From xdrudis@tinet.org  Tue Dec 12 10:50:41 2000
Date: Tue, 12 Dec 2000 11:50:41 +0100
From: Xavier Drudis Ferran xdrudis@tinet.org
Subject: (fwd) Re: [Patents] Report on software patents.


Gregory Aharonian va escriure el Mon, 11 Dec 2000 21:24:29 -0500 (EST)  :
>
>Hartmut writes:
>
>>At least if the process is to be program controlled, it must use
>>natural forces to cause a matter transformation not predictable
>>by computational means.  The result of sorting is computable.
>
>Which bans large numbers of mechanical, chemical, biological, and
>electrical patents as well.
>
>Look with a supercomputer and a superSPICE (a computer program for
>circuit analysis) I can predict computationally whatever "matter
>transformation" is being caused by an analog or digital circuit
>being patented.   Using mechanical CAD programs, I can predict the
>by computational means the matter transformation caused by a large
>number of the mechanical design patents being applied for (just
>determine the Langranging and integrate).
>

It is OK with me if you can predict it once you have invented it 
(imagined it). To get a patent you should think of the circuit first
(not just any circuit, one that represents an invention, is not obvious, 
uses physical forces in a new way, and any legal requirements for a 
patent) then simulate it, and then probably test a prototype (possibly with 
some iterations in the process to get it right). Then you are the inventor
and you may get a patent. It is OK to use tools like software to help you.
That's what they are for. I doubt that if what you invent is new enough to 
be given a patent, the existing simulators help too much, but if they 
do, then they are wellcome.

If you can invent it just by having a program generate
random circuits and SPICE analyze them to see if they work, then that
means the patent owner should be the random circuit generator, not 
you, and since computer software is not considered people in Europe, 
your random circuit generator couldn't be the patent owner,
the patent wouldn't be possible. 

The Eurolinux directive proposal does not talk about computationally 
predictable things. It talks about computer programs.

>But once again I will point out that this just shows that there
>is nothing peculiar about software to treat it different for
>patenting purposes.
>
What is peculiar about software is that it is information and 
a process to transform information at the same time. You have been 
told so a few times already, and there is a sentece about it 
in the directive proposal. A chemical process is not a molecule. 
An optical process is not light, a mechanical process is not a force.
An information process *is* an algorithm. So a chemical process is 
patentable, an optical process is patentable, a mechanical process is 
patentable, an algorithm is not patentable. The distinction is coherent.

And if you didn't like chemical, optical or mechanical patents, because 
you think they are not economically good for society, then that might or 
might not be true, but it is not what we are talking about. We are talking
about software patents, and software is different from chemistry or 
optics. 

>Any definition of how much "technical effect/industrial application"
>has to appear in a software patent will easily be met by filling,
>at great expense to the inventor, the patent specification with lots
>of boring examples.  The US courts finally realized the futility of
>trying to come up with such definitions, and you Europeans are going
>to waste a lot of time doing so as well.  Definitions are NOT the
>way to solve the software patent problem.
>

Definitions are the only way to talk about anything, unless you want 
dadaist law. If you don't define what is patentable and what isn't then 
you end up patenting pictures, novels, dreams and email messages, so 
be careful because I could patent your arguments if I ever see one.
Definitions use natural language, which is the best we have for law. 

Natural language is fundamentally ambiguous, so you can never reach a 
perfect definition, and you can always bend the best definition to 
make it mean something different. There is no useful unambigous way to 
code law. The only way to go is to try to avoid bendings in what you can 
by writing clear definitions in the laws, and expect the people that use 
the laws to be honest and don't bend them, and have mechanisms to 
redress those bendings. We have lawyers so that there are people who is 
trained to interpret law. They know how to bend it, but they know how much 
it should be bent and how. A stiff law is no good, a law so much bent
as the EPO is doing is too bad. So we must stop the EPO. One way would be to
include more democratic control and another one is to pass more clearly 
definitions on what they can do or not, so that they can't bend them
enough to do harm. It is not a perfect solution, but as with most 
things in real life, you have to do with an imperfect solution and that's 
better than no solution at all.

The directive proposal is clear enough, and I can't find a way to make
it any better, so it is a good solution. Europe is lucky not to have Greg as a judge


--
Xavier Drudis Ferran
xdrudis@tinet.org




--
Xavier Drudis Ferran
xdrudis@tinet.org







