That anti-patent pamphlet I mentioned

[Alex Hudson]

On Sun, 2002-12-15 at 11:46, Arnoud Galactus Engelfriet wrote:
> > And please also realize what you just said.  You said that everything will
> > be patentable.  Is this truely what you want?
> 
> I don't think I said that. Patents should be available for
> all technical inventions, and I believe it makes no difference
> whether the invention is based on software. But a purely
> mental technique, or a method of doing business should not
> be patentable.

This would be the problem I have with software patents - modulo all the
other arguments that are being put forth, I don't believe software is
technology. 

Software started as, and is, a branch of mathematics. Although the
systems being designed are extremely complex (e.g., wordprocessors),
they are no different to the simple programs that children write.

All software can be reduced to some set of math. Often, software is
designed upon a few core mathemetical ideas, which make up the basis of
the software. For example, a 3D CAD package. This would make use of a
number of areas of maths: vector/matrix math would be used to describe
the objects. As a side-point (more of which later), althought the
objects exist in the CAD, that's fairly meaningless: they don't exist in
real life, although obviously their ephemeral form is useful to us in a
number of ways.

The breakdown of software to maths is best shown by the "Gallery of
DeCSS Descramblers", at http://www-2.cs.cmu.edu/~dst/DeCSS/Gallery/.
It's aimed at showing the DMCA to be a bad law, and hence a number of
the examples are not relevant to this argument, but the basic argument
is the same: if you treat software as you would technology, the logical
extension of your argument has ridiculous implications. Here, the
application of the DMCA brings legal doubt over publishing mathematical
theorums, for example, but also (in the Steganography wing) publishing
certain (probable) prime numbers, which are equivilent to the software,
the algorithm, and hence "covered".

The examples of DSPs given were actually something I argued about with
Xavi when I first learned about the problems with software patents. On
the face of it, they are a very good example of why we should have
software patents. Let's say I design a DSP that is going to function in
an audio circuit as an echo chamber - a specific echo chamber. The echo
chamber I wish to replicate is a patented system (perhaps like a Leslie
Cabinet - I *think* that is patented..), and the obvious way for me to
go about this is to model in software the physical properties of the
cabinet, and replicate them. That way, the sound I generate is
identical.

Now, is the DSP covered by the patent? The obvious answer is yes: it's
doing exactly the same thing - causing the same technical effect - but
just in software, so it should be covered by the same patent. The
argument is, remember, that a technical patent should be available
whether or not the device is software driven.

However, I think that the obvious answer is the wrong answer, and it
becomes more obviously wrong if we look at it from the opposite
direction - coming back to my previous point that ephemeral forms can be
useful to us. That's not always the case. If, in my modelling of the
Leslie cabinet, I changed it so that the audio was bouncing off walls
several tens of metres away but not decreasing in amplitude, I have a
system with extreme echo but no audio degradation. 

Is that useful? As the model itself, yes it is useful, because I have
achieved a technical effect that was not possible before. However, it's
not physically useful: although I have modelled a cabinet, my model
bounces audio waves around with no loss of volume or clarity. Although
my audio effect depends on this, it's not something that I can reproduce
in real life: I could never actually *build* the cabinet, because no
such materials exist with those properties.

The fact that we can build such impossibilities in software does, I
think, point at the fact that software is different; radically
different. Take my altered DSP. If someone achieved this technical
effect physically, they would be hailed a genius. Someone achieving this
technical effect with software would not be similarly praised. 

To compare the technical achievements of a field such as electronics
with software is wrong, clearly wrong. Something achievable with
software may be completely unachievable with electronics, so to give a
patent based on the same technical acheivement is equally unbalanced. I
think the above example shows that the "technical effect" argument is
useless for software: yes, software can exhibit technical effects
similar to more conventional devices; but no, I don't think they can be
argued to be either inventive or novel or unobvious - usually the
contrary. A software patent in that context becomes a purely economic
device; there is no benefit to innovation from having it, and indeed it
would do great damage.

I think the reason people argue from the basis of "forces of nature" is
to do with the above reasoning: that the design of software is not
contrained physically in the same manner as other devices, and hence the
effects are not immediately comparable in terms of
innovation/achievement. 

The question that comes from that, I guess, is "are the technical
effects of software that are of sufficient invention that would warrant
a patent?". My answer is no. Software is constrained by a number of
things. Amount of memory - you need lots of memory to run some programs.
You often need a fast CPU. The definition of computability, though,
ignores these limits: the Turing machine has infinite memory and
infinite time to perform a computation. The real limit is on those
things which are computable, because not all problems are computable.
For some things, there is just no answer.

This is a very black and white limitation. Either you can compute
something, or you can't. For some problems, we're not sure yet - some
things which look to be uncomputable might actually turn out to be
computable - but that's our uncertainty, not a general uncertainty. For
those things that are computable, it's a fairly flat space: there are no
other constraints, other than your own time and your imagination.
Everything in that space is accessible, and it takes no stroke of genius
to find a new space, or discover a new plain. Hence, I don't see by what
criteria we could award patents, and don't see what could warrant
awarding one.

[This is beside all the other arguments: I believe the economic one to
be particularly strong - software patents should not be given simply
because they are not needed.]

Cheers,

Alex.


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
URL: <http://lists.fsfe.org/pipermail/discussion/attachments/20021215/28c90f58/attachment.sig>