From comp.org.usenix Tue Jul 16 12:24:15 1991
Xref: news-server.csri.toronto.edu gnu.misc.discuss:3645 alt.activism:
15396 misc.activism.progressive:222 misc.legal:28136 comp.org.usenix:2367
Path: news-server.csri.toronto.edu!rpi!uwm.edu!bionet!snorkelwacker.mit.edu!
ai-lab!ai.mit.edu!gnulists
From: rms@gnu.ai.mit.edu (Richard Stallman)
Newsgroups: gnu.misc.discuss,alt.activism,misc.activism.progressive,misc.legal,
misc.legal.software,comp.org.usenix
Subject: Easy way to reply to PTO request for comments
Message-ID: < gnusenet9107131615.AA14845@mole.gnu.ai.mit.edu>
Date: 13 Jul 91 16:15:59 GMT
Followup-To: gnu.misc.discuss
Lines: 1217
Approved: info-gnu@prep.ai.mit.edu
To: info-gnu@prep.ai.mit.edu, info-gnu-emacs@prep.ai.mit.edu
To: info-gcc@prep.ai.mit.edu, info-g++@prep.ai.mit.edu

[This is being broadcast by the FSF to its users as part of its duties
as a member of the the League for Programming Freedom.]

Here is the letter that the LPF is sending to the patent office in
response to its request for comments.  It will be touched up, but
there won't be substantial changes.  It is formatted with LaTeX, but
you ought to be able to read it without formatting it.

If you have written your own letter, that's good.  If you haven't
written one, and you don't have time to write a letter in your own
words, then perhaps you agree with this letter.  If so, please send a
letter saying simply that you agree with the letter being sent by the
League for Programming Freedom.  That will do some good, and is easy
to do.

If you don't agree with everything in this letter, then you can
identify the particular sections you agree with, or those you don't
agree with.  You can write whatever you want--referencing this letter
is just a way of saving time.

Note that this letter omits a few questions that the LPF didn't write
answers for.

Your letter should reach the patent office on Monday, July 15.  (I
believe they will also accept express mail postmarked that day.)  You
can fax your letter to the patent office using (703) 308-3718.  This
number works 24 hours a day.  If there is any trouble, you can phone
the patent office during business hours at (703) 308-1353.

If you can't send the letter Monday, you might as well do it anyway as
soon as you can.  Better late than never.

Please also send copies of your letter to the House and Senate
subcommittees on intellectual property.  There is no deadline for
this, and you can use ordinary snail.

    House Subcommittee on Intellectual Property
    2137 Rayburn Bldg
    Washington, DC 20515

    Senate Subcommittee on Patents, Trademarks and Copyrights
    United States Senate
    Washington, DC 20510

This is so easy that even the busiest programmers have time to do it.


\documentstyle[11pt]{article}
\addtolength{\topmargin}{-.75in}	% repairing LaTeX's huge margins...
\addtolength{\textwidth}{1.6in}		% and here...
\setlength{\textheight}{8.75in}		% more margin hacking
\addtolength{\oddsidemargin}{-0.75in}
\addtolength{\evensidemargin}{-0.75in}
\setlength{\parskip}{.1in}

\begin{document}

\begin{center}
{\Large\bf Response by the League for Programming Freedom}

{\Large\bf to the request for public comment}

{\Large\bf of the United States Patent and Trademark Office}
\end{center}

\vspace*{.5in}
\noindent{\bf I\@. Protection of Computer Related Inventions}

\begin{quote}
{\em (a) What problems, if any, exist in the current framework of laws
which protect computer-related inventions?}
\end{quote}

The existence of patents in the field of software is primarily a
problem.  In particular, the fact that techniques which may be used in
a program can be patented, and that features provided by programs can
be patented, introduces a large element of risk and considerable extra
difficulty and cost into the activity of software development.  It
generally benefits large companies (see appendix regarding IBM) and
harms the small companies which have done the most to advance the
state of software.

Here are the ways patents create risk for software developers:

\begin{itemize}
\item
It is hard to find the patents which might apply to a program being
developed.

\item
It is hard to determine whether a particular patent applies to a
particular program.

\item
It is hard to know whether a patent is valid.  Finding prior art is
difficult and whether the crucial piece of prior art is found is a
matter of luck.

\item
Whether a patent is actually valid often depends on chance occurrences
in the past.  Many patented techniques were used frequently in many
projects, but which of these projects happened to publish the
technique in question, and when, is largely a matter of luck.

\item
The only way to find certainty is through a lawsuit, which is
extremely costly in both money and time.  In addition, as lawyers say,
the outcome of any particular lawsuit involves a substantial element
of chance.
\end{itemize}

Patents create difficulties for software developers because they must
avoid using patented techniques or else negotiate and pay for licenses
for them.

If developers hope to profit from patents, they encounter an even
greater level of risk, because very few patents are significantly
profitable.  This is one of the reasons which led ``The Economist''
(15 Sep 1990) to call the patent system, ``A time-consuming lottery.''

Patents in software have little to do with contribution to advancing
the field of software.  This is because patents generally cover
building blocks, while the important contributions to the field
consist of software systems.  Whether a software system is a good one
has little to do with whether it uses any particular new technique of
great importance.  The quality of the software system is a matter of
the details of how numerous techniques are combined.

\begin{quote}
{\em (b) What changes, if any, should be made in the domestic and
international systems for protection of computer-related inventions?}
\end{quote}

The use of computer-related techniques by software developers should
be protected from encroachment by patents.

