Judge Invalidates Software Patent, Citing Bilski

bfwebster writes "US District Court Judge Andrew Gilford (Central District of California) granted a summary judgment motion in DealerTrack v. Huber et al., finding DealerTrack's patent (US 7,181,427) — for an automated credit application processing system — invalid due to the recent In re Bilski court decision that requires a patent to either involve 'transformation' or 'a specific machine.' According to Judge Gilford's ruling, DealerTrack 'appears to concede that the claims of the '427 Patent do not meet the "transformation" prong of the Bilski test.' He then applied the 'specific machine' test and noted that, post-Bilski the Board of Patent Appeals and Interferences has ruled several times that 'claims reciting the use of general purpose processors or computers do not satisfy the [Bilski] test.' Judge Gilford analyzes the claims of the '427 patent, notes that they state that the 'machine' involved could be a 'dumb terminal' and a 'personal computer,' and then concludes: 'None of the claims of the '427 Patent require the use of a "particular machine," and the patent is thus invalid under Bilski.' DealerTrack apparently plans to appeal the ruling. Interesting times ahead."

Comments

Re:Similar to Donald Knuth's Logic

[morgan_greywolf]

What is "non-mathematical software"?

There is no such thing as non-mathematical software. Even printing "Hello, World!\n" requires math. Taking math out of software is sort of akin to taking carbon out of food.

Software is equivalent to math.

[Anonymous Coward]

My degree is in mathematics. There's no such thing as non-mathematical software. There is mathematical proof of this. There's a nice equivalence theorem for the two, and the website linked shows the results of that equivalence.

I repeat: there's no such thing as "non-mathematical" software, because it is equivalent to math. The only people who think otherwise don't know what math is. It's like trying to claim that 1 != 1. And yes, people really do claim utter nonsense like that sometimes, especially those who don't understand the fact that infinite sequences like 0.99999[repeating] don't have a last digit by virtue of being infinitely long (if an infinite list had a last element, it would be a contradiction in terms, because part of the definition of infinite is that for every element x, there is a successor of x).

One might as well claim that pi is exactly 3.

Re:Software is equivalent to math.

[Maury Markowitz]

> There's no such thing as non-mathematical software

Which is great and all, but has absolutely nothing to do with the patent process.

Patentability has nothing to do with the implementation, and everything to do with the intent. If the intent of the program is to control a loom in a new way, that's (theoretically) patentable because the purpose of the program is not to solve a mathematical problem, but a real-world one. If the intent of my program is to run Newton's Method, it's not patentable, because the intent of the program is to solve a mathematical algorithm. It is the _intent_that_is_encoded_, not the form of the encoding, that is the only concern in terms of _theoretical_ patentability - there's other rules that govern whether or not you'll actually GET a patent.

The basic issue that the new caselaw concerns is the concept of "transformation" in a patent. The original intent of the system was to grant protect to novel devices, either improvements on existing concepts (disk brakes vs. drum), or to totally new devices. But then what is a "device"? Is a deck of cards a device? A new book? The law defined it as something that tranforms something. A machine for freeze-drying coffee to produce Nescafe transforms hot coffee into a powder, which can be re-constituted into something similar to the original. Something is actually "being done", and it's THAT that the patent covers, not the machine itself. You can invent a better freeze drying machine, and get a patent on it, but you'll still have to license the actual concept of freeze-drying coffee.

This is good, and worked for a long time. Then the CS wags came along. In the 1970s there were a couple of cases where it was argued that a program is literally a collection of instructions that transform one number into another. Thus every program is transformative, and patentable. Once that got into caselaw, then someone pointed out that that _purpose_ of the program is to run an algorithm, so why does it have to be in a program? Why not any process that does any processing? And thus we got into the mess we have today.

The new caselaw basically noted that something got lost in the 1970s rulings - that the transformation has to _produce_something_patentable_. So consider these cases:

1) I make a machine called the wingnut that produces gazzezas
- patentable, gazzezas are a new product
2) I make a machine called the gizifa that produces gazzezas
- patentable, but I'll need to license gazzezas, or wait for it's patent to run out. Additionally, someone else is free to make a different machine that makes gazzezas.
3) I make a machine that runs Newton's Method
- treated like (2), because - critically - the product is not patentable

In essence, all patents had to flow from (1). Basically, there had to be a product somewhere. Prior to Bilski, caselaw stated that programs are transforming inputs into outputs, so there is a "product". Oddly, (2) didn't really exist, because they noted that a program can run across a wide variety of machines, so the idea of a "specific machine" (or implementation) was difficult because patents on (1) were automatically so broad.

