Thursday, December 29, 2005

Bad software patents again

Greg Aharonian has named the worst software patent of the year in his PATNEWS mailing list, and I must say that I thoroughly agree with his choice. Behold U.S. Patent 6,910,071, which protects a "Surveillance monitoring and automated reporting method for detecting data changes". Sounds impressive, right? Here is the abstract:

A surveillance monitoring and automated reporting method is used for detecting observable changes in data sources over a network, such as the internet, for accessing changing data, such as world wide web content data, and for providing scheduled change detection notifications and results through user defined search criteria for automated monitored search criteria matches on a recurring basis by user defined scheduling. The method extracts content data from the data sources and updates a master database, then detects changes in the content data within the search criteria. Upon detection, the user is notified using graphical interfaces, electronic mail messages, pager messages, or personal data assistant messages.
Wait a second. A patent for a system that detects changes in database and sends a notification? This is beyond obvious, any database worth its salt will come with such a notification procedure. Not only that, imagine that any sort of notification of changes already in existence will be infringing.

And they ask why so many people are opposed to software patents.


kieran said...

But this isn't a problem explicity with software patents; obviousness disguised by technobabble is a problem in every area of patents. And when software patent critics cite examples like this they appear as strong critics of the entire patent system. While this is a view I would subscribe to, it's not the best way of convincing a patent supporter that the system should not be expanded - he thinks the system works, and strong critics don't.

Moreover, if you argue from this position then you box yourself into a corner. How do you combat obvious patents? You could provide specialist training to patent examiners. You could increase the cost of registration. You could encourage a re-registration system, increasing the cost of maintaining a patent, so that only strong, profitable patents endure for their full term of years. However, all of these things make patenting more expensive, and so favour larger companies. The debate about obviousness certainly affects software patents, but the argument is on patents as a whole, and not targeted at software.

To the patent supporter the argument is that maintaining a divide between software and hardware is illogical. And it is. For example, a recent development in audio processing is convolution, which detects variations in sine waves to build profiles of things like reverberation in a room. A computer knows what a sine wave should sound like, and will detect variations in the sound of the sine wave recorded in a room to find out how the room affects the sound. You can then apply this fingerprint to other sounds. Except it also allows you to run a sine wave through very expensive audio hardware, much coveted by professionals, and generate a fingerprint of what the hardware is doing. So companies selling convolution software bundle packages of profiles with it, much to the annoyance of the manufacturers of the harware. The point is that a lot of what can be done in hardware can also be done in software, and to maintain a divide between the two is to undermine the patents in hardware. Personally, I think the example I gave is sharp practise, but it would also be worrying if we privileged hardware through patents, as it is inefficient in comparison to software.

And this is where I think the argument against software patents is strongest. The problem is that the flaws and quirks of our patent system are exacerbated by factors unique to the manufacture of software. You have short development cycles, rapid technological advances, for which 10-15 years protection is utterly diproportionate. But above all in the manufacture of software you are pretty close to perfect competition, in that the cost of entry is negligible (cf. the group of software developers from around the world who developed a better web browser than microsoft, which is something to encourage.) If we allow software patents then we give more power to companies like microsoft to protect their markets. Large companies are better placed to file patents, litigate them and maintain them.

And so the debate widens again. Because this isn't a problem if you're a company manufacturing medicines as you're competing against similarly large companies. But software is easier to make and the only thing patents will do in software is to raise the cost entry to the point where only large companies will compete. I'd say that was inefficient.

Of course, this encourages further criticism of the patent system, particularly as software is also likely to be a complex product, for which no single patented technology is likely to be sufficient to bring a product to market. And where complex products involve many patents, all the system achieves is cross-licensing negotiations. And once again we find that large companies are favoured by this and will over patent (and patent obvious things) in order to strengthen their hand in negotiations. This is also inefficient, as we are no longer encouraging innovation, we are just patenting lots of things.

And so we reach the crunch point. To return to the audio manufacturers, do we offer some protection to them against the people profiling their machines, and allow them to license? (would they even bother licensing when they know that they can sit on their hands and let people pay $10k+ for the hardware?). Is it possible to offer a nuanced scheme of protection that does not raise the cost of market entry, does not allow obvious patents, and offers varying protection to varying areas of innovation? I have yet to hear anyone propose such a system.

I suspect this is because weak critics of software patents alone who follow the logic of their arguments become strong critics of the patent system itself. And it's been interested to note that the Economist (and so perhaps the voice of fiscal conservatism/market liberalism) has become increasingly critical of the patent system in relation to complex industries.

The point remains, however, if you want to argue against software patents you have to start with both feet inside the patent system and argue outwards on an economic basis. Obviousness is a peripheral point.

Andres Guadamuz said...

Excellent comment!

I tend to agree with you, but I find that the obviousness problem is not as peripheral as you would argue. The problem with patent quality is that too many bad patents pollute the system for all, and obvious patents are pollutants in the system. The EPO generally does a good job with patent quality, while the USPTO is an example of a bloated and inefficient body that is in the job of issuing bad patents.

kieran said...

I agree that bad patents pollute the system, and the example you cite in the original post is a particularly bad one. But it is not itself a principled reason to oppose software patents, as the implication is that non-obvious software patents would be justified/economically beneficial.

Bad patents may pollute the system, but I think the debate needs to address the negative externalities that the system itself produces, and indeed a pollutant-free system would still produce, and whether these can be justified by the benefits.