Doing this effectively requires that software be exempt from the scope
of patents.  That is to say, the development, distribution and/or use
of a computer program should never be held to infringe any patent.  A
combined hardware/software system should infringe a patent only if the
hardware alone does so, or nearly.  If the hardware is a
general-purpose computer, then the combination should infringe only
patents covering the construction of such computers---no matter what
the software does.

The natural idea for how to eliminate the harmful effects of patents
on software development would be to rule out patents ``on software''.
However, this won't work, because it is not possible to classify
patents into software patents and hardware patents based on how they
are written.  For example, a patent written to describe an electrical
circuit can also prohibit the use of a computer program.

Any approach for preventing patents from hampering software
development must, to be effective, be immune to circumvention by
writing patents differently.  Otherwise, it would merely cause patents
to be formulated differently (such as, in terms of hardware), without
really giving relief to the software developers.

Another reason why changing the criteria for issuing patents won't
solve the problem is that it would not provide any relief from the
thousands of patents already issued that restrict software
development.  The existing patents cause such serious problems that
something must be done about them.  In the past, the US government has
acted to clear away patents when impediments to progress were
considered unacceptable.  For example, this was done during World Wars
I and II\@.  Whether action is taken now will depend on whether a
blockage in the software field is acceptable now.

\begin{quote}
{\em (c) The Supreme Court has found that new and useful computer
program-related inventions are eligible for patent protection. 
What rationale, if any, exists in law or policy for Congress to now
remove patent protection for this field of science and technology?}
\end{quote}

\noindent
[This misstates what the Supreme Court did in Diamond vs.\ Diehr.  As
explained by Professor Samuelson in the ``Emory Law Journal'', Fall
1990, the case involved a method of curing rubber which made use of a
computer program to determine what the machinery should do.  The
ruling was that this was a patentable process.  This outcome is
consistent with continuing the old ``subject matter'' doctrine, under
which a patentable process had to be a process for transforming
matter.  The Court said nothing to indicate that software-related
inventions outside this narrow area can be patented.]

The reason to protect software developers from patents is that patents
hamper rather than promote progress in the field of software.  This is
because in the field of software, there is little room for patents to
do good, and much room for them to do harm.

In general, patents are supposed to benefit the public by encouraging
more inventions to be made and published.  The public pays a price for
this when product development is obstructed because certain techniques
cannot be used, and when license fees make projects economically
infeasible.  The uncertainty about which techniques are ruled out by
valid patents is also an important burden imposed by the patent
system.

Although these forms of harm and good exist in principle no matter
what the field, their magnitude varies greatly from field to field.
In the software field, because of the nature of software, the harm is
large and the benefit is small; for this reason, patents are
undesirable in software.

Let us first consider how much good patents can do in the field of
software, and then consider how much harm they do.

Some patents cover techniques that can be used in programs.  A glance
at the journals of computer science will show that there was a
profusion of new techniques being developed both ten years ago, when
they could not be patented at all, and five years ago, when few
workers in the field thought to patent them.

Other patents cover features that programs can offer.  A passing
acquaintance with history a few years ago will show a profusion of new
features being tried out in software products.  Yet at the time,
hardly any of these was patented.

Since patents can do good only by adding to this profusion, the good
they can do is small at best.  Progress in software is not being held
back by the aspects that patents can promote.

The introduction of patents has not substantially reduced trade
secrecy in the computer field.  Formerly, developers published the
general ideas that they felt would bring them credit, while keeping
the detailed code for a program a trade secret.  Today, they may
patent the general ideas, but they still keep the detailed code a
trade secret.

The benefit of discouraging trade secrecy is nonexistent when it comes
to patented features.  Every feature of a computer program is
published automatically by virtue of showing the program to users.
Even though a user may not know how a program works inside, he or she
cannot help seeing what features the program provides.

What about the harm that patents can do?  In the field of software,
the harm patents do is especially great.  This is because each
programmer can infringe far more patents in a day's work than a
hardware designer can.  The reason for this is that software design is
basically easier than designing mechanical systems, chemicals, or
electrical circuits---for a given size of system.

Why is software easier to design than hardware?  Other kinds of
engineers must deal with recalcitrant material objects which do what
they do, notwithstanding any mathematical models.  Hardware components
draw power; they dissipate heat; they vibrate or produce electrical
noise; they distort the outputs they are attached to; they leak; they
wear out; they break.  They require a means of access to replace them
when they break.  And they have to be assembled in each copy of the
hardware device, which can require a large factory costing as much as
hundreds of millions of dollars.  Software designers don't have these
problems of design, and copies of the program can be reproduced on a
workstation costing \$5,000, no matter how big it is.

The claim that software is easy to design might seem outrageous.
After all, there is much concern with the great difficulty of
designing software.  But there is no real contradiction here.
The reports that software design is difficult refer to typical
software systems, which have far more parts than typical hardware
systems.  This is because engineers push all capabilities to the
limit.  If software is a hundred times easier to design, the software
designer can tackle a project a hundred times as complicated.

As a result, one good programmer can write a program in a few months
which uses hundreds of techniques, and provides hundreds of features.
Such a system can easily infringe many dozens of patents.

A few good programmers working for a few years can design a system
with millions of components (perhaps a million lines of code).  Such a
project is hard work, but nothing compared with a hardware system of
the same size.  The hardware system would require a team of hundreds,
if not thousands, of designers.  (A hardware system composed of many
identical subunits is not so hard to design, but that is not a good
comparison for software: the components in a large program are not
normally repetitious.)

How does this relate to the effect of patents?  The work in software
design consists of putting together large numbers of different
techniques to form a large system for a specific purpose.  Patents do
not encourage this; instead, they make each design decision a legal
risk.  Each design decision might happen to combine the basic
operations of the computer in a way which is patented, and thus
subject the developer to future demands and threats.

