On Friday we got our first taste of the practical consequences of last month's landmark decision from the Supreme Court restricting patents on software. The Federal Circuit Appeals Court, which hears appeals in all patent cases, invalidated a software patent for being overly abstract. And the reasoning of the decision could lead to a lot of other software patents going down in flames, too.
The appeals court considered a lawsuit from Digitech, a firm that has sued numerous companies over a patent that was originally granted to Polaroid in 2000. The patent claims a method for ensuring that images display consistently on different kinds of digital devices such as computer screens and printers. At the time Polaroid sought the patent, companies such as Apple had already invented methods for ensuring that different devices displayed colors in a consistent fashion by building a profile describing the characteristics of each device. The Polaroid patent extended on that concept, claiming the idea of creating device profiles that not only include color information but also information needed to make images less noisy.
Unfortunately for Digitech, this kind of patent has faced growing skepticism in the courts. Digitech's patent didn't claim any specific type of hardware for capturing or manipulating images, it merely claimed the concept of storing and using information describing the optical characteristics of particular devices. That, the court said, wasn't a patentable invention.
The patent merely "describes a process of organizing information through mathematical correlations and is not tied to a specific structure or machine," the appeals court ruled. "Without additional limitations, a process that employs mathematical algorithms to manipulate existing information to generate additional information is not patent eligible."
Of course, a process that "employs mathematical algorithms to manipulate existing information to generate additional information" is a computer program. This reasoning could call into question the validity of thousands of existing software patents.
The legal scholar Robert Merges has argued that the Supreme Court's June ruling, the first time the Supreme Court has weighed in on software patents since 1981, could endanger a lot of existing software patents. Friday's Federal Circuit ruling underscores that conclusion. Consider Google's famous PageRank patent, which covers the algorithm at the heart of Google's search engine. In the language of the Federal Circuit, it claims the use of "mathematical algorithms" (involving eigenvectors) to "manipulate existing information" (a list of links between web pages) to "generate additional information" (a ranking of the pages).
This represents a significant shift for the Federal Circuit, which has traditionally been a strong supporter of patents on software. Indeed, the modern software patent era began in the 1990s, when the Federal Circuit began to loosen early Supreme Court rules against patenting software. In a key 1998 ruling, the Federal Circuit upheld a patent on a method for managing mutual funds with software. Like the Digitech patent, it merely involved manipulating information (about market prices) to generate additional information (buy and sell orders). It's hard to imagine the court reaching the same conclusion today that it did in 1998.
The Supreme Court is in an awkward position when it comes to software patents. There's mounting evidence that patents are bad for the software industry, and the court's own precedents, dating from the 1970s, cast doubt over the patentability of software. Yet invalidating software patents worth billions of dollars would be disruptive and controversial move, and courts hate to make waves.
So the Supreme Court has taken an incremental approach. Its recent decisions have signaled that the Federal Circuit has been too lax about granting software patents without establishing a clear standard for when (and if) software can be patented. Yet the Federal Circuit seems to be reading last month's ruling as a significant new restriction on software patents.
That's significant not just because other patents are likely to be invalidated in the future, but also because it creates greater uncertainty about the validity of all software patents. It's hard for patent holders to predict whether their software patents will be invalidated in court. But the threat of suffering Digitech's fate will make all software patent owners more reluctant to enforce their patents aggressively.
Correction: This post originally stated that the decision was made this week, instead of last Friday. Also, I described eigenvectors as an algorithm. In fact, the Pagerank algorithm uses eigenvectors.