Path: sparky!uunet!think.com!mips!mips!munnari.oz.au!metro!cluster!
zale...@bme.unc.edu
From: zale...@bme.unc.edu (Rudy Zalesak)
Newsgroups: comp.patents
Subject: Patents and their form of expression
Message-ID: <4169@cluster.cs.su.oz.au>
Date: 20 Feb 92 18:08:05 GMT
Sender: n...@cluster.cs.su.oz.au
Lines: 26
Approved: pate...@cs.su.oz.au
Programs expressed in functional languages can look like
(in fact, ought to look like) ordinary expressions in
mathematics. If it is possible to express an algorithm in
ordinary mathematical notation, then that notation can be
used as a program itself- expressions in the language Mathematica
come very close to this; some work with TeX could erase
the distinction entirely. Stan Wagon's book on Mathematica
includes working Mathematica code for the RSA algorithm which
is remarkably close to simply being plain mathematics
("FirstPrimeAbove[10^100]").
I think this shows it doesn't make sense
to distinguish between the explanation
of an algorithm in a textbook and the implementation of the
algorithm in a computer language. Some pro-patents writers here
seem to think that only the advent of natural-language understanding
programs (able to read patents and compile them automatically!)
would erase the distinction between speech about an
algorithm and the implementation itself. But this example
shows that we are much closer to that situation than they think.
Now, consider the case of a translation of Sedgewick's *Algorithms* into
Loglan... :)
Rudy Zalesak
Path: sparky!uunet!munnari.oz.au!metro!cluster!l...@ai.sri.com
From: l...@ai.sri.com (Kenneth I. Laws)
Newsgroups: comp.patents
Subject: Re: Patents and their form of expression
Message-ID: <4239@cluster.cs.su.oz.au>
Date: 24 Feb 92 18:22:25 GMT
References: <4169@cluster.cs.su.oz.au>
Sender: n...@cluster.cs.su.oz.au
Organization: Computists International
Lines: 31
Approved: pate...@cs.su.oz.au
> Programs expressed in functional languages can look like
> (in fact, ought to look like) ordinary expressions in
> mathematics.
Agreed. I suspect that there's no problem here, but I'd like
to know more about the legal definition of "algorithm." We
know that mental processes are not patentable, nor are laws
of nature (and, by inference, mathematical formulas). If there
is only one way to accomplish a result, courts have considered
this a law of nature or otherwise unpatentable. (Am I right?
Richard Stallman once told me "this was not his understanding,"
but I consider it reasonable and believe it's true.)
On the other hand, processes are often patentable. A process
appears to be an "algorithm" applied to a particular useful task.
The patent does not cover the algorithm itself, just the application.
Again, am I right?
The arguments I've heard invoking the unpatentability of algorithms
(and hence of software, although I think that's too big a jump) have
never mentioned the legal or philosophical basis of this exclusion.
Can anyone fill me in? I think it would help to clear up how this
exclusion does or does not apply to software.
-- Ken
--
Dr. Kenneth I. Laws, (415) 493-7390, l...@ai.sri.com.
Moderator of the Computists International AI/IS/CS mutual-aid association.
Ask about my weekly online career newsletter, The Computists' Communique.
[mod- good point re algorithm.
For a process patent covers the process per se and this appears to
me to be very close to a recipe and in turn very close to an
algorithm. Hmm.. food for thought....
]
Path: sparky!uunet!munnari.oz.au!metro!cluster!n...@Think.COM
From: n...@Think.COM
Newsgroups: comp.patents
Subject: Re: Patents and their form of expression
Message-ID: <kqo2jfINNorf@early-bird.think.com>
Date: 29 Feb 92 05:49:04 GMT
References: <4169@cluster.cs.su.oz.au> <4239@cluster.cs.su.oz.au>
Sender: n...@cluster.cs.su.oz.au
Organization: Thinking Machines Corporation, Cambridge MA, USA
Lines: 21
Approved: pate...@cs.su.oz.au
NNTP-Posting-Host: gandalf.think.com
In article <4...@cluster.cs.su.oz.au> l...@ai.sri.com (Kenneth I. Laws) writes:
>On the other hand, processes are often patentable. A process
>appears to be an "algorithm" applied to a particular useful task.
>The patent does not cover the algorithm itself, just the application.