Another way of putting this is that the barriers to entry in the
software field are very low.  Or, that is, they were very low in the
past.  Patents threaten to make them much higher.  Recently, for example,
one software company was started with 10 programmers and 4 lawyers.  In
the past, there would have been no need for these lawyers.

The basic assumption behind the patent system is that inventions are
rare and precious.  Perhaps in some fields this is true, but not in
software.  Because the barriers to entry are so low, and the number of
software developers is so great, new techniques are likely to be
reinvented more frequently than in other fields.  Society only loses
by issuing a patent on a technique likely to be reinvented within the
next few years.

The extent to which software techniques are likely to be reinvented
can be obscured in a systematic fashion that is due to the constant
improvement in the power of computers.  This has to do with the fact
that computers get steadily larger and faster.  As a result, what
is not obvious today may become obvious in a few years.  By this we
mean that, a few years from now, the same technique would be subject
to frequent reinvention, whereas today that is not so.

A computer of a given power may today be a rare, specialized item that
only highly funded research groups (of large companies) can afford.
Six years from now, a computer of the same power will be common.  The
natural lines of investigation using these powerful computers may lead
to techniques which are not applicable to less powerful computers.
Today, these techniques could only be found by the largest companies,
which implies they will not be rediscovered frequently.  Six years
from now, many people will be in a position to follow the natural
lines of investigation for equally powerful computers.  The same
techniques which today seem like rarefied research will then be
reinvented each month.  They will be effectively obvious.

Today, with patents, the large companies patent these techniques.
The techniques will still be patented when they become obvious, and
for many years thereafter.  (Considering the years of delay before a
typical software patent is issued, the patent could easily last 15
years after the time when the techniques become obvious.)

In essense, these companies have a chance to explore virgin territory
and patent everything they find there---even hills and rivers which
would inevitably become public knowledge once the area is settled.

If patents are not available, then the large companies will probably
publish their discoveries anyway, since that is what they customarily
did in the 1970's when software patents were discouraged.  But even if
they do not, the techniques will become available due to reinvention
when they are needed.  Society needs the techniques to be known when
they become useful, which is when computers powerful enough to support
them become common enough to make a large software market.  This is
precisely when they will be rediscovered frequently.

\begin{quote}
{\em (d) What evidence exists, if any, that patents issued on new and
useful computer program-related inventions do or do not provide an
incentive to conduct research and development on new products, and
that such patents do or do not promote the development of new
technology?}
\end{quote}

It is necessary to distinguish between development of products and
development of technology.

It is possible that some additional new technology is developed as a
result of the existence of patents.  However, there is good evidence
that a large amount of new technology in this field would be developed
and published in the absence of patents.  This evidence consists of
the large number of papers published in computer science before the
era when most of these authors thought of getting patents.  Further
evidence is found in the large number of new features of computer
programs that were developed.

What this shows is that patents probably do not increase the
development of new technology by a very large ratio.

The main effect of patents on new technology in our field is to make
it less useful---because today it will be patented, whereas ten years
ago it would have been published without a patent.

When it comes to products, patents are an impediment to development.
This is because developing a product means putting together a large
number of techniques, each of which might be patented by someone else,
and therefore exposes the developer to threats and demands to pay, or
even to shut down.

\begin{quote}
{\em (e) What conflict or overlap is created by the existence of a patent
for a new and useful computer program-related invention and either (1)
a copyright for original expression embodied in the computer program
fixed therein, or (2) a mask work in the invention; and if any exists,
is it harmful or helpful?}
\end{quote}

Conflict, overlap and compounding can multiply the problems of the
industry.  It is possible to have a mask work monopoly, successive
copyrights, design patents, and a utility patents all on the same
apparatus.  These various monopolies will begin at different times,
so that one might start when another ends.  The end result
is a complicated patchwork of prolonged monopoly.

\begin{quote}
{\em (f) Should Congress legislate the boundary for patents in the
computer program area, or is it preferable to permit the courts to
continue to define the outer boundary?}
\end{quote}

There is reason to believe the Supreme Court would likely abolish much
of the problem, if it gets the chance; but we cannot be certain of
this, and there is no telling when the occasion would arise.

However, if Congress passes the appropriate legislation, it can be
sure of solving the problem.

Thus, Congress should legislate.

Legislation is a vital input to the judicial process; if the law is
such that judges arrive at interpretations which don't serve the
public interest, changing the law is the usual solution.

\begin{quote}
{\em (g) Are the tests of patentability for computer program-related
inventions implemented by the USPTO in the notice published in the
Official Gazette on August 9, 1989, consistent with the patent statute
and/or court decisions?  If there are inconsistencies, what are they?}
\end{quote}

The wording of this question implies that the PTO has been doing what
it was supposed to do---setting forth guidance and tests for Examiners
to use in judging applications for software-related patents.

The Patent Office didn't even remember where the notice appeared (the
September 5, 1989 issue).  A more substantial problem is that the
notice actually deals only with Sections 101 and 102, and ignores more
important and difficult problems with respect to Sections 103 and 112.

It appears that the notice is the only guidance the PTO offers to
applicants and examiners concerning tests for patentability for
computer program-related inventions.  There is nothing to be found in
the Manual for Patent Examining Procedure to supplement it.

Not being a mere abstraction or algorithm (Section 101) is but one of
four requirements for patentability.  The invention must also be novel
to comply with Section 102, unobvious to comply with Section 103, and
the applicant must also enable the part of the public which has
ordinary skill in the art to which the invention pertains to make and
use the invention in order to comply with Section 112 of the act.

