Difference between revisions of "Why software is different"
(→Mathematical nature: I think argument is stronger now (with link). otherwise a naysayer would point out how software is a solution/application like any other engineering)
(→Mathematical nature: a bit of rewording)
|Line 16:||Line 16:|
:''(see also: [[Software is math]])''
:''(see also: [[Software is math]])''
The rules that exclude software from patentability do so on the grounds of similarity with mathematics: you should not be able to patent pure expressions of mathematics and logic [[#Software is unlike applied science|
The rules that exclude software from patentability do so on the grounds of similarity with mathematics: you should not be able to patent pure expressions of mathematics and logic [[#Software is unlike applied science|no novel invention in the physical sciences]]. This is exactly what we have, either when we consider software as text, or when looked at as a set of steps (algorithms) being carried out by a machine purely to determine a new answer to something, changing saved information, and communicating the results (acts that require just a mindand trivial uses of physical parts ).
Revision as of 06:47, 24 October 2010
Note: this is not a page about why software ideas shouldn't be patentable.
This page only documents why software can be treated as a distinct domain i.e. why the patent system doesn't necessarily produce the same effects for software as it does for mechanical engineering or pharmaceutical research.
Conversely, software is similar to innovative fields such as literature, film and fashion.
Studies on usage patterns
The 2008 Berkeley Patent Survey revealed big differences between software companies and biotechnology companies in their use of patents.
"[s]oftware patents are more than twice as likely to be litigated as other patents" - Bessen & Meurer, at 22.
- (see also: Software is math)
The rules that exclude software from patentability do so on the grounds of similarity with mathematics: you should not be able to patent pure expressions of mathematics and logic or their use by an existing machine leading to no novel invention in the physical sciences. This is exactly what we have, either when we consider software as text, or when looked at as a set of steps (algorithms) being carried out by a machine purely to determine a new answer to something, changing saved information, and communicating the results (acts that require just a "mind" and trivial uses of existing physical parts used in their ordinary ways, much as a human uses a mind while using eyes and ears to acquire extra information and finally a mouth and hands with basic utensils to communicate and record what was processed in the mind -- no novel invention in the physical sciences got created).
Software is highly abstract: you can't look at two pieces of software from the outside to see whether they look the same. Two engineers will describe the same piece of software in very different terms. It's therefore highly subjective whether the ideas embodied by one piece of software are essentially equivalent to the ideas in another. In short, there is no objective test for infringement.
Problems and solutions are tied
There are many ways to cure rubber. These methods can be used to cure various types of rubber and the output will be usable for any product that requires cured rubber. The inputs and outputs don't define what method of rubber curing must be used - the curer is not required to use a specific method.
In contrast, for a software video player to be functional, it must read videos in precisely the format in which videos exist. That is to say, software video players have to read MPEG video formats, for example, in order to be useful. So a patent excludes developers from the whole field rather than excluding them just from one method of doing that field of activity.
The test of obviousness
Because software is abstract, the test for whether an idea is obvious "to one skilled in the art" is very difficult to apply. The test cannot be based on the amount of effort expended in working out the idea, because the least obvious ideas are often the simplest. Nor is it clear what it means: what if an idea is obvious to a thousand of the best programmers in the world, but not obvious to a million others? Simultaneous and independent invention is not unique to the software field, but it is certainly a characteristic feature of software that good programmers will tend to come to the same solution independently.
Because software is invisible, determining prior art is notoriously difficult. Many software patents appear to be granted despite the ideas being well known to practitioners, but it can be very difficult to find acceptable documentary evidence that the ideas were already in widespread use. Most programmers do not publish written papers about their work, for a variety of reasons: they are too busy, they don't consider the idea significant, or they don't want their competitors to know what they are doing.
Programming is by its nature innovative. Programmers have new ideas every day, in a way that is not true of engineers in other fields. If every new idea were patented we would have billions of patents rather than merely tens of thousands. Many of these new ideas have little significance, but there is no meaningful way of distinguishing those that deserve a patent monopoly from those that do not.
By the same token, it is impossible for programmers to check every time they write a line of code whether it infringes a patent. Programmers take the risk of patent infringement every day, because they have no realistic alternative. It is unfair that society should penalise them for this.
- (see also: Low risk)
The devil's bargain can be called a bargain only if it benefits someone besides the devil. A major justification for the bargain, the encouragement of entry by new people with new ideas into industries with high cost of entry, i.e. high risk, does not apply to software development. Software development is exceptionally cheap, thus low risk.
Software is unlike applied science
Software has one fundamental difference as compared to other sciences, and that's of predictability. Other industries, such as the pharmaceutical industry and manufacturing, require a significant amount of research into nature and a sort of alchemy in order to learn the secrets held by nature alone necessary for progress. Patents protect the investment that go into such research, in exchange for sharing the acquired knowledge with the world so it will not be lost.
With software, it's different. By its very definition, the outcome of software is always deterministic and predictable. It is fairly predictable that combining disk compression with an operating system will yield an operating system that can squeeze more files onto a disk. It is fairly predictable that one-click ordering will be faster and more convenient than two- or three- click ordering. Neither of these require any significant amount of research. A main purpose of a patent is to incentivize disclosure of nature's relevant secrets for the public good, with exclusivity that comes at public expense.
Unlike the recipe for penicillin, the "disclosure" that one-click ordering is convenient did not come after a billion dollars of research. There is no good reason it should be rewarded at the public's expense the same way.
How patents on different domains affect society
One thing to keep in mind is that plenty of things are excluded from patentability. Being excluded doesn't mean that software is special.
For the manufacturing of cars, you have to consider how patents will affect:
- the cost of mass production
- the impact on quality/safety of what's offered to citizens
- the impact on the economy overall
Campaigners against software patents are usually not general experts on those topics, so we might not know if innovations in car manufacturing should be patentable or not.
For software, there are similar questions to the three above, plus there is the fourth question of individual liberty and the effectiveness of communities.
This question isn't pertinent to car manufacturing because individuals and communities don't manufacture cars in significant quantities - or at all.
If someone patents a method for making a car, that doesn't reduce people's liberty. Making cars requires a lot of cash and materials, and there are already many laws that places regulations and restrictions on making cars. So people are already excluded. Adding a patent problem doesn't change anything.
For one thing, software is mass produced by individuals and groups who don't get paid directly for that work (or do it for non-commercial reasons). Adding the cost of the patent system is unfair to these people.
Communities write great software (a community wrote most of GNU/Linux). People should continue to have their right to participate in the development and distribution of software, and to continue to benefit from the work of the vast community that develops the software that people use. Patents would create problems for individuals' liberty, and for the general software development which society as a whole benefits.
Related pages on en.swpat.org
- Why focus only on software?
- Software is too abstract, patent quality is bad
- Software patents are unreadable
- 2008 Berkeley Patent Survey - showing differences between patenting of software and biotech
- Breaks common software distribution models - a patent problem unique to software
- Cost of defending yourself against patent litigation
- In Defense of Software Patents, by Martin Goetz, argues that software development is just like machine development
- Are software patents evil?, by Paul Graham, argues that software patents are the same as hardware patents
This wiki is part of the End Software Patents (ESP) campaign (donate). For more information, see: