From phm@a2e.de  Sat Sep 16 08:10:52 2000
Date: Sat, 16 Sep 2000 10:10:52 +0200 (CEST)
From: PILCH Hartmut phm@a2e.de
Subject: [Patents] Gregs myth

Greg Aharonian wrote in Patnews:

>             SOFTWARE SHOULD NOT BE PATENTED
> 
> He starts out this section with "The first myth is that software simply
> should not be patented."  Actually, I agree with him that this should be
> a myth, that software should be patentable.  Unfortunately, he rambles
> on talking to software engineers about Linux, copyright, and expression in
> three paragraphs, when he could of, and should of, been more professional
> and said that "In light of Turing/Church/Post, hardware/software codesign
> tools and the Doctrine of Equivalents, if hardware is patentable, then so
> should be software".  That's all you have to say to a bunch of software
> engineers to convince them that software should be patentable, that is if
> you know anything about software engineering and are writing for a
> professional software engineering journals.

I don't think a simple "doctrine of equivalents" or any other systematic
reasoning is enough to justify the application of the patent system.  The
primary justification for applying the patent system lies in the economic
results, finding a systematic explanation is only secondary. And below I
show that it can be done nicely, even given the equivalence of hardware
and software.

Besides, I think it is a myth of Greg's own making if he says that the
"doctrine of equivalence" reasoning is enough to persuade any professional
software engineer of the need to make software patentable.  Apparently
lots of people in that group are not persuaded, and a survey from 1998
sets this figure to 84% (among software developpers in large American
software companies).

If you set a borderline between hardware and software, you can get some
difficult cases that require further reasoning for making a
decision.  Which doesn't invalidate the wisdom of setting the borderline
between sw and hw, as the European Patent Convention does.

You could for example have the case of an anti-blocking system, which is
based on new insights into the effects and possible use of natural forces
in a car on a road.  From these insights a solution is derived, which can
be implemented in software on an existing engine control device.  Such a
solution would be patentable in Europe, but the software itself would not,
nor would a manual which teaches the solution.  This may create some
slight practical difficulties for the patentee, who would probably prefer
to be able to ban the publication of the software, and, even better, the
publication of instruction manuals and even of his patent description.
But you can't have your cake and eat it: if you want to monopolise the
solution, then you must allow the publication of descriptions, which can
also be formulated in formal languages.

This is also a systematic reason for drawing a borderline between software
and hardware.  Software patents in the extreme case will infringe on
themselves:  in order to properly teach the man skilled in the art, you
would need to give an implentation example in a standardised programming
language.  But this implementation example itself would infringe on the
patent, if a "computer program" can be patented.

Inventions therefore, even for systematic reasons, need to be such that
the publication of software does not destroy the monopoly.

In the ABS case, this is still true, because the sale of the car can be
monopolised.  Real theoretical difficulty will however arise if the car
itself is controlled by a universal computer which can be reprogrammed by
the car's user.  Here we are in a gray area, but there are several
approaches at solving this.

One way is to negate patentability of the ABS innovation because it
requires nothing more than a program for a known device.  Another is to
uphold it, because it is based on new insights into natural forces, but to
make clear that it becomes less and less enforcable the more it can be
implemented on a universal computer.

In principle, solution which is not based on new knowledge of the real
world should not be considered inventive, simply because it could have
been found by an AI robot.  This is true for most "software patents" and
also quite a lot of "hardware patents".

This requirement will certainly also kill a lot of hardware patents, and
justly so I think.  Granting patents for things that can be calculated
from known models without costly real-world experiments makes little
sense.

Maybe exactly those patents would be killed, for which the "doctrine of
equivalents" holds?

-phm