The notice addresses Section 102 by saying that if the Examiner's
computer system finds a past anticipation of the invention, the
invention is not patentable.  Otherwise, he'll assume that it is
patentable.  This scheme cannot possibly work, because the data base
is woefully incomplete, and because you cannot find examples of even
the identical technique through a straightforward computer search of a
data base (see (h) 4 below).

The notice totally fails to tell the Examiner how to determine what is
obvious in view of related, but not identical art.  This is something
that calls for human intelligence.

The notice also does not tell us what is enough to comply with Section
112.  What if the ``preferred embodiment'', the only embodiment for
which there has been a reduction to practice, involves a computer
program of one million lines of code.  In days gone by the program
could be provided on microfiche as an appendix to the application.

But this issue may be of little importance, because the idea that a
patent ``teaches'' a software technique is basically ludicrous.
Reading patents is very difficult; programmers don't know how.  It's
not useful for them to learn how.  When a programmer looks for
techniques to study, he wants to find something applicable to
designing a particular system.  It's no use looking for this in
patents, because the techniques found there are patented, and thus not
safe to use.

It is said that the examiners on these cases have only passing
familiarity with the art and have neither the time, nor the ability,
nor the inclination to give to the applicant a reasonable presumption
of validity as to Sections 103 and 112.

It is remarkable that this notice does not consider a single
decision of the U.S. Court of Appeals for the Federal Circuit.  It is
obvious that the notice should be brought up to date.

\begin{quote}
{\em (h) What concrete steps should be taken, if necessary, to revise the
PTO examination procedure for computer program-related inventions in
order to achieve high-quality issued patents, particularly with
respect to: [a list of possible ways of improving the functioning of
the patent system for software]?}
\end{quote}

Any of these changes could only be an improvement, because they would
reduce the number of patents which might subsequently hamper software
development.  But even taken together, they will not prevent the
existence of a large number of software patents, and those patents
will still cause serious problems.

Even if these measures cut the number of patents in half, that will
not change the situation very much.  A large computer system is likely
in the future to infringe many dozens of patents.  Half of that will
still be a large problem.

You can think of this as the equivalent of nuclear overkill.  How much
good does it do to block half (on the average) of the warheads in a
nuclear attack?  If the attack consists of a single warhead, this
could prevent all damage---a big improvement, though only 50\% probable.
If the attack consists of twenty warheads, blocking half could do a
great deal of good, even though the problem caused by the remaining
warheads would still be large.  If the attack consists of 5000
warheads, knocking out half of them at random will do no good
whatever.

\begin{quote}
{\em  (h) 2. Providing patent examiners with training in this field to
raise/maintain their expertise;}
\end{quote}

Computer Science should be recognized in the same light as
mathematics, ``the Queen of Sciences''.  Both contribute importantly
to the understanding of inventions in all fields of technology.

Realizing that innovations in programming are properly nonstatutory,
the ``obviousness'' of the program becomes immaterial.  The Examiner
does not need to be skilled in that art; but he needs to be ``computer
literate''.  He still needs to apply his judgment to the hardware
class of invention before him.

\begin{quote}
{\em  (h) 4. Providing an optimum system of classification to maximize the
searchability of inventions in this field;}
\end{quote}

This is more easily said than done.  No one knows of a system that
will do the job, or if a system exists.

Patents are usually classified by the end results they are used for.
This will not work for software.  For example, the patent (owned by
Stanford University) on a faster algorithm for the discrete Fourier
transform may be used in programs for signal processing, and in
programs for multiplying large numbers.  And perhaps in other sorts of
programs as well.

A simple keyword-based searching system does not even begin to work.
For example, Lotus is now being sued on account of a patent 4,398,249
which covers the technique of ``natural order recalculation'' used in
spreadsheets.  The patent does not mention the terms ``natural order
recalculation'' or ``spreadsheet''.  There is no way that developers
writing a spreadsheet could have recognized its possible applicability
without studying the patent in detail.  Other related applications
also covered by the patent include ``constraint propagation'',
``antecedent reasoning'' and ``topological sort''---but these terms do
not appear in the patent either.

In other words, searching cannot be easy because actual thought (using
both a programmer's education and a lawyer's education) is needed to
determine that any given patent is inapplicable to a planned project.

\begin{quote}
{\em (i) What procedures, not currently available, should be considered
to correct problems caused by improperly granted patents (i.e., post
grant oppositions, court nullification), and how would these
procedures particularly relate to computer program-related inventions?}
\end{quote}

Post grant oppositions would help by invalidating some patents.  Like
other things that would prevent or invalidate a fraction of all
software patents, this is not enough to solve the problem.

\vspace*{.08in}
\noindent{\bf II\@. Federal Protection for Trade Secrets}

\begin{quote}
{\em Trade Secret statutes now exist only at the state level, and are not
uniform in the states wherein they exist.

  The principal issue to be examined is whether Federal Statutory
protection should be enacted and if so, how it should relate to state
statutes?

  (a) Is there a need for a Federal law on trade secrets?  What
problems have been caused by the absence of such a law?  What problems
would such a law create?}
\end{quote}

Trade secrecy is an antisocial activity.  The founding fathers
considered trade secrecy so undesirable that they set up the patent
system to discourage it, paying a price in the general freedom.

Indeed, there was a line of decisions that held that an inventor had
the option of trade secret or patent, that he could not keep the
process secret until the secret got out, then apply for a patent.
This makes sense given the idea that granting a patent means society
is paying a price, and society should get something in return.

Perhaps a better way to discourage the practice of trade secrecy would
be to stop government assistance for it---to eliminate the trade secret
laws that help people force others to keep secrets for them.

