Re: Why software patents cannot work
From: email@example.com (Dennis Ritchie <7549-15328> 0112710)
Date: Apr 03 1995
Piercarlo Grandi ponders,
You mean there have been thousands of "software" patents a year in the
40 years after 194x? Perhaps thousands of patents ayear on mathematics
in the past 200 years? This is amazing news!
There are many people instead that had formed the funny impression that
for those 40 years [1940-1980] the PTO routinely rejected algorithm patents
(allowing patents on industrial and other physical processes,
e.g. curing rubber), and that it still is supposed to reject patents on
mathematical theories/theorems/... after 200 years of doing so.
I hold in my hand US Patent 3,568,156, entitled "Text Matching
Algorithm," filed 9 Aug 1967, issued 2 Mar 1971, inventor Ken
Thompson. It concerns a clever way of implementing a NDFA for doing
regular-expression searches; the method was used in several of our
versions of the qed editor. (Not ed; that used a simpler scheme, and
not [ef]grep, they used other ideas). The disclosure is chock full of
pseudo-Algol and IBM 7094 code for the preferred implementation. Then
it says, "Although it is less likely that the algorithm of the present
invention will be implemented by means of special-purpose circuitry,
such circuitry is illustrated in fig 2...."
I don't hold in hand my own patent on the Unix SUID bit, and don't
remember the number. Although the algorithmic content is trivial
(check a bit; if it's set do an assignment) it is very visibly a
software patent, and it issued well back in the '70s.
What I'm arguing is that neither software nor essentially algorithmic
patents are particularly new. The general metes and bounds in the US
may have shifted somewhat with court decisions and PTO's desire either
to avoid or to take on work. However, my guess is that recent
increase in software patent activity is much more affected by software
producers' perception that their product is valuable and worth
protecting than by changes in the legal situation.