[A2k] CLS Bank v Alice May Signal Turning Point in Software Patents

Seth Johnson seth.p.johnson at gmail.com
Sun Feb 10 14:26:04 PST 2013

On Sat, Feb 9, 2013 at 4:02 AM, "Gérald Sédrati-Dinet (Gibus)"
<gibus at april.org> wrote:
> Le 08/02/2013 21:32, Seth Johnson a écrit :
>> https://www.infoworld.com/d/open-source-software/federal-hearing-today-marks-turning-point-in-war-software-patents-212445
>> Speaking up for open source and the digital society, we find the
>> Electronic Frontier Foundation (EFF) and Public Knowledge. Their brief
>> is significant because it offers the approach devised by Stanford's
>> Professor Mark Lemley as a solution. As I wrote here in September,
>> Lemley suggests that the patent statute of 1952 has been incorrectly
>> interpreted concerning software.
> I disagree strongly, Lemley's proposal is taking us back ten years ago
> when we wondered whether disclosing an algorithm would solve the
> software patent problem. The answer we've reached was clearly NO.

<Haphazard insert point; Gerald has good stuff below>

An exchange with Mark Lemley right after I posted that article:
> https://www.facebook.com/mark.lemley/posts/10151246380590846

Seth Johnson posted to Mark Lemley
Friday at 4:01pm ·

    To Mark Lemley: Sure, there's a good, solid, unambiguous
definition of a software patent: A patent that covers instructions (in
any physical form) provided to a general purpose logic processor. It's
software because it's instructions, and it's pure abstraction because
that's all a logic processor does. Ergo, simple legal solution: 1) is
there a logic processor in a claimed thingy? 2) are there separate
instructions being provided to it? If both are yes, those instructions
are not covered by a patent (which might nevertheless claim the
overall apparatus, just not the pure logic instructions inside it).

        Mark Lemley A couple of questions. (1) why permit a claim to
the programmed apparatus if the only novelty is the instructions? (2)
why treat a general purpose and a special purpose processor
differently, so that software to control an MRI machine is patentable
but software to control an MRI machine from a PC isn't? (3) is music
software under this definition? the text I typed in this comment?
        Friday at 5:10pm · Like

        Seth Johnson 1) I wouldn't presume the instructions are novel.
There is the possibility that the instructions are controlling a novel
physical process. The instructions, even controlling such a process,
are pure, not necessarily designating any particular process except
that they happen to have particular devices connected to them. But, on
the other hand, if the only thing that's novel in the whole doohicky
is the instructions, the instructions don't qualify, because they are
solely pure logic, not necessarily about any particular thing in the
real world. They are not just abstract (like general physical laws),
but *pure* abstraction (like math or logic), which is independent of
empirical particulars. The most novel mathematical discovery doesn't
qualify for patenting. Now of course, if you start interpreting the
numerical code as referencing particulars in the real world, then
you're interpreting it according to a higher level language (and not
solely as math or logic), and at that point it's a description of the
physical process "on the other side" of the processor, and you are not
considering the actual pure op codes that the logic processor receives
in their nature as such. 2) There's no software in the MRI that you
say is patentable, unless it has a general purpose logic processor in
it. If the MRI has no general purpose logic processor, but some sort
of unique controller that is not a Turing-complete logic controller
(conceivable, but increasingly unlikely and inefficient design), then
the instructions controlling it would not qualify as software (pure
abstract logical operations) in this scheme. In an approximate way of
speaking, the "special purpose processor" to be relevant as a
counter-example, would most likely be some sort of mechanical gizmo.
You can be Turing complete with really reduced instruction sets, so
most of the time, your first guess should be a processor is a general
purpose one at core, and only if it's not being fed pure logic but
just a narrow set of triggers activating a narrow set of particular
mechanical or chemical reactions would it really be special purpose at
core. You can make a general purpose processor seem to be special
purpose, by the devices you attach. But (generally, almost always) you
still have a general purpose processor in the device, and "on the
instruction side" it's receiving pure logical operations. What
translates those operations into particular processes, is the devices
plugged into it "on the other side." 3) Music (coded by some
convention) might conceivably be fed into a logic processor, and
probably wouldn't in very many cases actually do anything useful if
used that way. However, if you file a claim with instructions fed to a
processor that just happen to also correlate with Beethoven's Fifth by
some convention, then those notes translated into op codes are
software in that device. A similar point applies for words/text.
        Friday at 6:00pm · Edited · Like