Perhaps if we did not have trade secret laws to make the keeping of
secrets more reliable, we would have less need for patent laws to
encourage disclosure.

\vspace*{.08in}
\noindent{\bf III\@. Cost and Complexity of Patent Enforcement}

\begin{quote}
{\em   Although many patents are enforced through negotiation, resort is
frequently made to either litigation in the Federal district courts or
to arbitration or one of several other forms of alternative dispute
resolution (ADR).

  Patent litigation is said to be complex, expensive, unpredictable
and heavily dependent on the ability of the judge to exert fair and
effective control. Arbitration and other forms of ADR would seem
to offer significant advantages in many types of dispute, but are
not frequently used.

  (a) Is the cost of patent enforcement too high?}
\end{quote}

Note the bias here for patent holders and against the victims of
patents.  Nearly everything connected with the patent system is too
expensive, with one exception: the cost of prosecuting and winning a
patent infringement suit is not expensive at all.  Costs, attorney
fees, and treble damages may be awarded.

Costs to defendants are equal or greater, particularly since a
defendant may not have any warning or chance to prepare.  They are far
too high for a defendant who is a small software developer.

As a result, even patents generally thought to be bogus cause real
problems for software developers.

\vspace*{.08in}
\noindent{\bf IV\@. First-to-File System}

\begin{quote}
{\em In the United States patent system, when more than one patent
application is filed claiming the same invention, the patent is
awarded to the applicant who establishes the earliest acts of
invention in the United States and who has not thereafter suppressed,
abandoned or concealed his invention (``first-to-invent'' system).  In
international treaty negotiations aimed at more uniform patent laws
throughout the world, particularly the World Intellectual Property
Organization (WIPO) patent harmonization proceedings, the United
States is being urged to adopt a different procedure, almost
universally used in other countries, under which we would award the
patent to the first applicant to file a patent application for the
invention (``first-to-file'' system).  The principal issue presented
is whether the United States should change from a first-to-invent
procedure to a first-to-file procedure (1) only as part of a
comprehensive treaty to harmonize all of the world's patent systems,
(2) without regard to a harmonization treaty, or (3) not at all.  The
following questions are designed to help frame this issue:

  (f) If the United States adopts a first-to-file procedure, would
it be desirable to codify the rights, if any, of a prior user of
the invention?}
\end{quote}

Regardless of whether the US adopts a first-to-file procedure, anyone
who has demonstrated knowledge of a technique before a suitable date
(perhaps the date of issuance of the patent) should be permitted to
use it.  This should extend to all the customers or users of the
products made by the exempt user of the technique.

\begin{quote}
{\em   (g) Would a first-to-file patent procedure run afoul of the
reference to ``inventors'' in Article I, Section 8, Clause 8 of the
Constitution?}
\end{quote}

\vspace*{.08in}
\noindent{\bf VII\@. Automatic Publication of Applications}

\begin{quote}
{\em   In the United States patent system, patent applications are held
in confidence by the Patent and Trademark Office and not published
until a patent is granted.  Patent rights begin on the date of
grant.  In international treaty negotiations aimed at more uniform
patent laws throughout the world, it has been proposed that patent
applications be published and confer provisional protection
commencing on the date of publication.  The Japanese and European
patent systems provide for publication of patent applications 18
months after their effective filing dates and feature provisional
protection, generally a right to compensation from infringers with
actual notice.  The principal issue is whether United States patent
applications should be published before grant of a patent (1) only
as part of a comprehensive system to harmonize all of the world's
patent system, (2) without regard to a harmonization treaty, or (3)
not at all.  The following questions are designed to help frame
this issue:

  (a) What benefit do United States patent holders and the public
realize from keeping patent applications confidential until a patent
is granted?  What are the detriments?}
\end{quote}

This is detrimental to the public, because it facilitates the practice
of trade secrecy.  It is beneficial only to those who apply for
patents and do not receive them.

\begin{quote}
{\em   (b) What benefits are United States patent holders and the public
expected to realize from a procedure requiring automatic publication
of patent applications?  What are the detriments?}
\end{quote}

This would be an improvement, because it would reduce the uncertainty
that independent developers of the same technique must face, as to
whether they will be faced in the future with a patent.

\begin{quote}
{\em   (c) Should U.S.\ patent applications be published by some fixed time
measured from their U.S.\ or foreign priority filing dates?  Does the
need for a pre-grant publication differ depending upon whether patent
terms are measured from filing date or issue date?}
\end{quote}

The advantage exists no matter how patent terms are measured.

\vspace*{.08in}
\noindent{\bf VIII\@. Patent Term}

\begin{quote}
{\em   In the United States patent system the term of a patent is 17
years from the date when the patent is granted.  Many other
countries provide a 20-year term for their patents from the date
when the application for the patent was filed. The appropriate term
for a patent is under consideration in negotiations aimed at more
uniform patent laws throughout the world, as well as in the
negotiations on the Trade-Related Aspects of Intellectual Property
(TRIPs) in the current General Agreement on Tariffs and Trade
(GATT) Round.  The principal issue is whether the United States
should measure the life of a patent from the date when the
application for it was filed (1) only as part of comprehensive
treaties to harmonize all of the world's patent systems, (2)
without regard to harmonization treaties, or (3) not at all.  The
following questions are designed to help frame this issue.

