algorithm isn’t tied to any specific machine or physical process. So you’d only have to worry about patent infringement if you were implementing an algorithm in silicon. Building hardware is much more expensive than building software, so the requirement to hire a patent lawyer is much less burdensome.