Path: sparky!uunet!munnari.oz.au!metro!cluster!...@gnu.ai.mit.edu
From: r...@gnu.ai.mit.edu (Richard Stallman)
Newsgroups: comp.patents
Subject: [ARTICLE] Reply to "Why Patents are Bad for Software"
Message-ID: <4949@cluster.cs.su.oz.au>
Date: 19 May 92 17:32:43 GMT
Sender: n...@cluster.cs.su.oz.au
Lines: 137
Approved: pate...@cs.su.oz.au


This letter was sent to Issues in Science and Technology
opposing the article, "Why Patents are Bad for Software".


We disagree with the conclusions expressed by Simson L. Garfinkel,
Richard M. Stallman, and Mitchell Kapor in "Why Patents Are Bad for
Software" (_Issues,_ Fall 1991), particularly their statement that the
patent law should me amended immediately "to disallow software
patents."  The patent system has served the United States well over
the past two hundred years by adapting to new areas of technology as
they emerge, and today it provides vital support to U.S.
competitiveness around the world.  Improvements to the system are
certainly possible, but the beneficial effects of any major changes
need to be clearly proven before they are made.  The authors do not
meet this burden of proof.

First, they demonstrate that they do not understand the current law.
Their analysis of the Supreme Court decisions on this issue is
inaccurate and superficial, and tehy omit any mention of the numerous
decisions by lower apellate tribunals.  They appear to allege that
originality is a requirement for patentability and that mathematical
laws are patentable subject matter.  They are wrong on both counts.
Their discussion of the concept of "public domain" is confused at
best.  In our view, recommendations for amending current law that are
based on misconceptions such as these should not be given any
credence.

Second, the authors tend to make sweeping statements that reflect
their point of view.  Most of these statements, however, do not appear
to be the result of a balanced and reasoned inquiry and do not appear
to be supported by facts.  For example, the authors allege that
granting patents on computer-related inventions will harm small and
even mid-sized companies and that these businesses cannot use the
patent system.  Yet, there is no quantitative evidence to suggest that
such companies are not applying for, receiving, or enforcing patents.
Indeed, we expect the contrary.  Also, the authors forget that the
patent system has traditionally been a major factor in the growth of
many small companies into large businesses, regardless of the area of
technology.

Third, the authors make several inconsistent statements.  They assert,
for instance, that the patent system is "backfiring in the computer
industry" because, among other reasons, the search for prior art in
the realm of computer software is all but impossible to conduct.  Yet
they later state that one of the traditional roles of the patent
system--encouraging the dissemination of information--is not necessary
for computer software because computer scientists routinely publish
their discoveries.  If such publications are available, they may be
evidence of prior art and should be available to patent examiners for
consideration.  In fact, the authors' concern that the prior art is
too enormous for patent examiners to search is groundless, as the
modern patent examiner has a plethora of resources to assist such a
search, and every patent applicant has a positive duty to disclose any
prior art known to be relevant to the claimed invention.

Finally, the authors report in detail the views of thsoe who believe
they have been economically disadvantaged by the operation of the
patent system.  On the other hand, they cavalierly dismiss the views
of those who appear to have used the patent system successfully and
impugn their motives for acquiring patent protection or enforcing
their patent rights.  Furthermore, the authors' "recommended reading"
list cites only articles that support their own point of view.

We welcome constructive suggestions for improving the U.S. patent system.  
We believe, however, that constructive suggestions must be based on an 
informed and objective viewpoint.

HARRY F. MANBECK, Jr.
Assistant Secretary and Commissioner of Patents and Trademarks
Patent and Trademark Office
U.S. Department of Commerce
Washington, D.C.

^M

Garfinkel, Kapor and Stallman reply:

We are software developers, not lawyers.  While Commissioner Manbeck
surely understands better than we how the legal system analyzes its
actions, we are better placed to understand their effects on software
developers.

What does it mean to ask whether algorithms are patented?  From a
legal point of view, the question is whether this patent or that is
considered to be "a patent on an algorithm".  The patent system does
not consider any patent to be "on an algorithm" since the Supreme
Court has ruled there may not be such patents.

But the question that matters to a programmer is whether patents
prohibit the use of a certain algorithm.  As a matter of practical
fact, they often do.  If you ask an expression on data compression
what patent 4,558,302 or 4,814,746 covers, the response will be, "It
covers the LZW compression algorithm."  (Both of these patents
cover *the same* algorithm.  Yes, this is not supposed to happen,
but it does.)

A claim in a patent prohibits any system with a certain combination
of elements.  

Commissioner Manbeck mentions a seeming contradiction in our article
which does call for an explanation.  We stated that there was no
shortage of publication in the software field.  We also stated that it
is hard to find prior art for many techniques now patented because no
one bothered to publish them.  But this is not a contradiction--both
statements are true, but in different domains.

Software developers have always been eager to publish papers
describing impressive achievements.  These papers describe new
algorithms and techniques that programmers think bring someone credit.
The peer review system makes sure of this.

But what of techniques that did not meet this standard?  Minor
clevernesses?  They could not have been published in a serious
journal, so we told them to our friends to win a chuckle.  They were
not important enough to publish--but some of them are now patented.

For example, in the mid-1970's it took just one instruction on the
PDP-10 at MIT to perform the exclusive-or operation to modify the
screen contents.  It would have been ludicrous to submit a journal
article about such a simple thing.  And dishonest as well, since the
MIT programmers did not believe they were the first to use this
technique.  Far from considering this their secret, they thought it
was generally known.  It is uncertain whether their work results in
legally valid prior art for patent 4,197,590.

Repeated experience shows that this example alone is enough to
convince an audience of programmers that the patent system is off
course by miles, not just inches.  Yet advocates of software patents
continue to defend it as legitimate.

We do agree with one thing said by Commissioner Manbeck: a major
change in the patent system should only follow from a clearly
demonstrated need.  It is unfortunate that this principle was ignored
when patents were imposed on our field; only a small number of lawyers
and a smaller number of judges participated in the decision, and the
need for the change was never demonstrated to the satisfaction of
programmers.  Since the requisite scrutiny was omitted then, better
late than never: this change should undergo be scritinized properly
now before society accepts it as permanent.