  (a) What benefits do United States patent holders and the public
realize from measuring the life of a patent from its date of grant
What detriments are there?}
\end{quote}

This system is disadvantageous because it creates the possibility that
a technique will become patented that is already very old.  It means
that one can never be confident that any commonplace technique will 
not become patented next year as a result of an application filed 
decades ago.

\begin{quote}
{\em   (b) What benefits are United States patent holders and the public
expected to realize from a procedure that would measure the life of a
patent from its filing date What would the detriments be?}
\end{quote}

The benefit of this change is the elimination of the disadvantages of
the current system.

\begin{quote}
{\em (c) If the term of a United States patent is to be measured from
the filing date, should the term be extended to compensate for delays
due to (1) secrecy orders, (2) marketing delays due to FDA or EPA
regulatory procedures, (3) NASA or DOE ownership proceedings, (4)
appeals, or (5) any other delays outside of the control of the patent
applicant?}
\end{quote}

Any possibility of extending the term is greatly threatening to the
public ability to benefit from public disclosure of the invention
before it is obsolete.  So in general the term should not be extended.
The patent system is already a lottery, so the possibility of a
reduced effective term in an unlikely circumstance will not change its
nature, nor make much quantitative difference on the average.  If an
appeal is necessary for a patent, it is likely to be marginal in its
merit, which is additional reason not to extend the monopoly.

In particular, the term should not be extended because of NASA or DOE
ownership proceedings.  These agencies should act for the public,
which means encourage rather than interfere with the use of a useful
technique.

However, for drugs, where delays due to regulatory procedures are
normal and predictable, and where obsolescence typically comes more
slowly, a basic term longer than used in other fields might be
appropriate.

\begin{quote}
{\em   (e) If the term of a United States patent is to be measured from
a filing date, should that date be (1) the earliest United States
filing upon which priority is based, (2) the earliest filing upon
which priority is based, be it foreign or United States filing
date, or (3) the filing date of the application on which the patent
is granted?}
\end{quote}

If the applicant is granted the benefit of priority as regards rejecting
prior art, the applicant should have to accept the same date as regards
when the patent will end.

\begin{quote}
{\em   (f) If the United States adopts a patent term based upon a filing
date, should that term be 20 years?  If not, how long should the term
be?}
\end{quote}

For some fields of industry, 20 years may be reasonable.  For the
field of software, however, it is absurdly long.  Even five years is
too long for our field.  20 years is also unreasonably long for the
field of computer hardware: it spans several generations of computers.

\vspace*{.08in}
\noindent{\bf IX\@. Deferred Examination}

\begin{quote}
{\em United States patent laws permit an applicant to petition for a
deferral of the examination process only for a short period, and then
only for ``good and sufficient cause.'' Deferred examination at the
option of the applicant is a feature of several foreign patent
systems.

  (a) What would be the benefits of a procedure whereby a patent
application is not examined in due course, but instead, the applicant
is given the option, for a period of years, to request that the
application be examined What would be the detriments?}
\end{quote}

If the term of a patent is measured from the date of issue, then 
permitting the applicant to defer the examination permits an abuse:
while the technique spreads, or is reinvented, and its usage becomes
more entrenched, the applicant stands only to gain.

Whenever patents are a factor, the public deserves to be told quickly
whether people are permitted to use any new technique.  The Patent
Office has a responsibility to resolve this question with no undue
delay.

\begin{quote}
{\em (d) Should the applicant have the option, in lieu of requesting
formal examination, to obtain a lower-cost patent of shortened term
and reduced enforcement rights?}
\end{quote}

Using patents of this sort for software instead of the current 17-year
patent, with a term of only three years, might greatly reduce the
problem of software patents.  No matter how much trouble a given
patent may cause, it can't be too bad if it is limited to three years.

\vspace*{.08in}
\noindent{\bf XI\@. Reexamination}

\begin{quote}
{\em The reexamination statute permits any person to file, with the
requisite fee, a request for the reexamination of a patent in view of
prior art limited to prior patents and printed publications.  Issues
of inequitable conduct, public use and sale, and the like, may not be
addressed in reexamination.  If the USPTO determines that the request
raises ``a substantial new question of patentability,'' the patent is
reexamined under the same procedures applicable to the examination of
new applications and concludes with the issuance of a reexamination
certificate.  The reexamination process is basically ex parte in
nature, with very limited participation by a third party requestor.
Since the present system of reexamination was established by Congress
in 1980, there have been numerous proposals to modify reexamination to
increase the degree of participation of the requestor and/or to
enlarge the scope of reexamination.

  (a) Do you believe the present reexamination system is working
effectively?  If not, identify each deficiency that you believe
exists in the system.}
\end{quote}

The present reexamination system is not a reliable way of clearing
away bad patents.  Because reexamination works through a discussion
between the Patent Office and the patent holder, it is likely to lead,
not to the elimination of a bad patent, nor even to the elimination of
a single claim, but only to making it slightly narrower.

\begin{quote}
{\em   (b) Do you believe that the present reexamination system should
be modified to permit complete inter partes participation by a
protestor at all phases of the reexamination proceeding?  If so,
explain why.  If not, explain why.

  (c) If your answer to question (b) is ``no,'' to what extent, if
any, should third parties be allowed greater participation than at
present in reexamination proceedings?}
\end{quote}

It is not clear that either of these would improve the outcome of
reexamination, unless by helping to convince examiners that certain
techniques are obvious.

\begin{quote}
{\em   (d) Would you be in favor of modifying the present reexamination
system to permit or require one or more of the following:}
\end{quote}

Anything that gives opponents of the patent a greater effective voice
will be an improvement, given the lax criteria now being employed by
the US Patent Office.  However, as long as reexamination permits the
patent holder to respond by gerrymandering the claims, it is unlikely
to be very effective for correcting the mistakes of the patent system.

