Oh my god. I can only hope his views aren't mainstream. Quoting from the link:
"This doesn’t mean that moving forward software won’t be patentable, it just means that getting a software patent will be much more difficult than it ever has been. Software can be described by reference to a series of physical actions operating through gates. This type of micro level description of what happens is going to be required, which means getting a patent for software has just become much more expensive and time consuming."
The disconnect between how he (apparently) thinks software works and what really happens is mind-boggling. He seems to have absolutely no idea that the same program (even down to the same machine code) will create radically different "physical actions operating through gates" on processors with different micro-architectures (or with an emulator or ...). Even if you could get a patent based on a description of these "physical actions" in a specific execution context, such a patent would be practically worthless. It would be like having a patent on implementing your idea on a Pentium. Anyone could trivially get around that by choosing a Pentium Pro (or an Athlon or some random ARM core or whatever).
I'd really like to see him try. Even the most trivial operation will run to a few thousand pages of very fine print. Patent examiners will have a new use for applications: fuel.
"This doesn’t mean that moving forward software won’t be patentable, it just means that getting a software patent will be much more difficult than it ever has been. Software can be described by reference to a series of physical actions operating through gates. This type of micro level description of what happens is going to be required, which means getting a patent for software has just become much more expensive and time consuming."
The disconnect between how he (apparently) thinks software works and what really happens is mind-boggling. He seems to have absolutely no idea that the same program (even down to the same machine code) will create radically different "physical actions operating through gates" on processors with different micro-architectures (or with an emulator or ...). Even if you could get a patent based on a description of these "physical actions" in a specific execution context, such a patent would be practically worthless. It would be like having a patent on implementing your idea on a Pentium. Anyone could trivially get around that by choosing a Pentium Pro (or an Athlon or some random ARM core or whatever).