My father, a US patent examiner, suggested to me that I should make a push to patent my company’s software algorithms. Apparently, there was recent (< 2 years) change in the law allowing algorithms to be patented. Dad told me that a patent can be listed as a company asset, and the value can be declared in the patent application. Has anyone had any experience in this arena? Do you think the payoff would be worth the investment?
Re: Patenting algorithms.
You can certainly do this and this was possible for many years now, going back at least to 1996 but even well before that in some cases.
A little (I would think well-known) history to put this in context: in the 1972 Gottschalk vs. Benson case, the famous patent lawsuit focused upon a computer program algorithm that did nothing more than convert binary-coded decimal numbers to pure binary. According to the Supreme Court, the 1952 Patent Act did not cover an algorithm. However the Supreme Court did not define the term "algorithm". However this decision did allow the Patent and Trademark Office to pretty much summarily reject any patent applications if they contained the word "algorithm". Now, in 1978 the equally famous Parker vs. Flook case centered upon a patent that incorporated an alarm-generating algorithm for computer systems. Here the Supreme Court decided that other than the algorithm itself (which was not patentable, by their own previous dictate), the patent had no significant innovation, which was the defining characteristic for a patent irrespective of the algorithm used. Thus, the 1952 law stood. Moving along, 1981 was interesting because that saw the Diamond vs. Diehr case, where the Supreme Court looked at the idea of a process that employs an algorithm and decided that this could be patentable. In that particular case, the Supreme Court found that the invention incorporated an algorithm, but also found that other portions of the patent, which covered the entire process (which was the molding of rubber, incidentally) were innovative, and thus eligible for patent protection. In other words, the algorithm fell in between the cracks because of the presence of "significant innovation" relative to the process as a whole. Note, however, that the algorithm itself was still not strictly protected as a lone entity but was protected as part of that specific process.
Now we come to 1994. This was the Alappat case presented before the Court of Appeals for the Federal Circuit. Here the ruling was that a patent could not be denied solely because it contained an algorithm. What happened was that the Patent and Trademark Office used something called the "Freeman-Walter-Abele" criterion whose purpose was essentially to determine "patentable algorithms" if they were algorithms that manipulated more than abstract ideas. (This ruled out purely mathematical algorithms for example.) This test had two primary components: mathematical algorithm tests and physical transformation tests. Failure of those tests could exclude a patent application relative to an algorithm. And then we come to the very famous (and precedent setting) 1996 case of The State Street Bank and Trust Co. vs. The Signature Financial Group, where, again, the Circuit of Appeals for the Federal Circuit extended legal protection to business methods that were embedded in software and that could include algorithms.
So it has been a gradual process. Since 1996 many banks have done this with their rate-flow algorithms, for example. Trading and stock systems have patented certain algorithms, although only those proprietary to their own system. Queuing network algorithms for networks, for example, cannot really be patented in terms of return on investment but they can be patented in terms of recognition or origination. The RSA (cryptographic security) is probably the most popular algorithm patent (Patent No. 4,405,829) that I know of and that payoff was certainly worth the investment. Or consider Internet Server Access Control and Monitoring Systems (Patent No. 5,708,780), Network Sales System (Patent No. 5,715,314), or Digital Active Advertising (Patent No. 5,724,424). Or consider the run-length encoding algorithms that were patented in 1986 (Patent No. 4,586,027) and 1989 (Patent No. 4,872,009). Or consider the LZ77 compression patent (Patent No. 4,701,745). So were those worth the investment? Many would say yes. However, how much this will be a return on investment for you (or your company) is, of course, impossible for us to say. That depends on a lot of factors as well as the market viability or the widespread applicability of the algorithms you are talking about. Note that not all algorithms can be summarily protected by a patent as there are still certain conditions that apply.
Re: Patenting algorithms.
Thank you for giving such an informative response! I had no idea there was so much history surrounding this issue.
After considering your reply, I don't think it would be worth it, as the algorithms we employ would probably fail the "significant innovation" test, and don't offer much in terms of wide spread application.
Re: Patenting algorithms.
<BLOCKQUOTE><font size="1" face="Verdana, Arial, Helvetica">quote:</font><HR>Originally posted by bpolitzer:
After considering your reply, I don't think it would be worth it, as the algorithms we employ would probably fail the "significant innovation" test, and don't offer much in terms of wide spread application.<HR></BLOCKQUOTE>
Understand, also, that everything I talked about here is relevant to the United States. The United Kingdom and Australia, to name just two, have different laws in different situations. The UKPTO and European Patent Organization (EPO), for example, has had some interesting papers come out on the nature of algorithms. Trade concerns, of course, make this an issue that many people want settled because they take in much software from the United States.
What is interesting in most cases is that algorithms are considered not patentable unless they can be "properly applied", in the legal jargon. But consider this: when is an algorithm "properly applied"? That is a difficult question to answer. Legally speaking, I can tell you this as an example: the courts consider that an algorithm used in a data compression process as being "applied" because data compression is a recognized statutory type of process. With that, however, if one is considering using an algorithm in accounting software, this will not be a "properly applied" algorithm if that algorithm just calculates a new balance or other number which does not "control the flow" of the software. Consider that the 1996 case stated very specifically that "the mere fact that a claimed invention involves inputting numbers, calculating numbers, outputting numbers, and storing numbers, in and of itself, would not render it nonstatutory subject matter unless, of course, its operation does not produce a useful, concrete and tangible result." And it is that last part that is interesting because the court concluded that the question of "statutory subject matter", such as it is, "should not focus on which of the four categories of subject matter a claim is directed to - process, machine, manufacture, or composition of matter - but rather on the essential characteristics of the subject matter, in particular, its practical utility." The court defined "practical utility" as producing "useful, concrete, and tangible result." As far as what those "results" are, those can, according to the court, be measured in numerical terms "as price, profit, percentage, cost, or loss."
Finally, understand also that there does exist the Software Module Patent system, and here the idea is that you are not patenting algorithms, but specific implementations of these algorithms in software. So, for example, an on-line car buying service cannot implement the algorithm of context-relevant circular-based radius searching queries but they can patent how they specifically tie that algorithm into their particular software. But sometimes this can get away from you. Consider the whole GIF/LZW debacle between CompuServe and Unisys, for example, or between Microsoft and Stac Electronics as another example. In the current situation, some of the "automatic music preference" engines that exist (to automatically suggest music that you might like based on other music you like) has gone through a similar round of battles of just what it means to "properly apply" a given algorithm and what can be considered a "significant innovation". The legal proceedings on these issues are actually fascinating to read through in some respects.
[This message has been edited by Cryptonomic (edited 10-13-2002).]