\vspace*{.08in}
\noindent{\bf XII\@. Assignee Filing of Applications}

\begin{quote}
{\em   The United States currently requires the actual inventor or
inventors to file patent applications, subject to a few limited
exceptions such as death, unavailability, or insanity.  Each
application must be accompanied by an oath from the actual inventor
which states that the applicant believes he was the first to invent
the claimed invention.  Nearly all foreign patent systems permit
the owner of an invention (whether the actual inventor or the
assignee) to file the patent application without an oath signed by the
inventor.

  (a) Should the United States allow filing of patent applications
by assignees?}
\end{quote}

No.  Three hundred years ago when our forebears elected to limit
monopolies to inventors, they were reacting to their experience with
``projectors'' who purported to represent the inventors.

This decision is restated explicitly in the United States
Constitution.

The inventor's signature provides a check on the desire of companies
to file for patents of a nature regarded as obnoxious by the actual
practitioners of the art.  As the jury system makes public consent
necessary in order to convict someone of a crime, requiring the actual
inventor to sign the patent application makes the issuance of a patent
depend on an inventor's belief that a patent is legitimate.

\begin{quote}
{\em   (b) What benefits and drawbacks would occur by allowing filing
of patent applications by assignees?

  (c) What contours should an assignee filing proposal have,
including possible safeguards for inventors and the public?}
\end{quote}

\vspace*{.08in}
\noindent{\bf XIII\@. PTO Funding and Fee Structure}

\begin{quote}
{\em The Patent and Trademark Office historically recovered its costs
through a combination of fees and funding from general tax revenues.
The major patent fees are filing fees, and issue and maintenance fees
in a two-tier system with lower rates for small entities.  These fees
have undergone substantial increases recently as Congress and the
Administration have changed from a policy under which applicants and
patent owners paid only a portion of these costs to one under which
they pay full costs.

  (a) What impact, if any, will a system in which PTO costs are almost
entirely user-fee funded have on (1) U.S.\ inventors generally, (2) on
small entities, including independent inventors, and (3) the public?}
\end{quote}

User funding means funding primarily by patent applicants.  This
raises the danger that the Patent Office will view its role as one of
serving patent applicants, rather than serving the public.  It also
removes all direct incentive for Congress to rein in the activities of
the Patent Office.

\begin{quote}
{\em   (e) Would it be desirable to increase the average application
pendency time in the PTO to a level above 18 months if such increase
permitted a significant reduction in PTO fees?}
\end{quote}

Under the current system, where the term of a patent is measured from
the date of issue, this change would be harmful in two ways:

\begin{itemize}
\item It would be likely to increase the number of patents.

\item It would increase the surprise effect of each patent.
\end{itemize}

\begin{quote}
{\em (l) In discussions of patent harmonization, the question of
pre-issuance publication of U.S.\ patents has been raised.  Given the
possible reluctance of Congress to fund the costs of such publication
from general revenues, would you support such publication if its cost
were totally user-funded?}
\end{quote}

It may be possible to publish the applications very cheaply using
computerized media and on-line access, and publication would still be
useful even if its costs were funded by fees for obtaining copies from
the Patent Office or for searching the computerized files of the
Patent Office.

\newpage

\begin{center}
{\Large\bf Appendix}
\end{center}

\vspace*{.5in}
\noindent{\bf An Egregious Patent Explained}

U.S.\ patent number 4,956,809 covers using a single standard byte
ordering for transfer of data between machines whose normal byte
ordering is different.  Unless invalidated, this patent prohibits the
only principal means for transfering data between all types of
computer systems.

Here is a layman's explanation of the communication problem, and of
the solution which this patent covers.  The solution is considered so
obvious by programmers that most of them laugh and find it hard to
believe it was patented in 1990.

The memory of today's computers is divided into units called bytes.
Each byte holds a single text character, or a number from 0 to 255.
Each byte of storage has a unique address.  These addresses start at 0
and range upward.

When you need to represent a number in a range bigger than 0 to 255,
you need more than one byte.  For example, two bytes can hold a number
in the range from 0 to 65535.  Four bytes can hold a number in the
range from 0 to 4294967295.  This four-byte number might be held in
bytes 72, 73, 74 and 75---four consecutive bytes.

There are various options for distributing the value of the number
among the bytes.  In practice, most of today's computers use one of
two methods.  The ``little endian'' method is to put the less
significant figures in the low-numbered bytes.  The ``big endian''
method is to put the most significant figures in the low numbered
bytes.  (These names come from @cite{Gulliver's Travels}.)

You can think of this as the computerized equivalent of deciding which
way to write the digits in a number.  Normally we put the most
significant digit on the left, and the other digits follow in
decending significance as we move to the right.  Thus, one hundred and
four is written 104.  Since the most significant digit comes first (in
our left-to-right convention for ordering all text), this is
big-endian.

It would be just as practical to write the digits in the opposite
order---first the one's digit, then the ten's digit, and so on.  Then
one hundred and four would be 401.  That would be little-endian.  One
advantage of little-endian format is that addition and subtraction can
be done by considering the digits in the order they are written.  Big
endian order requires reading the digits from back to front.

If you transfer a block of data from a big-endian computer to a
little-endian computer, preserving the order of the bytes, all the
multi-byte numbers will be scrambled.  Likewise, if a big-endian
person writes the number one hundred and four as 104, a little-endian
person will misread it as four hundred and one.

We can solve this confusion by establishing a conventional publication
order for the digits of numbers.  For example, we can declare that all
numbers in published works will be written with the one's digit last
(big-endian).  People who normally think of numbers this way will read
and write them normally.  People who normally think of numbers with
the one's digit first (little endian) would read and write numbers
``backwards'' (from their point of view) in published works.

