Submission by Linux Users Victoria, Inc., on the Options Paper for the Review of the Innovation Patent System.
Secretariat Advisory Council on Intellectual Property
PO Box 200 Woden ACT 2606
Dear Mr. Carl,
The following is the submission by Linux Users Victoria, Inc., on the Options Paper for the Review of the Innovation Patent System.
Linux Users of Victoria (A0040056C) is one of the world's largest and oldest Linux user groups, with close to 1500 members and with regional groups in Ballarat, Geelong, and Shepparton, as well as our Melbourne-based group. Linux is the most deployed operating system on supercomputers, webservers, routers, and other critical IT infrastructure. The Linux kernel is also incorporated in the Android system on mobile phones and tablets.
We address our response specifically on the issue of software patents.
President, Linux Users of Victoria
October 4, 2013
1. Core Proposition and Recommendation
Our core proposition is that Standard Patents and especially Innovation patents, attentuate innovation in software. If the purpose of the review includes the stimulation of innovation in software, then software (specifically "computational ideas"), should be excluded from the patent system as a whole, and especially Innovation Patents.
The existing patent system does not cover software in a particularly useful manner. A patent provides a temporary monopoly as an incentive for innovation. This arguably works with physical products, but in software the monopoly is not granted over the actual code (which would be covered by copyright if asserted), but the algorithms and ideas used and expressed in the software. This is particularly damaging considering that even a small software product, produced over several months with a small team of developers, can easily contain hundreds of computational ideas independently developed. The time period in Innovation Patents (8 years) and Standard Patents (20 years) is excessive for the industry.
2. Rational and Empirical Evidence
In software, code reuse is considered to be the best and appropriate practise, taught from the most basic and introductory programming classes. Programming is primarily about the implementation of ideas, but if the ideas themselves are patented then software cannot develop efficiently.
As a result of software patents, software written in any programming language (even a language not yet invented) can infringe by implementing a patented idea. Software developers, working on their own code, can find themselves paralysed if they wish to be attentive to the existing patent system, or liable if they are not. In either case, innovation suffers.
Audio development has been particularly hampered by software patents, with a large number of projects (Ogg Vorbis, Virtual Dub, Broadcast 2000, BladeEnc etc.) all requiring programmers to conduct patent searches and then program around them. From the perspective of software developers, this is a waste of time, money, and effort that would be best spent on software innovation. The same applies to graphics developers with patent restrictions on GIF image formats limiting the GD Library (this particular patent has since expired), the patent on Pantone colour matching limiting image manipulation programs, hinting mechanisms in TrueType fonts.
Even with simple system utilities, such as the development of compression algorithms, have been hampered by the presence of software patents such as bzip and zlib development.
A compiler suite such as GCC (GNU Compiler Collection), the most common compiler for scientific computing, is significantly less efficient than it could be, due to patents in optimisation algorithms, register allocation, and memory checking. Programming development in tree visualisation has been limited because of patents, as has the implementation of virtual functions.
These are far from trivial instances. If GCC's efficiency has been reduced by a mere 1% due to software patents, this means that Australian scientific research, on High Performance Computing clusters alone, loses billions of hours of compute time per annum.
Because any major piece of software contains hundreds, if not thousands, of software ideas it is impractical, damaging, and anti-innovative (especially for small and medium enterprises) to ensure that they have not breached any existing patents in their development. This is prior to even negotiating patent licenses, which in themselves would be extremely limiting to developers of free and open-source software (FOSS), on which most academic and scientific computational research depends, and which is necessary for widespread secure critical computer infrastructure throughout Australia. FOSS is particularly disadvantaged by software patents as it allows scrutiny of the code (which also assists its security), whereas proprietary software requires disassembly or reverse-engineering. That is, the best type of software is disproportionally restricted by allowing for the patent of computational ideas.
As Ben Klemen's (http://spectrum.ieee.org/computing/software/software-patents-dont-compute/) has pointed out, following Church and Turing, that there is no clear boundary between mathematical algorithms and software. If mathematical equations and laws of nature already exist, then no idea expressed in software, should be subject to a patent, and certainly not for an Innovation Patent, which lowers the criteria for this anti-innovative system.
This submission was written by Lev Lafayette with contributions and reviews by Les Kitchen, Rodney Brown, and Daniel Jitnah, committee members of Linux Users of Victoria.