        Seth Johnson By identifying the presence of the logic
processor and the separate instructions, you are proving two things:
1) that they are instructions (not just a physical medium, but
instructions conveyed by the medium that do in fact represent
instructions); and 2) that they are pure logic (because that's all a
logic processor does).
        Friday at 5:45pm · Edited · Like

        Mark Lemley Ah, so you mean something different by "general
purpose logic processor" than I would have imagined, and certainly
than courts do when talking about a general-purpose computer. With
that broad definition music in any computer format would certainly
qualify as software. And many, many technologies would be unpatentable
because they are controlled by instructions that go into some device
with a processor in it.
        Friday at 6:06pm · Like

        Seth Johnson Those technologies could be patentable, inasmuch
as the physical things they're doing qualify as novel independently of
the software. Just look for whether the physical stuff is like that.
But if inside them they're using pure logic, applying formulas and
logical algorithms, those aren't covered.
        Friday at 6:08pm · Like

        Seth Johnson The patentable device should never be able to
"capture" software -- math or logic -- inside it.
        Friday at 6:17pm · Edited · Like

        Seth Johnson Now, yes, still a lot of things that have gotten
away with being patentable would no longer be -- but in that respect,
we're seeing how the gee whiz of computers has motivated confusion
about what patents actually are. Design patents? Ridiculous. Is there
actually an empirical invention? That's the real question. And here's
the thing: while lots of things patent attorneys have gotten away with
would fall away, with huge implications for industry, what happens
once you exclude software by this definition, is it's never mixed in
with what's considered among the claims -- you get back to considering
whether there's a real empirical invention there.
        Friday at 6:13pm · Edited · Like

        Seth Johnson No, music would not qualify as software unless it
is being provided as instructions to a logic processor. If it's not
being used to do that, it's maybe a higher level language designating
notes for a performer. If you turn the notes into op codes and that
actually does something, sure it's software.
        Friday at 6:16pm · Like

        Mark Lemley How about the hybrid gasoline-electric engine? The
mechanical pieces aren't new, just the system for controlling which
one we draw power from when. And that's software.
        Friday at 6:18pm · Like

        Seth Johnson Well, the software in it isn't about gas or
electric, except maybe in the higher level language code that humans
use to write software. At that level, it's a description of a
particular use of what is actually at bottom a mathematical algorithm
that doesn't have any necessary relation to gas or electricity or
engines. As a description, maybe that's novel (maybe at least once?),
to say let's have an engine that will switch power sources under
software control. It's a cheap trick, what you talk about in your
work, but even assuming the first time somebody says that's a nifty
thing to do, and so we give a patent to that guy, still that software
that's doing the switching and whatever else it does, ain't covered.
There's a logic processor in there (very likely), and it's just doing
pure logical operations. It's the engine attached that's doing the
potentially novel thing. Bad example of a novel thing, as you and I
know, but the principle applies here as anywhere.
        Friday at 6:28pm · Edited · Like

        Seth Johnson We'd probably very quickly apply your principle,
recognizing that the trick of just saying a computer attached makes it
novel is illegitimate. But I'm just showing the principle works
regardless of the patentability of the physical device. We should just
recognize from the outset that in every case, that code, in the form
the logic processor understands, is pure, and just not covered.
        Friday at 6:35pm · Edited · Like

        Robert Sachs · Friends with Joshua Sarnoff and 2 others
        Mark: "It's pure abstraction because that's all a logic
processor does" Nope. That's literally a non-sequitor. A logic
processor does not "do" pure abstraction. A logic processor processes
bits (electrical signals) using a complex of logic gates. "Pure
abstraction" is something else entirely. If you mean by this what the
Supremes called "abstract intellectual ideas", then this refers to
concepts which are disembodied attributes: Virtue, Humility, Justice.
Even "roundness" or "redness". The latter two are pure abstraction
because you cannot "drop them on your foot". You can say something is
"round" or "red", but you cannot have "roundness" by itself.
Similarly, you can say "Mark is humble", but you cannot have
"humility" as a concrete thing that exists independently of something.

