🚀 go-pugleaf

In article <BURLEY.91Sep1202411@churchy.gnu.ai.mit.edu>
burley@churchy.gnu.ai.mit.edu (Craig Burley) writes:

>Admittedly the term "software patent" needs a better legal definition.
>One approach to finding one that I could use is whether human thought
>processes can substitute for the complete process described in the
>patent.  If so, it is a software patent.

It would be better to use "algorithm patent" rather than "software patent".

An algorithm is a mathematical object, like a number, or a vector space,
or a theorem, or a proof.  An algorithm should be no more patentable than
any of these other mathematical objects.

What should not be allowed is any charge of patent infringement based on
purely mathematical isomorphism.  One should have to copy some code or some
circuitry.  Something more than the mathematics should be involved.

>In summary, we should go back to the 1981-and-earlier standards for
>patentability, and abolish software patents retroactively.  Processes that
>include software may still be patented, but use of the software (equations)
>that is in the patent should not, by itself, ever constitute infringement.

This reads much better if changed to ... and abolish algorithm patents
retroactively.  Process that use software may still be patented, but the
use of the same mathematical algorithms as those used by the patented
process should not, by itself ...

Charles Geyer
Department of Statistics
University of Chicago
geyer@galton.uchicago.edu