On 5/19/07, [EMAIL PROTECTED] <[EMAIL PROTECTED]> wrote:

It doesn't seem as clear-cut to me.  An arrangement of transistors
is patentable.  A ROM is a complex arrangement of transistors.
By implication you can patent the pattern stored in the ROM.
(I'm not referring to a PROM.  But a ROM, where the wiring actually
encodes the data.)

Many decades ago I heard a rumor that IBM implemented a an algorithm
in transistors in order to patent it.  At that point the patent
was claimed to cover software implementations of the same thing.

Historically I have been a chip designer.  Did it for 3 decades.
I've also done a moderate quantity of software.  To me the distinction
between the two is pretty hazy.  One of the first tasks in defining
a new ASIC is to decide how much of it is done in HW and how much
in SW.  The decision is based on cost, both silicon cost and cost
of development.  Is it reasonable that we can patent one implementation
but not the other?  Seems odd to me.


You base this idea on the belief that patenting the ROMs was the "right"
thing to do, that they are clearly patentable, and extend that idea to
software. But what if they're not? What about the other way? Say we
establish beyond a shadow of a doubt that software is not patentable because
it is a "description of an abstract idea or mathematical algorithm"[1].
According to existing patent law, this _should_ be a pretty clear and
accepted statement, "Logic and algorithms aren't patentable". Then we extend
from there the other direction, into the hardware. In that case the ROMs you
speak of would not _entirely_ be patentable. Only the physical
implementation or circuit layout would be, but not the logic that they
encode. A real zealot could even take it further and argue that unless their
physical attributes were particularly novel, say they created some new kind
of logic gate, that the physical attributes are really just the "language"
that expresses the logic, and that's not actually patentable either! In that
case, only the packaging of the ROM would be patentable, but then you start
losing footing because of prior art and obviousness.

In this light, even Marbux's point about the "bright line" of patentable and
unpatentable being between physical and not becomes muddy. As he points out
though, the line has to be drawn somewhere. I'd argue that the physical /
not break is a fairly good place to do it, but it must also specifically
state that the patent covers only the particular physical implementation of
the logic, not the logic itself.

Patents were created to protects "inventions", and it has been widely
established that software is an "expression" like math, music, art, or
literature. That's why it's already protected by copyrights. And like art,
music, and literature, software (and more generally, logic and algorithms)
are built upon the prior art of others. Each piece being a derivative of the
work that came before it. Since patents are designed to protect "certain
details of a device, method, process or composition of matter (substance)
(known as an invention <http://en.wikipedia.org/wiki/Invention>) which is
new <http://en.wikipedia.org/wiki/Novelty_%28patent%29>,
inventive<http://en.wikipedia.org/wiki/Inventive_step_and_non-obviousness>,
and useful <http://en.wikipedia.org/wiki/Utility_%28patent%29> or industrially
applicable <http://en.wikipedia.org/wiki/Industrial_applicability>."[2]. I
think in that portion of the definition of what is patentable is the core of
our problem, the word "process". How do you define the difference between a
"process" and an "algorithm"? It would be difficult to do so, and I don't
believe that the USPTO has sufficiently differentiated the two. Perhaps they
should have left out the word process?

Arguably, the entire software patent issue is a result of monied
corporations wanting to have their cake and eat it too, which was enabled by
a broken patent system that issued patents that it should not have. This
point of view is further strengthened by the _mountains_ of questionable
patents that have been issued in the last 20 years, for software or
otherwise. That's the real reason that this has become such a quagmire, the
organization that is supposed to be regulating the system is so incompetent
and self-contradictory that it breeds confusion rather than increasing
clarity. Add to this dodgy legislation like the DMCA, and the slide to
treating unpatentable "intellectual property"[3] (I hate that phrase) like
the patentable "real property" of physical inventions becomes easy and fast,
and here we are.

"The illusion of freedom [in America] will continue as long as it's
profitable to continue the illusion. At the point where the illusion
becomes too expensive to maintain, they will just take down the
scenery, they will pull back the curtains, they will move the tables
and chairs out of the way and you will see the brick wall at the back
of the theater." --- Frank Zappa


Well put Mr. Zappa.

[1] -
http://www.freesoftwaremagazine.com/blogs/software_aint_patentable_damn_it
[2] - http://en.wikipedia.org/wiki/Patent ( I know it's lame to refer to
wikipedia, but I couldn't find another good online reference for this point
quickly.)
[3] - http://www.uspto.gov/web/offices/pac/doc/general/whatis.htm
--
-Regards-

-Quentin Hartman-
_______________________________________________
EUGLUG mailing list
[email protected]
http://www.euglug.org/mailman/listinfo/euglug

Reply via email to