The judge in the Bilski case noted that while it is true that all algorithms are transformative, the patent in question failed to actually make a "product" that was patentable, like case (3). Risk hedging is not a product you can patent, so then the only patent you might be able to get would be like (2). So this business practice moved from (1) to (3) - or similar to (2). He then noted that the patent application itself stated that it could be run using any number of different methods, it couldn't be a (2), by definition. So the patent was invalid.

Bilski was about a business practice, which generally don't produce anything. So now it appears some enormous subset of these are invalid. But what about software?

DealerTrack shows that the same principal applies. The patent in question did not produce a patentable

Double Edge

[duanes1967]

This is really a tough situation. Consider the CODEC. It is primarily a series of mathematical algorithms, but is quite complex and provides a function never before found. This is the brunt of intellectual property. We have moved beyond mechanical devices to the point that the device is not unique, but it's application is. On the other hand, what if the patent on a pencil covered the output from the pencil? In my opinion, a codec is definitely a process that is non-obvious, while Amazon one-click purchasing is a natural evolution. I have a device and software that I want to patent that falls into between "Duh-Why didn't I think of that" and "Holy Sh#$, that's awesome". While not obvious, it is not rocket science but no one has come up with anything like it yet. It is not merely and extension of current ideas. The device itself is only required in some situations in which an adequate general processor is not available (ie, stand alone operation). In the case of Dealer Track, I think that computer based credit application is simply an evolution of computer based forms processing. There is nothing new or non-obvious here.

Re:And so, it begins.

[jank1887]

Done. From Bitlaw (emphasis mine):

Section 101 of the U.S. Patent Act sets forth the general requirements for a utility patent:

        Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvements thereof, may obtain a patent, subject to the conditions and requirements of this title.

In other words, for an invention to be patentable it must:

      1. be statutory,
      2. be new,
      3. be useful, and
      4. be nonobvious.

Re:Similar to Donald Knuth's Logic

[danaris]

If fifty years ago I came up with a way to manufacture ball bearings - independently of an existing, patented method - would I not be sued by the patent holder of the bearing production process if I brought a product to market using my bearings?

Only if your method was identical (or very similar) to his method.

Despite modern corruptions, particularly in software patents, most patents are not, and should not be, of the form "A patent on making type of object X". They are and should be "A patent on a method for making type of object X."

In the patent, the entire method is clearly spelled out—it is made "patent," or obvious—and from the patent, anyone in the field and with the requisite equipment/money could produce the same object X by the same method. This, too, is missing from software patents, because to truly match a regular patent in this, the software patent would need to include the source code.

Dan Aris

Decision Text Here

[Theaetetus]

Summary points to a press release. The actual decision is available here: http://bfwa.com/docs/dealertrack.pdf (7 page pdf)

Re:Babies and bathwater

[Theaetetus]

But if the bath water is going to include such notorious crap patents as 1-Click, Desire2Learn, NTP, and many others, then I would have to say that the bathwater is so rank and disgusting that it's not too high a price to pay to lose a handful of babies, as Bilski does.

But can't we do better? Can't we find an "obviousness" test that works?

Bilski wasn't about obviousness - Bilski was about patentability of certain types of inventions. For obviousness, you want to look at KSR v. Teleflex, where the Supreme Court laid out 9 different ways to find something obvious.

Re:Similar to Donald Knuth's Logic

[CastrTroy]

Great point. I may write a book, and it may have some very non obvious and novel story lines, but the book isn't patentable. Books have copyright protection. The way patent law was written it specifically says you can't patent an algorithm, no matter now original it may be.

The C definition, same token on both sides.

[Xenographic]

I wasn't logged in before, GP anon was me. Anyhow, the period was the end of the sentence, not some attempt to make it into a float/string/boolean/whatever and I certainly didn't use the Python operators. It's supposed to be the same token (1) on both sides. But that's why we use formal languages that are picky about syntax and which can be checked automatically to avoid people finding weird ambiguities to question.

The theorem I was mentioning above is called Curry-Howard-Lambek correspondence (it took me a while to find all the links):

The Curry-Howard-Lambek correspondance is a three way isomorphism between types (in programming languages), propositions (in logic) and objects of a Cartesian closed category. Interestingly, the isomorphism maps programs (functions in Haskell) to (constructive) proofs in logic (and vice versa).

(Wiki links added because most people are too lazy to Google the terms they don't understand. Especially if they don't realize that they don't actually understand them.)

So even if you find some crazy language where they define != to be an equality operator or something equally unusual, software is still equivalent to math. Metamath wouldn't be possible otherwise. And as you can see, they're doing just fine.

Comment removed

[account_deleted]

Comment removed based on user account deletion