OT: How I learned to stop worrying and love software patents
[Note to d-l readers: the subject is tongue-in-cheek, mmmkay? Film reference.]
On 7/24/05, Nathanael Nerode <firstname.lastname@example.org> wrote:
> Michael K. Edwards wrote:
> > Patent is not copyright; you don't obtain a monopoly on describing
> > your method, you obtain a monopoly on its commercial application. No
> > patent prohibits you from making a computer program implementing any
> > algorithm you like; but if you sell it as a solution to the problem
> > addressed in the patent, without authorization from the patent holder,
> > you are infringing. The same goes for selling its output, if that's
> > covered by the patent -- compare against the enforcement of chemical
> > process patents.
> Thanks for this informative comment. So I guess you would phrase the problem
> differently, but perhaps you agree on the existence of the problem. As far
> as I can tell, (a) mathematical problems are being used as "problems" in the
> patent domain (apart from "solving a system of linear equations", a cipher is
> a mathematical transformation and the problem of finding one is a
> mathematical problem); (b) giving things away is considered just as bad as
> selling them; and (c) selling it as a solution for a different problem is
> considered just as much a violation as selling it as a solution for the same
> problem. I really hope one or all of these is not true, but every time I
> look at something in software, they all seem to be true.
I think these are all judgment calls under the current system, and
doomed to be judgment calls under any system that is making an honest
effort to "promote progress in the useful arts and sciences". Let's
take your example of ciphers as "mathematical problems".
Cryptanalysis is higher math, all right, and no patent on a technique
of cryptanalysis should be permitted to issue (and as far as I know,
none ever has). But the design of a cipher resistant to cryptanalysis
shades over into industrial application of that math as well as the
more mundane techniques of efficient coding, just as the design of a
corrosion-resistant alloy is an industrial application of surface
chemistry as well as metallurgy per se.
As for (b) and (c): There is not ordinarily a distinction between
giving away and selling an infringing item -- after all, there are
ways to profit from undermining part of a rival's business model other
than extracting per-unit revenues from his potential customers. And
while the domain of application and the nature of the useful result
are supposed to be part of the statement of a patent's claims, there
is some leeway to argue that the disclosure equally teaches an
invention in a closely related area (such as video vs. still image
compression). But constructing a device or writing a program that has
both infringing and non-infringing uses, and taking some care not to
encourage or extract revenues from infringing use, can be a pretty
good defense against a contributory infringement claim -- especially
where the novelty claimed in the patent is limited to applying a
well-known technique to an unexpected problem. (IANAL, TINLA.)
There isn't a lot of litigation (as opposed to bluster, brinksmanship,
and political posturing) on software patents, and AFAICT what there is
is pretty benign. I think the State Street decision is somewhat
dubious (as I have written earlier), simply because it risks a major
change in the effective scope of patent protection without any
indication of a legislative or Supreme Court mandate to do so. Ditto
EPO parallels such as Sohei (T 769/92) and Pettersson (T 1002/92).
The "useful arts and sciences" are traditionally understood to be
those that manipulate the natural world, not markets. But I don't
really have any problem with jettisoning the dualist baggage and
adopting the Alappat test to call digital results "tangible" if they
represent a useful advance in the state of the art in an applied
In any case, just because a patent issues doesn't mean it isn't
complete crap. Don't be fooled into lumping USPTO howlers (or, for
that matter, their EPO parallels) in with those which have survived
scrutiny in an adversarial setting.
> > applications of software techniques to practical problems
> > are just as patentable when stated using "process" lingo as when using
> > "machine" lingo, certainly now (per AT&T v. Excel) but AFAICT all
> > along.
> Well, it's straightforward, anyway. Any algorithm is a "process" by
> definition. The problem is that it's a *mathematical* process. If you don't
> have a prohibition on the patenting of mathematics -- and apparently we don't
> have one on the statute books in this country -- algorithms are obviously
Whether or not the statute calls it out explicitly, mathematics per se
is in the domain of "laws of nature, natural phenomena, and abstract
ideas" (quoted from Diehr, paraphrasing cases going back to 1853),
which are reliably excluded from patentable subject matter. I invite
you to question the assumption that "algorithms are mathematics". My
preferred US dictionary (American Heritage, third edition) has it that
an algorithm is "a step-by-step problem solving procedure", and goes
on to describe the computational specialization of this idea. That's
not really theoretical mathematics any more than a titration technique
is theoretical chemistry.