        There is nothing wrong with "abstractions": They are inherent
in all claim drafting, whether it's software, hardware, chemical,
biological, etc. "Mammal" is an abstraction over humans, dogs,
dolphins, and horses, but it still is a 'concrete' concept.
        Friday at 6:40pm · Like

        Seth Johnson A logic processor is a device designed to be
general purpose in the most general sense, by providing for the
performing of a set of operations that covers all logical operations.
If the bit operations the processor "does" can be used to represent
any logical algorithm, it is a device that supports pure abstraction.
(That was me you referenced, not Mark) Yes, "pure abstraction" is not
even attached to a logic processor. But a logic processor is capable
of serving to represent pure abstraction -- and then automate it (thus
"do" it), just as you can write logic with pen and paper.
        Friday at 6:49pm · Like

        Seth Johnson And your definition of pure abstraction is not
the pure abstraction that is generally understood as associated with
logic. Logic isn't about any qualia like redness or roundness. Things
like Virtue or Justice might be ideal by philosophical insight, but
typically are related to empirical referents of some sort by most
folks you're going to talk to, and here I'm talking about the pure
abstraction that logic and math represent.
        Friday at 7:00pm · Like

        Robert Sachs · Friends with Joshua Sarnoff and 2 others
        What does it mean to "support pure abstraction"? The only
'device' that support pure abstraction is your brain. What the logic
processor does is process electrical signals. Under this analysis, the
electrical signal that control my toaster are pure abstraction too
because all they do is measure time and temperature and cause the
toast to pop up. Or are you drawing a distinction between analog and
digital signals? Or discrete circuits and ICs or CPUs? And given that
you say a logic process is designed to be "general purpose" then do
you consider algorithms on FPGA and and DSPs to be something not a
pure abstraction?
        Friday at 7:02pm · Like

        Seth Johnson You're uselessly failing to distinguish
representation from what's represented. The logic processor supports
pure abstraction by representing it and executing it. It counts as a
logic processor if it can do that for any logical algorithm. If you
want to say a physical implementation that does that doesn't represent
logic, you're only kidding yourself. My brain does logic. I do logic
with symbols on paper. The formulas I write do logic, because they
communicate not just to me, but to anybody who uses the same symbols.
        Friday at 7:07pm · Like

        Robert Sachs · Friends with Joshua Sarnoff and 2 others
        By "logical abstraction" do you mean stuff like predicate
logic representations, or similar symbolic manipulations? In that
case, you are conflating three levels of analysis into one. The top
level if you will is the real world solution/functional model (e.g,
computing network bandwidth, or determining optimal bid price for an
ad), the second level is the symbolic representation (e.g.,
algorithmic, formula, pseudo code, etc.) and third level is the
machine level presentation (binary representation, and specific RTL or
similar 'semantic' free). In many cases, the "invention", the thing
that has commercial and functional value is at the top level. The
implementation (2nd and 3rd level) are variable and not the locus of
invention. So yes, a purely symbolic set of instructions that are
divorced from their real world semantics can be said to be pure
abstraction "supported" by a logic device. So constrained, that's a
very limited conclusion, and certainly does not speak what counts as
an abstract idea in regards to boundaries of patent eligibility
generally, or even in software.
        Friday at 7:12pm · Like

        Robert Sachs · Friends with Joshua Sarnoff and 2 others
        "You're uselessly failing to distinguish representation from
what's represented. The logic processor supports pure abstraction by
representing it and executing it." I think not. Example: If A then B.
A. Therefore B. That's pure logical abstraction regardless of what A
and B "represent". There is a distinction between the form of
representation (symbols and their syntax) from what they represent
(their semantics, meaning). This distinction is recognized by
linguists, philosophers, computer scientists, etc. Perhaps I'm not
eloquent enough to make it clear, then that is my failing.
        Friday at 7:17pm · Like

        Seth Johnson Nope, it's very generally applicable. It very
simply explains why the software in a patentable device does not
qualify for patentability. Though I would be more specific in saying
that what makes the machine level a logic processor (or software,
whichever way you look at it) is that it is capable of representing
and executing any logical algorithm.
        Friday at 7:18pm · Like

        Robert Sachs · Friends with Joshua Sarnoff and 2 others
        "It very simply explains why the software in a patentable
device does not qualify for patentability" So your position is that
'software' based inventions are never patentable?
        Friday at 7:20pm · Like

        Seth Johnson Um, I think you are saying what I am saying, just
failing to apply it.
        Friday at 7:21pm · Like

        Seth Johnson See above. Inventions that use software could be
patentable. Just not the software in them.
        Friday at 7:22pm · Like

