THE SOLUTION TO THE PATENT PROBLEM
By Gene Michael Stover
created Sunday, 8 August 2004
updated Sunday, 8 August 2004

   Copyright (c) 2004 Gene Michael Stover. All rights
   reserved. Permission to copy, store, & view this document
   unmodified & in its entirety is granted.

********

The "patent problem" I'm discussing is software patents.
Patents are granted to algorithms & data structures which
are obvious, or for which there is prior art, or on which
society already depends.  I know the solution to that
problem.

The real & total solution would be to refrain from granting
patents to algorithms that are obvious or for which there is
prior art.  The patent agency should do their best at this,
but they can't be perfect, so we need another way to help
minimize the damage done by patents which shouldn't have
been granted.

The point of a patent is to reward an inventor (or an
algorithm or technique or new kind of widget or whatever) by
granting him an exclusive right to make use of his
invention.  A patent has a finite life-time after which the
protection ceases.

The life-time of a patent is usually 15 years.  (Ir is it
30?  I don't know, & the exact duration doesn't matter, as
you'll see.)

The duration of a patent should be the time until someone
else would have invented the algorithm, technique, widget,
or whatever if the first inventor hadn't.  The first
inventor shouldn't have protection after someone else would
have invented it anyway.

How do you know when someone else would have invented it?
It's probably impossible to know for sure, but it could be
approximated by looking at the time spans between inventions
in that field.  A simple formula would be to take the
arithmetic mean between the dates at which patents in that
field were filed.  A better formula would be to use the
dates at which patents in that field were filed & use them
to derive a polynomial which accounts for a trend of
accelerating or decelerating rate of discoveries in that
field.  I'm sure there are even better techniques for
approximating the time at which someone else would have
discovered the invention in question.  The actual method of
approximating that time is a detail that could be decided
later.

So the duration of the patent should only be until someone
else probably would have discovered the invention
independently if the original inventor hadn't done it first.

This is an "appropriate duration" for the patent.  Patents
for inventions from different fields, industries, or
disciplines would have differing durations.  Patents within
a single field might have different appropriate durations if
the period of time between their filing dates was length,
like maybe 100 years, because the rate of progress in that
field might have changed between the earlier patent's filing
& the later patent's filing.

SO HOW DOES THIS SOLVE THE PROBLEM OF PATENTS IN SOFTWARE?

Appropriate durations don't prevent patents that shouldn't
have been granted in the first place, but they minimize the
damage done by such patents.

Though I haven't heard people say it in years, a maxim of
the computer software & hardware industry used to be "a
product is out of date six months before it is released".
That sure sounds like "someone will have invented a better
way of doing it six months before you release your product
that does it".  If the average development time for such a
product is, say, 24 months, & if someone discovered a better
way of doing it 6 months before the product was released,
then the appropriate duration for the patentable ideas on
which the product was based would be 18 months.

Eighteen months is just a quick approximation; it's based on
folklore, not research, but it shows what I'm talking about.
What if software patents were granted for a mere 18 months?
Would it be so bad for a corporation to patent an algorithm
or data structure for that duration?  Would corporations
even try?

Besides that, an appropriate duration to patents is
ethically appropriate.  If the techniques of brick-laying
haven't changed in, say, one hundred years, then the person
who discovers a significantly better way of laying bricks
deserves long protection for his invention.  (I'm not sure a
patent duration of 100 years would be a good thing; maybe
appropriate durations should have a maximum of something
like one generation, which would be about 25 years.)  A
person who makes a discovery in a field in which discoveries
are made multiple times a day (& I mean that literally)
shouldn't receive more than a few months of protection.

********

$Header: /home/gene/library/website/docsrc/pat/RCS/pat.txt,v 395.1 2008/04/20 17:25:47 gene Exp $

--- end of file ---