On 31/01/2011 12:53, Andy Farnell wrote:
..
AXIOM: Ideas should not be patentable. Period.
Er, they aren't, and never have been.
Any patent has to describe a ~device~ - a machine, a thing that can be
"built" - hence the ubiquitous term "preferred embodiment". If an
"invention" cannot be expressed in a buildable machine or device, it
cannot be granted a patent. the patent is supposed to supply sufficient
information for someone ordinarily "skilled in the art' to build it,
while also being something that that person cannot be expected to invent
for himself based on that ordinary art.
You can't patent the 'idea' of a "sorting machine"; but in principle you
can patent a specific buildable machine that implements sorting to solve
some problem. So you may be able to patent a machine that sorts
potatoes, threshes wheat, etc. Or, given the existence of such a
machine, patent a new one that does it faster/cheaper/safer/whatever
using some defined 'embodyable' process.
The modern difficulty arises simply because we (public at large, etc,
and so also patent offices) now recognise the notion of a "virtual
device" - a machine that does something, solves some "problem", but
which exists predominantly in software form. So a machine that
reverberates a sound is a patentable device. An audio plugin is not in
this sense "an algorithm" any more than a spring reverb is an algorithm
(though it can of course be modelled mathematically)- it is [claimed to
be] a ~device~ that does something. The distinction is that multiple
devices that are outwardly identical, may embody different processes
internally. Maybe literally the only difference is that one does it
faster than the other. So you can't patent the idea of a reverb plugin,
any more than you can patent the idea of a weaving machine. But you may
be able to patent an embodiment that does it
faster/cheaper/better-blahblahblah.
Hence the classic original FM patent. It uses multiplication (can't be
patented), and a trigonometrical identity (can't be patented), and even
the idea of modulating one oscillator with another (can't be patented),
but does so in a particuarly novel and structured way so as to solve a
problem - how to generate lots of dynamically variable partials, whether
harmonic or inharmonic, controllably, at audio rates, very
economically; at a time when the ordinary art consisted of fundamentally
analog subtractive or additive synthesis. FM already existed in the
context of radio transmission; the Chowning patent applied a known
principle in a new area of application; and included the instructions
and knowledge required for controlling it (theory of Bessel functions, etc).
Thus, you can't patent the known general process (algorithm) of
frequency modulation, but you manifestly could patent its particular
embodiment in a music synthesiser.
Thus also, a patent that describes some new form of partitioned
convolution could be patentable within the established terms of
reference for a patent ~if~ it describes some "non-trivial" new
improvement to existing methods, for some ~embodyable~ device. It is not
trying to patent the "idea" of partitioned convolution (it can't,
obviously). Whether it is in fact novel, or beyond the skills of those
classified today as "ordinarily skilled in the art" is another question
entirely.
Now I may well dislike the idea of software patents myself, and agree
there are far too many trivial non-novel patents around, but thinking
they can somehow disappear is virtually on a par with hoping to design a
perpetual motion machine. Governments around the globe (at least, if the
UK is at all representative) are pressuring universities to "transfer
technology" - realise as much economic value as possible from
innovation, so that the pressure to patent first, then [optionally]
publish, is increasing exponentially. Vice-Chancellors no long ask "how
does this increase knowledge?" but "do we own it?". Once it was realised
that software could make money (wish I knew how!), the horse had bolted
from the stable, never to return.
So, arguments in favour of reform, and better scrutiny/peer review etc,
are all good to make; but the modern software patent is a cash cow not
only for companies (pity the poor individual inventor indeed!) but also
de facto for government exchequers everywhere, and is I suspect here to
stay.
Richard Dobson
--
dupswapdrop -- the music-dsp mailing list and website:
subscription info, FAQ, source code archive, list archive, book reviews, dsp
links
http://music.columbia.edu/cmc/music-dsp
http://music.columbia.edu/mailman/listinfo/music-dsp