>Again, am I right?
I think that a "process" has to control something that causes matter
transformation of some sort. Thus, a process for curing rubber is
patentable, as would be a cooking recipe. This is presumably also the
basis for many patents in chemistry and biotechnology.
An example of an algorithm, which wouldn't be patentable, is long division.
It's purely a mental process, although physical devices (paper&pencil, an
abacus, a computer, etc.) might be used to assist the person (mainly
substituting for his memory or slowness).
--
Barry Margolin
System Manager, Thinking Machines Corp.
bar...@think.com {uunet,harvard}!think!barmar
Path: sparky!uunet!munnari.oz.au!metro!cluster!0001811...@mcimail.com
From: 0001811...@mcimail.com (Carl Oppedahl)
Newsgroups: comp.patents
Subject: Re: Patents and their form of expression
Message-ID: <4265@cluster.cs.su.oz.au>
Date: 29 Feb 92 06:13:54 GMT
Sender: n...@cluster.cs.su.oz.au
Lines: 26
Approved: pate...@cs.su.oz.au
Kenneth I. Laws (l...@ai.sri.com) writes:
[mod- extensive repetition of article with general agreement
statements deleted by moderator]
>The arguments I've heard invoking the unpatentability of algorithms
>(and hence of software, although I think that's too big a jump) have
>never mentioned the legal or philosophical basis of this exclusion.
>Can anyone fill me in? I think it would help to clear up how this
>exclusion does or does not apply to software.
Proposition: any given stored-program computer with a stored program, and
having input and output ports, is for all purposes that matter (intellectually,
legally, patentwise, Turing-test-wise) indistinguishable from a random-logic
device that has the same inputs and outputs, and is indeed indistinguishable
from a device with gears and shafts and cogs having the same inputs and
outputs.
If we accept the proposition, then software patents (which I define as patents
covering physical systems that happen to have lots of software in them) have
as much right to exist as patents on mousetraps.
Carl Oppedahl
30 Rockefeller Plaza
New York, NY 10112-0228
Path: sparky!uunet!think.com!mips!mips!munnari.oz.au!metro!cluster!
rockw...@socrates.umd.edu
From: rockw...@socrates.umd.edu (Raul Deluth Miller-Rockwell)
Newsgroups: comp.patents
Subject: Re: Patents and their form of expression
Message-ID: <4285@cluster.cs.su.oz.au>
Date: 29 Feb 92 15:06:02 GMT
References: <4265@cluster.cs.su.oz.au>
Sender: n...@cluster.cs.su.oz.au
Lines: 43
Approved: pate...@cs.su.oz.au
Carl Oppedahl:
Proposition: any given stored-program computer with a stored
program, and having input and output ports, is for all purposes
that matter (intellectually, legally, patentwise, Turing-test-wise)
indistinguishable from a random-logic device that has the same
inputs and outputs, and is indeed indistinguishable from a device
with gears and shafts and cogs having the same inputs and outputs.
If we're going to take a black box approach let's add that it's also
indistinguishable from a device with a human operator, where the human
is making the decisions. Incidentally, note that this thesis is the
converse of artificial intelligence.
If we accept the proposition, then software patents (which I define
as patents covering physical systems that happen to have lots of
software in them) have as much right to exist as patents on
mousetraps.
If we accept the extended proposition, then software patents (defined
similarly) have as much right to exist as patents on mice having been
caught.
If you're going to treat "physical systems with lots of software" as a
patentable entity, and the patent would not hold with a human operator
providing services analogous to those which the software provides,
then the patent should also fail when functionally equivalent (but
stylistically different) software is substituted.
In accordance with Carl Oppedahl's earlier post, suggesting that we
post the proposed statue in one's posting:
... mathematical equivalence, symbolic equivalence, or equivalence of
result without equivalence of the step by step procedures, can not be
grounds for patent violation. If a patent involves a device which is
turing equivalent for some class of problems then, for the purpose of
determining the validity of the patent, the patent must also be
considered to cover, as identically equivalent, a person serving in
place of that device.
--
Raul Deluth Miller-Rockwell <rockw...@socrates.umd.edu>
The U.S. federal government went another billion dollars into debt today.