        Robert Sachs · Friends with Joshua Sarnoff and 2 others
        My apologies for failing to apply this. I think we're
addressing different parts of the same, or perhaps I misunderstood
your position. But I'll persist in my thickheadness. If by "inventions
that use software" you mean a series of functional operations
independent of their software code implementation, then I certainly
agree that's patentable stuff. However, I also hold the view that
particular software logic itself (not bound to a specific language,
but the formal operations) is also potentially patentable. In other
words, even if we don't care what the bits mean, the particular way of
manipulation the bits/data/symbols can be an invention subject to
        Friday at 7:38pm · Like

        Seth Johnson I would simply say the presence of the logic
processor that those instructions are being provided to, disqualifies
them. A very simple way to identify when stuff is being used to
represent pure logical instructions, and thus exclude pure logic from
        Friday at 7:41pm · Edited · Like

        Seth Johnson A definition of software, as those instructions
that are pure logic because they are being used by a logic processor.
        Friday at 7:42pm · Like

        Edward Reines · Friends with Mark Lemley
        I suggest investing these ideas in a legislative effort rather
than via proposed judicial interpretations of Section 101. Old school.
        Friday at 9:40pm · Like

        Seth Johnson My usual effort is to recommend this as a base
for legislation. It has the benefit of fitting right in with
long-recognized traditional principles. Just makes it an easy formula.
There's certainly long been a lot of confusion and debate on the
matter of software patents; this cuts through it all.
        Friday at 9:46pm · Edited · Like

</Haphazard insert point>

> Groklaw has rightfully identified that Lemley's solution is not
> sufficient, the most important point in their answer to USPTO
> consultation is section C) :
> <http://www.groklaw.net/article.php?story=20130131141515562>
> This is what I've written some months ago to Julie Samuels from EFF:
> * CLS Bank v. Alice Amicus Brief
> I fully support the first amicus brief from September 2012 requiring a
> re-hearing. The problem is with the December amicus brief where EFF
> endorses, too much IMHO, Professor Lemley's solution based on means-plus
> function.
> First, there is some kind of a contradiction between insisting on the
> "use of Section 112(f) as a vehicle to analyze and resolve potentially
> abstract claims" and the refusal to "substitute §§ 102, 103, and 112
> inquiries for the better established inquiry under § 101", as expressed
> in Mayo v. Prometheus, which is underlined in EFF amicus brief from
> September. This may be important for what CAFC may conclude from amicus
> briefs, but this is not the main point (and anyway I really doubt that
> anything good can come from CAFC).
> The main point is that Lemley's proposal would be going in the right
> direction if it was not based on a false assumption: that some software
> can be patented. Reducing the number of granted software patents and the
> unclear boundaries thereof by forcing to disclose algorithm implementing
> the claimed function is not THE solution to the software patents
> problem, contrary to what Mark Lemley voices in his paper. There are
> software patents with clear boundaries and disclosed algorithm which
> nevertheless impede innovation and economy. Just think about infamous
> GIF/LZW patents. They actually included quite defined algorithms, IMHO
> with enough details to pass Section 112(f). Nevertheless, there is no
> need to recall harms caused by these patents.
> This is really worrying because, we've already had these discussions in
> Europe in the beginning of the century. We have identified that
> disclosure of algorithm was only a partial solution which don't really
> help with the bigger problem of software patents.
> Here again I'm worried that the momentum in US could be wasted by bad
> partial solutions. Just read 2011 Federal Register notice
> (https://www.federalregister.gov/articles/2011/02/09/2011-2841/supplementary-examination-guidelines-for-determining-compliance-with-35-usc-112-and-for-treatment-of))
> on "Computer-Implemented Means-Plus-Function Limitations". They are
> advising disclosure of algorithm "in any understandable terms including
> as a mathematical formula […]". Hey, don't you have a Supreme Court
> having ruled that mathematical formula cannot be patented, as being an
> abstract idea?
> I think this is the way to go: exclude software patent for the inherent
> abstractness of software. This is the way that was debated and chosen in
> Europe (with "non-technical" as European counterpart for "abstract").
> And this is  actually what you seemed to support in EFF amicus brief
> from September. And even in the final December amicus brief, you've
> realized that § 112(f) test is eventually defined as a coarse filter to
> determine abstractness under § 101. So go for abstractness!
> --
> Gérald Sédrati-Dinet
> http://www.unitary-patent.eu/      http://www.april.org/
> http://www.brevet-unitaire.eu/     http://laquadrature.net/
> _______________________________________________
> A2k mailing list
> A2k at lists.keionline.org
> http://lists.keionline.org/mailman/listinfo/a2k_lists.keionline.org

More information about the A2k mailing list