What Do You Think About Software Patents?

John McHugh, Portland State University:

I am of a mixed mind about the issue. First, let me say that I was a patent examiner for several years in the mid 1960s. In my day, software per se was not considered patentable. A program or algorithm could be the subject of patent claims only if it was embodied as hardware--in effect, a special purpose computer to carry out the invention. At that time, it was also the case that mathematical methods in general were not considered as patentable, limiting the protections that might be obtained for algorithms in the abstract.

I left the office in mid-1965, and am not sure whether the current patentability of software is due to changes in statutes or court decisions, though I have the impression that it started due to the latter.

It seems to me that there are two issues: 1) Should inventions and discoveries that are embodied in software obtain the same kinds of protections as inventions and discoveries that are embodied in more traditional forms such as mechanical devices, manufacturing processes, asexually reproduced plants, etc.? and 2) Does the patent do a good job of ensuring that the patents that it issues in this or other fields are valid?

In examining the question, we should probably consider alternatives. Note that a patent is supposed to be a trade. The inventor discloses the invention in sufficient detail to allow its practice by one of ordinary skill in the field to which it pertains in return for rights to control its practice for a limited time--17 years when I was involved. Alternatives include protection by trade secret, which provides no protection against independent rediscovery or disclosure by individuals not covered by license or non-disclosure agreements, but does provide indefinite protection. Copyright protection is also claimed for software, but this has always seemed to me to be strained and really should be related more to form than to concept. In the absence of a mechanism more suited to the ephemeral and malleable characteristics of software, I suspect that patents are the best we can do, if we believe that this kind of invention would be protected.

So, now we come down to the issue of should inventions embodied as software be protected. In the old days, it was settled that mathematical methods in the abstract were not patentable. Thus one did not see patents for "A method for factoring large numbers" or "A method for the logical combination of symbols." One could patent a machine for factoring large numbers that used a novel method and obtain such a broad claim that any embodiment of the method beyond performing the computation in your head or with pencil and paper would be covered. So, is a computer program a machine, or is it more like thinking or working with pencil and paper? If it is a machine, then algorithms embodied as programs are patentable - and I suspect that this was probably the reasoning of the court.

Remember that the patent has two parts, the disclosure that ultimately allows the public to practice the invention, and the claims which carve out the protected turf. During the examination process (see below) the inventor (actually the inventor's lawyer) tries to get claims that are as broad as possible, while the patent office tries to show that the claimed material has been anticipated by prior work. If prior work that would be covered by the claims does not exist or cannot be found by the patent office and if the argument cannot be made that the claimed invention would be obvious (to the mythical practitioner of ordinary skill) given the prior art that does exist, then the patent will issue. Finding the prior art is the patent examiner's job. Determining "obviousness" is harder. Being naturally curious, I think that lots of things are obvious, given some prior knowledge and a need. Most inventors and patent lawyers disagree.

As evidence for my view, I note that we often have a run of similar or identical discoveries as soon as there is a generally recognized need and certain enabling technologies. I think that exploiting any well known mathematical property, such as the fact that A xor B xor A = B in a computer program should be obvious, but one of the first software patents covers an application of this fact.

The patent office may not be very well equipped to deal with new or rapidly changing technical areas. The primary repository for "prior art" is previously issued U.S. and foreign patents. When a new field starts producing patents, there is little or no prior art of this form. The small, everyday tricks of the trade are not easily available to the examiner. The vast majority of patent examiners were, in my day, recent graduates with technical backgrounds who worked in the patent office and attended law school part time. After some years in the office and a law degree, they became patent lawyers and usually left the office. Senior examiners knew the field primarily from the patent literature, but were seldom practitioners. I once had an application that claimed the use of a junction transistor with one burned out junction as a diode to be the invention. I knew that I had seen this in a "hints" column in a magazine such as "Popular Electronics" years earlier, but could not find the reference. I rejected it as "obvious," the applicant appealed, and the appeals board found it to be unobvious and the patent issued. This was harmless as the claim was very specific and it is unlikely that a major industry based on defective junction transistors would be thwarted by the existence of the claim. Knowing that references exist and ing them are two different things. Like most businesses, patent examiners are under pressure to produce. This pressure limits the time that they can spend on the examination of a given application. The fact that most examiners are not skilled practitioners in the field, coupled with a low standard for unobviousness, makes the examination of applications from fields that do not have a long history of patentability somewhat haphazard. Time will change this, but the fact that we are seeing a number of companies that appear to exist primarily to collect and litigate software patents is disturbing, given that many of the early software patents seem to cover ideas that were probably either well-known at the time the patents were filed or at least obvious to many practitioners. The fact that many ideas get passed down through the folklore and non-archival repositories further complicates the problem both for the patent office and for the developer faced with an infringement suit for using a well-established technique.

Another View:
David Barstow, Instant Sports:

I have a fair amount of personal experience in this regard, being the inventor on 4, soon to be 5, patents, all of which have a software flavor. The most interesting ones are those related to the use of simulation techniques to do interactive broadcasting of sporting events. The patented technology is the basis for my little company Instant Sports. We provide pitch-by-pitch coverage of Major League Baseball games with about a 60 second lag from reality. We even have an animated service, which works whenever Java is working.

My perspective is that of a little guy who invented something very interesting and now finds himself competing with some very big guys. My patents are the only reason the company is still alive at all. So I think I'm a case where the patent process is achieving one of its objectives, enabling small inventors to do interesting and valuable things and have some hope of the results not being taken over by somebody big.

Another View:
Dewayne Perry, Bell Labs:

I don't think software patents are the right thing. It is not really clear what is being protected or why. Copyright isn't appropriate either. We of course patent everything we can think of, but it is not clear what good it is. How do you check for it? How can you tell that what is used is the same or substantially different?

No one patents mathematics--if programs are mathematical objects, then what is the difference? That we can make money out of software objects but we can't out of semi-groups or Cartesian products?