Kelly Talcott of Pennie & Edmonds recently posted some interesting thoughts on software and business method patents. Of particular interest, he said:
Let's not, however, ignore the role that practicing patent attorneys have played in this. We all have an obligation to disclose relevant prior art to the PTO as part of the patent application process. For Internet and software patents, our clients are frequently the best sources for that art. Are we really doing our clients a favor when we file applications without making sure that someone has performed a diligent prior art search (commensurate with client resources) as part of the patent application process? Are we helping a client when we secure a patent upon which the client then relies, only to see it invalidated when tested in the fires of a patent infringement lawsuit?
Unfortunately, Mr. Talcott's musings are all too relevant in intellectual property as a whole. His underlying point is valid, too—that the whole concept of "business method" and software patents is open to question.
I do not know where Mr. Talcott comes down on the concept. In the context of intellectual property as a whole—something that is seldom given enough thought—neither business methods nor software are most appropriately treated as patentable material. Business methods deal with a concept that can be reduced to practice in multiple ways. The key is the concept; and the "right" way to analyze a pure concept is as a trade secret. Software, on the other hand, actually has two components, and looking at those components separately helps illuminate the issue. At least, it is a slightly less charcoal shade of grey.
Software includes both executable object code—the binary instructions that a computer can actually follow—and some higher-level abstraction of instructions to create that code. It remains possible to write programs directly in machine code. I've done it; and I still bear the scars from that experience. The closest analogy in existing intellectual property law is to the concept of a derivative property in copyright. Consider, for example, Gone With the Wind. (Or, if you have good taste in literature, something else.) The film is a derivative property from the novel. The film is an independent expression based upon the source code. More than one different version of a film could have been made from that same source code; one can even argue that merely adapting it for the 4:3 aspect ratio of the television screen is the same thing, analogous to porting a program from Windows to Linux. The difference, though, is that the exact "expression" of object code for a computer does not matter. The key is still the source code, whether in assembly language, Java, or whatever.
The farther one's exact expression gets from the source code, the less protection that source code can/should have. A core algorithm is not really expressed in the object code; it is expressed in the source code. The Quiksort algorithm is an excellent (and all-too-current, due to SCO's questionable litigation over rights to Unix source code) example. There are a limited number of ways to express the algorithm. Under copyright law, that greatly weakens the protection given a particular expression. That is exactly what should happen.
Of course, the intricate intertwining of copyright and patent law, as in the DVDCCA case, presents a Gordian knot. I propose to unravel that knot the same way as Alexander did: by cutting right through it. Patent is just not the right reference frame for analyzing protectability for software. The law of "how much different X must be from Y to avoid infringement" is so much clearer, and better developed, in copyright law that only those who wish to obstruct the progress of science and useful arts have an interest in treating software as part of the patent field.