Patent Office finds voice, calls for software patent sanity

The US Patent and Trademark Office is a convenient whipping boy for problems with the patent system. The USPTO famously approved the junk patents at the heart of the legal battle between Research in Motion and a patent-trolling firm called NTP. The USPTO belatedly recognized its mistake and began invalidating the patents, but didn't finish the job in time to save RIM from being forced to pay $612.5 million to avoid having its whole BlackBerry network shut down.

Perhaps shaken by the negative publicity generated from that case and others, the USPTO seems to have gotten religion on patent quality. Over the last year, the courts have heard a series of cases in which the Patent Office has rejected software-related applications. The legal theories advanced by the Patent Office in these cases are causing consternation among software patent supporters.

In an article last week at Patently-O, law professor John Duffy warns that the Patent Office has staked out positions that, if accepted by the courts, would amount to the de facto abolition of software patents. He's right that the Patent Office has become increasingly hostile to software patents in the last couple of years. However, it's far from clear that the end of software patents is imminent. And Duffy is dead wrong to suggest that fewer software patents would be bad for innovation.

Growing hostility

Duffy cites four recent cases that illustrate the Patent Office's growing hostility to the patenting of software and other abstract concepts. While the USPTO hasn't formally called for the abolition of software patents, the positions it took in these cases do suggest a growing skepticism.

In the first two cases, decided last fall, the United States Court of Appeals for the Federal Circuit (which has jurisdiction over patent appeals) upheld patent rejections by the USPTO. They were not software patent cases, as such. In In Re Nuijten, the court considered a patent related to an algorithm for adding a watermark to a digital media file. The Federal Circuit did not invalidate the claims relating to the watermarking algorithm itself; everyone seemed to agree that the algorithm was patentable. Rather, the decision focused on whether a digital signal could be the subject of a patent claim. The court concluded that it could not. A victory for common sense, perhaps, but hardly a rejection of software patents.

In the case of In Re Comiskey, the Federal Circuit found that adding binding arbitration language to a legal document is an unpatentable mental process. However, it sent the case back to the lower courts to consider whether automatically performing this process with a computer would have been obvious at the time the application was filed.

Two decisions from May and June of this year show the Patent Office moving more decisively toward rejecting patents on software. They were heard by the Board of Patent Appeals and Interferences (which is above the Patent Office but below the Federal Circuit in the appeals process) and concerned patents on simulating physical processes using mathematical models. The Patent Office rejected both patents on the grounds that they involved only the manipulation of abstract mathematical formulas and were not tied to a specific machine or physical process. The Board of Patent Appeals upheld the Patent Office's decision in one case but, in the other, it overturned the Patent Office because the patent described a simulation that occurred on two processors operating in parallel.

The Board ruled that because the patent described a specific machine—a computer with two processors instead of just one—the patent was sufficiently tied to the real world to merit patent protection.

The underwhelming case for software patents

Duffy rightly ridicules this result. Clearly, it makes no sense for the patentability of an algorithm to depend on the number of processors used to implement it. And it would surely be pointless to have firms gratuitously adding or removing processors from their software algorithms simply to dodge patent liability.

Duffy would have granted both patents. But the Patent Office's position—that both applications should be rejected—is not only more consistent with Supreme Court precedents, but it's better policy as well. The Supreme Court has repeatedly ruled that abstract concepts and mathematical algorithms are not patentable. Unfortunately, the Federal Circuit has ruled that merely loading an algorithm onto a general-purpose computer and executing it transforms it into a patentable invention. This doctrine makes a mockery of Supreme Court precedents. However, the Supreme Court's precedents are still technically the law of the land, and the Patent Office seems to be increasingly willing to stand up for those precedents even at the risk of being overturned by the Federal Circuit.

Duffy seems to regard the end of software patents as a calamity for innovative companies, but his argument is awfully thin. Duffy focuses on Google's PageRank patent, which he has long regarded as a poster child for software patenting. He describes it as "surely one of the most famous and valuable of all modern software patents," and suggests that the invalidation of Google's patents would be a calamity for the company. Curiously, however, he never explains how Google benefits from this or other patents in its portfolio.