The same solution works for computer communication.  The simplest way
to communicate data between different computers is to set up a
convention for the ordering of the bytes in any multi-byte number.  If
the big-endian ordering convention is chosen, then little-endian
computers must reverse the byte order when they read and write network
messages.

This is exactly what is covered by US patent 4,956,809.

Here is a message containing big-endian numbers:

\begin{quote}
March 16 1953
\end{quote}

The same message, containing little-endian numbers:

\begin{quote}
March 61 3591
\end{quote}

Note that words like ``March'' are not reversed, because they are
composed of characters, which are single-byte values.

\newpage

\begin{center}
{\Large\bf Appendix}
\end{center}

\vspace*{.5in}

\noindent{\bf How Patents Prevent Use of Public-Key Mail Encryption}

Programmers who haven't grasped the full scope of the harm done by
software patents sometimes offer the RSA patent as an example of a
software patent that ought to exist.  This is because they feel that
RSA encryption is not obvious, the way backing store is obvious, or
natural order recalculation is obvious.

We can't deny that the RSA encryption algorithm is an interesting
discovery.  But this does not imply that issuing such patents benefits
the public.  The example of RSA shows how much harm such a patent can
do.

Public key encryption was developed in the 1970s by professors at MIT
and Stanford.  These schools gave an exclusive license to a company,
called RSA Inc., which has done its level best to prevent the most
obvious and useful application of public key encryption: the delivery
of encrypted mail on the Internet and Usenet.

The adoption of any new method of communication encounters a ``chicken
and egg'' problem: it is not very useful until a large number of
people are using it.  (Recall how only explicit government
intervention to require UHF support in all new TV sets was able to
bring about use of the new UHF television channels.  Until then,
viewers would not pay for UHF capability because nothing was being
broadcast on those channels, and no one would set up broadcasting
because there was no one to receive.)

Making encrypted mail useful likewise depends on having many people to
talk to through it.  But, unlike the case of UHF TV, this does not
require the government to force anyone to support it.  There is a
simple way to give a large number of people the ability to use this
form of communication: just produce a public domain program and
distribute it on computer networks.  This has taken place in Europe.
There have been several attempts to do this in the United States; RSA
Inc.\ has quashed every one.

The RSA patent was obtained for an algorithm called RSA after the
names of the professors that developed it---Rivest, Shamir and
Adelman.  Recently a public domain mail encryption program was
released using a related algorithm, called the Rabin algorithm.  RSA
Inc.\ claims their patent covers this algorithm, too.

In fact, they claim their monopoly extends to any encryption algorithm
that performs two or more exponentiations and combines the results.
This is a very broad class of algorithms.  (The president of RSA likes
to boast of this broadness as if it were a measure of his merit.)

This illustrates an important half truth in statements made by the US
Patent and Trademark Office that they do not grant patents on
algorithms.  It is true that no patent covers a single algorithm; this
is because a patent monopolizes a class of many related algorithms.

While you might naively think that prohibiting a broad class of 
algorithms is even worse than prohibiting one, the Patent Office
doesn't think so.  By careful choice of words, they can deny the
existence of algorithm patents, and give the impression that algorithms
are not being prohibited by patents.  For example:

\begin{quote}
First, many assumed that we would grant a patent on a computer program
or a mathematical algorithm.  As we have stated, we will issue patents
on computer processes..., not on programs or mathematical algorithms.
\end{quote}

\noindent
(This comes from testimony of Jeffrey Samuels, acting commissioner of
patents and trademarks, in the house subcommittee on intellectual
property.)

This case also shows how the patent system can do great harm while
doing absolutely no good.  The algorithms concerned were developed by
professors who inevitably would have published them, patent system or
no.  There was not even a possibility that society would benefit by
issuing this patent.

Another interesting aspect of this case is that several MIT professors
now receive government funding to do research which cannot be used
without infringing the RSA patent.  The effect is to provide a
government subsidy for RSA, Inc., under the guise of research done for
the public good.

\newpage

\begin{center}
{\Large\bf Appendix}
\end{center}

\vspace*{.5in}

>From an article in ``Think'' magazine, \#5, 1990 (a magazine published
by IBM):

\begin{quote}
    ``You get value from patents in two ways,'' says Roger Smith, IBM
    Assistant General Counsel, intellectual property law.  ``Through fees,
    and through licensing negotiations that give IBM access to other
    patents.

    ``The IBM patent portfolio gains us the freedom to do what we need to
    do through cross-licensing---it gives us access to the inventions of
    others that are the key to rapid innovation.  Access is far more
    valuable to IBM than the fees it receives from its 9,000 active
    patents.  There's no direct calculation of this value, but it's many
    times larger than the fee income, perhaps an order of magnitude
    larger.''
\end{quote}

This information should dispell the belief that the patent system will
``protect'' a small software developer from competition from IBM\@.  IBM
can always find patents in its collection which the small developer is
infringing, and thus obtain a cross-license.

However, the patent system does cause trouble for the smaller
companies which, like IBM, need access to patented techniques in order
to do useful work in software.  Unlike IBM, the smaller companies do
not have 9,000 patents and cannot usually get a cross-license.  No
matter how hard they try, they cannot have enough patents to do this.

Only the elimination of patents from the software field can enable
most software developers to continue with their work.

The value IBM gets from cross-licensing is a measure of the amount of
harm that the patent system would do to IBM if IBM could not avoid it.
IBM's estimate is that the trouble could easily be ten times the good
one can expect from one's own patents---even for a company with 9,000
of them.

IBM does not actually experience this burden---they avoid it through
cross-licensing.  For ordinary companies which cannot do likewise,
this burden is real.  IBM's estimate suggests that for a typical
software company, patents will do ten times as much harm as good.

\end{document}