Photo by Indrani Soemardjan
I have a visceral aversion to software patents, after several incidents in my career where I’ve been forced to deliver a lower quality product to the customer purely because of absurd patents. We ended up having to cut out not only a ghost car mode from one game, but any way of comparing your current time against your previous laps since that was also covered. I was also involved in finding prior art for a crazy patent for retiming video based on audio output, something that had been done for years but still got granted.
I know I’m not alone, pretty much every engineer I know agrees the system is broken, but I’ve had a hard time explaining that to people outside the industry. I recently came across a new book that might help explain the impact, Patent Failure by James Messen and Micheal J Muerer. Tim Lee has an extended review in three parts (1, 2, 3) on Megan McArdle’s blog, and for me the most fascinating evidence were these two graphs:
The dotted lines on each graph are the profits made from licensing patents, and the solid lines are the total costs to alleged infringers. The top shows the chemical and pharmaceutical industries, where the costs of defending against patent litigation are a small fraction of the licensing fees. By contrast, in all other industries a lot more money changes hands to defend infringement actions than is made by licensing.
The graph only goes up to 2000, and if my experience is anything to go by, the gold rush by patent trolls to sue deep pocket companies has only increased since then. This seems like a pretty clear and quantitative sign that something is rotten in the state of non-chemical patents, since infringement penalties were meant to be a deterrent to trespassing on someone else’s IP, not the main way of making money from inventions. Either the non-chemical industries are full of bad actors stealing other people’s ideas, or it’s simply not possible to avoid being vulnerable to patent infringement actions, since there’s obviously a big financial incentive to not infringe and it’s still happening.
Tim digs up a couple of reasons why there might be such a difference between the two sectors. With chemical compounds, there’s a very standard way of describing them in a formula, and so it’s a simple process to search patents and see if what you’ve discovered has already been claimed. By contrast, with software patents it’s literally impossible to first find all the relevant patents, and then tell if you’re infringing.
Searching patents for software and expecting to find all or even most is like searching the web for every page that talks about a subject. There’s an almost infinite set of natural language variations that might find what you’re after, and any search will also bring up a large number of irrelevant results. Worse, it’s not obvious even to patent attorneys if a given piece of engineering infringes a patent. Tim gives the great example of the point-of-sale kiosk patent that suddenly got accepted as applying to websites with no physical presence. Not even an attorney would have imagined that was relevant until the court decision accepted its broad scope.
The original rationale behind patents was to distribute information about new inventions in exchange for a temporary monopoly for the inventor. Software patents are almost entirely useless as an information source, and it’s pretty obvious from these figures that their main effect is as an employment boon for lawyers and patent trolls who produce nothing that helps society. The damage to the software industry in stifled innovation is huge, when patents were supposed to do exactly the opposite.