Path: sparky!uunet!think.com!mips!mips!munnari.oz.au!metro!cluster!
egg...@twinsun.com
From: egg...@twinsun.com (Paul Eggert)
Newsgroups: comp.patents
Subject: legal and philosophical basis for excluding software from patents
Message-ID: <4287@cluster.cs.su.oz.au>
Date: 3 Mar 92 09:01:49 GMT
References: <4169@cluster.cs.su.oz.au> <4239@cluster.cs.su.oz.au>
Sender: n...@cluster.cs.su.oz.au
Organization: Twin Sun, Inc
Lines: 38
Approved: pate...@cs.su.oz.au
Nntp-Posting-Host: bi
Apparently-To: comp-pate...@uunet.uu.net
l...@ai.sri.com (Kenneth I. Laws) writes:
>The arguments I've heard invoking the unpatentability of algorithms
>have never mentioned the legal or philosophical basis of this exclusion.
A legal basis for this exclusion is that patents are not granted on
all ``useful tasks''. If they were, one could obtain patents on all
sorts of useful tasks that are in fact unpatentable, e.g. new ways of
legally sidestepping campaign contribution laws. Instead, patents are
limited to the domain of physical processes, i.e. processes whose
principal goal is changing material state. (Patents are also granted
in a few other domains which are not relevant here.) Patents are not
granted on pencil-and-paper methods, because these are not considered
to be physical processes. I suppose one might argue that pencil marks
on paper are ``material state'', but that's not the intent of the
patent laws, and that's not the way the courts have ruled. Legally
speaking, the presence of software is largely irrelevant to the
question of whether a process is patentable. A physical process like
curing rubber may be patentable, but a non-physical process like
solving linear programming problems is not patentable, and the presence
of software should not magically change this. I suppose one could
argue that electrical signals inside a computer are ``material state'',
but that argument is no more valid for the internal signals inside a
computer than it is for pencil marks.
A philosophical basis for excluding software from patents is that
patents are a blunt instrument and should not be used carelessly.
There is no completely just and practical way to compensate
intellectual labor. The patent system is no exception: it is harmful
in several ways, e.g. it unjustly prohibits independent reinvention and
reuse, and it diverts significant resources to patent lawyers and their
ilk. We should put up with these harms only in fields of endeavor
where the benefits of the patent system outweigh the harms, and where
there is no better way to compensate intellectual workers. But for
software, the benefits of the patent system are far smaller than the
harms, and there is already a proven way to compensate intellectual
labor, namely the copyright system.
Path: sparky!uunet!think.com!mips!mips!munnari.oz.au!metro!cluster!
kas...@cunixa.cc.columbia.edu
From: kas...@cunixa.cc.columbia.edu (John Kasdan)
Newsgroups: comp.patents
Subject: Re: Patents and their form of expression
Message-ID: <4292@cluster.cs.su.oz.au>
Date: 3 Mar 92 09:14:02 GMT
References: <4169@cluster.cs.su.oz.au> <4239@cluster.cs.su.oz.au>
<kqo2jfINNorf@early-bird.think.com>
Sender: n...@cluster.cs.su.oz.au
Organization: Columbia University School of Law
Lines: 29
Approved: pate...@cs.su.oz.au
Nntp-Posting-Host: cunixa.cc.columbia.edu
In article <kqo2jfINN...@early-bird.think.com> n...@Think.COM writes:
>An example of an algorithm, which wouldn't be patentable, is long division.
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>It's purely a mental process, although physical devices (paper&pencil, an
>abacus, a computer, etc.) might be used to assist the person (mainly
>substituting for his memory or slowness).
>
I know that the CAFC continues to give lip service to the holding in
Benson that an algorithm cannot be patented (although they have
restricted that to "mathematical algorithms" simply to get around
holding that processes are not patentable.) However the "natural
order recalculation" patent which has been discussed recently is just
a bad (O(n*(n+m))) algorithm for doing topological sort. And that
patent was actually litigated before the CAFC predecessor court, the
CCPA.
If you look at some of the patents that are being granted, things get
even worse. E.g. the Kamarkar patent which, in its first claim,
claims all methods of solving the LP problem by computing a non-linear
path through the feasible set. Now if that isn't an algorithm patent,
I'ld like to know what is.
Finally, the "mental process" doctrine has, I believe, been disavowed
by the CAFC; the only court whose opinion on the matter is of any
importance.