So I don't see a good public policy reason why a sequence of
arithmetical steps, reduced to practice as a solution to a concrete
industrial problem, shouldn't be every bit as much patentable subject
matter as a sequence of chemical steps that result in a useful polymer
or of mechanical steps that result in seed-free cotton. There are
still other hurdles to jump (novelty, non-obviousness, disclosure of
best mode) that eliminate the parade of horribles with which some
anti-software-patent polemics would like to frighten you.
> The really nasty thing is that I can losslessly transform a
> large number of other mathematical constructs into algorithms.
> "Practical problems", of course, is not a restriction at all, since a patent
> on using a process to do one thing apparently applies to using the exact same
> process to do anything else too. :-P (Although a new patent can be granted
> for the novel use, AFAICT that just means that a user has to license *both*
Where do you get the idea that the techniques in a process patent are
detachable from the "method of [achieving some useful result]"
language with which an independent claim conventionally begins? When
we parse claim language, we're delving into the area of administrative
guidance here rather than statutory requirements as such; but the
whole idea is that a patent doesn't cover applications that aren't,
within reason, both taught in the disclosure and stated in the claims.
Refinements, yes, even if they involve an additional quantum of
patentable novelty (and therefore qualify for subordinate patents);
but not inventions that the claim language (including domain of
application) doesn't fit.
For a real application of this principle, read the Alappat opinion.
Do you think Alappat could claim to have patented anti-aliasing in any
context other than displaying waveforms on a digital oscilloscope (or
a functional equivalent implemented in software)? I think not.
Again, just because a patent issues doesn't mean its claims aren't
improperly broad. A claim that begins with "a method of processing
digital signals, representing anything under the sun" is (or should
be) a target for special scrutiny in district court. If the
disclosure fails to show how the process achieves a concrete objective
in a patentable subject area, then it should (and probably will) be
tossed out on Section 101 grounds.
Where encryption is concerned, I could go either way. It's certainly
got industrial uses; but whether a patent like #4,405,829 ought to be
granted as written, or perhaps limited to its "communications system"
claims and not enforceable against mere transformations of blobs of
data, is an interesting question. Not, as far as I know, one that has
ever been tested in court; and the opportunity has also existed to
test it in European courts (not that same patent, but EP 0482154) and
has not been seized there either.
[snip fulmination against judicial conduct without any grounding in
> See above. Oh, for a chaser: data in RAM is "fixed in a medium of
> expression", because the electricity *might* not be turned off. Whether or
> not the electricity actually is likely to be turned off or not (in the
> context of the ruling, it was almost certainly going to be). Actually, RAM
> is essentially "continually recopied" in the internal circuit; I know just
> enough EE to know that that's the actual facts. There's another one which can
> be reduced to absurdity easily. According to the same logic, a pair of
> people repeating a phrase to each other alternately would fix the phrase in a
> medium of expression (because they might not stop), even if they had every
> intention of stopping.
To what actual case are you alluding? Perhaps you're talking about
the obsolete Freeman-Walter-Abele test for "physical limitations", and
the peculiar contortions that patent agents used to go through to
dodge Benson and Flook by claiming "machines" and "apparatus" instead
of "processes". If you read AT&T v. Excel, I think you will find that
the Federal Circuit agrees that the whole obsession with "physical"
effects in the 70's and 80's was pretty silly.
> Look, judges have been fairly clueless about technical matters -- on any
> technology! -- most of the time. The wise ones appoint special masters.
> Most of them don't. Eventually, we may expect that the consensus of the
> specialists in a field may penetrate the world of the judges, but it seems to
> require generations. The early rulings about telephones and telegraphs were
> all wrong, too, IIRC.
Special masters are useful for establishing some kinds of facts,
including the realities of business practice in a particular industry.
Judges are emphatically prohibited from delegating questions of law
and statutory interpretation to special masters or anyone else.
That's not only how it should be, that's how it is.
> I still suspect that judges haven't noticed that the algorithms being patented
> are pure mathematics. If they had, they would have had to face up to the
> question of whether mathematics should be patentable honestly. Instead, they
> keep coming up with weasel-words. But if a mathematical solution to a
> mathematical problem is patentable, then any sort of mathematics is
> patentable, and there's no getting around it.
I see your "weasel-words" and raise you "horse-pucky". You are
impugning the intelligence and integrity of a whole class of dedicated
public servants, whose actions are subject to more public scrutiny
that any other branch of government, on pure hearsay. Tell me what
cases you would have decided differently and why -- taking into
account the actual constraints of statutory language and stare decisis
-- and then we can talk about where the error lies on the spectrum
from ignorance to corruption.