/JK (.sig closed for repairs)
Path: sparky!uunet!munnari.oz.au!metro!cluster!0001811...@mcimail.com
From: 0001811...@mcimail.com (Carl Oppedahl)
Newsgroups: comp.patents
Subject: Patents and their form of expression
Message-ID: <4320@cluster.cs.su.oz.au>
Date: 4 Mar 92 02:57:00 GMT
Sender: n...@cluster.cs.su.oz.au
Lines: 29
Approved: pate...@cs.su.oz.au
Raul Deluth Miller-Rockwell (rockw...@socrates.umd.edu) gives a proposed
change to the patent statute:
> ... mathematical equivalence, symbolic equivalence, or equivalence of
> result without equivalence of the step by step procedures, can not be
> grounds for patent violation.
The patent law, the way it is now, largely does just what Mr. Rockwell
suggests. One is only an infringer if one satisfies _each_ of the steps of
the method (in the case of method claims). An accused infringer who only
brings about the same result without having performed each of the steps in the
claim (or their equivalents) will not be an infringer.
> If a patent involves a device which is
> turing equivalent for some class of problems then, for the purpose of
> determining the validity of the patent, the patent must also be
> considered to cover, as identically equivalent, a person serving in
> place of that device.
Here, too, the patent law already does what Mr. Rockwell suggests. Suppose
the claim being discussed recites a method of performing the steps of doing
A, doing B, doing C, doing D, and doing E. The way the law is right now, a
human doing those five steps will matter. If a human doing those five steps
was described in a printed publication more than a year before the filing
date then the claim is invalid. And so on. The human having done those
steps in the past counts as prior art just as much as a machine having
done those steps in the past.
Path: sparky!uunet!munnari.oz.au!metro!cluster!brns...@KRAMDEN.ACF.NYU.EDU
From: brns...@KRAMDEN.ACF.NYU.EDU (Dan Bernstein)
Newsgroups: comp.patents
Subject: Re: Patents and their form of expression
Message-ID: <4338@cluster.cs.su.oz.au>
Date: 4 Mar 92 13:35:07 GMT
References: <4169@cluster.cs.su.oz.au> <4239@cluster.cs.su.oz.au>
<kqo2jfINNorf@early-bird.think.com>
Sender: n...@cluster.cs.su.oz.au
Organization: IR
Lines: 53
Approved: pate...@cs.su.oz.au
> Finally, the "mental process" doctrine has, I believe, been disavowed
> by the CAFC; the only court whose opinion on the matter is of any
> importance.
I'd like to emphasize that the main anti-software-patent position is
that software _should not_ be patentable. Whether software _is_
patentable is a secondary question. So what the CAFC may or may not have
done shouldn't deter us from rational discussion of what it _should_ do.
At the risk of repetition (since someone asked again for a precise
definition of ``mental process''), I've attached another copy of my
proposed regulations, which in my opinion _should be_ part of the law.
Nobody's claiming that these _are_ part of the law.
---Dan
Statement of Proposed Mental-Process Patent Regulations
Daniel J. Bernstein
July 5, 1991
I support the adoption of the following definitions and resolutions
into statute or regulation.
A mental process per se is not statutory subject matter for a
patent. The term ``mental process'' includes, but is not limited
to, any process which may be carried out within a person's mind.
The term ``mental process'' includes, but is not limited to, any
method by which a set of numbers or symbols is computed from a
different set of numbers or symbols. The term ``mental process''
includes, but is not limited to, a mathematical algorithm. The
term ``mental process'' includes, but is not limited to, a
mental process performed with the aid of a computer. The term
``mental process'' includes, but is not limited to, any
combination of mental processes.
A mental process applied in a non-essential manner to physical
elements is not statutory subject matter for a patent. A mental
process limited to a particular technological environment is not
statutory subject matter for a patent. A mental process combined
with data-gathering steps which merely determine values for
variables used in the process is not statutory subject matter
for a patent. The combination of a mental process with a process
or product which is not statutory subject matter for a patent
is not statutory subject matter for a patent. A product defined
by its result or function, when that result or function is also
the result or function of a mental process, is not statutory
subject matter for a patent.
Use of a mental process per se cannot infringe upon a patent.
Use of a mental process with the aid of a computer or partially
or entirely carried out upon a computer cannot infringe upon a
patent.