[seibelj]

What software is even patentable anymore? Some light reading of articles and wikipedia[0] makes it appear that post-Alice Supreme Court ruling anything mathematical, or anything that is an abstract idea implemented by a computer, is not patentable. I'm struggling to think of what is left. Cryptography is pure math, as is data compression. Is a specific type of UI patentable?

[0] https://en.wikipedia.org/wiki/Software_patents_under_United_...

[rayiner]

Alice/Mayo are more complicated than that. Those cases lay out a two-step process:

(1) Is the patent directed at an abstract idea or law of nature? If not, it is subject-matter eligible.

(2) If the patent is directed at an abstract idea, does the patent add any "inventive concept" so that the patent isn't just an attempt to monopolize the abstract idea itself? For purposes of this second step, simply performing the abstract idea on a computer is insufficient. If the patent adds something extra, it's subject-matter eligible.

"Abstract idea" almost certainly means something narrower to the courts than it does to many programmers. The Supreme Court declined the opportunity in Alice to rule that all algorithms are abstract ideas that are patent ineligible.

Say I come up with a circuit that controls engine timing to minimize emissions. If I use an electronic circuit, that's definitely patentable. Under broad definitions of "abstract idea," if I replace the circuit with a general purpose processor, then suddenly that's not patentable. The Supreme Court has tried very hard to avoid that result. Something isn't patent eligible just because you run it on a computer, but something also isn't patent-ineligible just because it runs on a computer.

[seibelj]

Thanks for your detailed response. I'm still confused unfortunately. So my algorithm needs something more that is subject-matter specific. As a contrived example, let's say I take the quicksort algorithm. If I add something extra, like quicksort applied to data objects tagged in a very specific way that I design, that could then be a patent?

[rayiner]

To be fair, the Supreme Court's standard is pretty confusing!

As to your example, I suspect he answer is still "no." It's not enough to just "add something" to the abstract idea--the thing you add has to have an "inventive concept." Field of use limitations ("quicksort to sort birthdays") or post-solution activity ("quicksort then you do something else") do not make the abstract idea patentable. Routine activity--and arguably, adapting algorithms to different data formats is routine in programming--aren't sufficient either.

One thing that's left unsaid in the cases but that matters in practice is that the patent has to "seem like a real invention" that produces a useful result. Courts deny that they do this, but post-Alice there has been a tendency to inject elements of obviousness and novelty into the issue of patentability. If your algorithm doesn't seem like it was hard to come up with, or it just produces numbers that are not directly useful to the end user, it's going to get labeled "abstract idea."

E.g. consider an algorithm for distributed reference counting. That'll get dinged as an abstract idea. However, if you come up with a way to manage image/video resources in a distributed cache that happens to use a distributed reference counting algorithm, that might pass muster.

Finally, the law has changed a lot in the last 6-7 years. But there are still lots of patents out there from back when the law was laxer that are getting litigated (and often invalidated in light of Alice).

[kirrent]

If you weren't confused then something would be wrong. People in the industry get confused by this subject matter eligibility stuff. Alice's tests are pretty vague and leave a lot up to individual judgement on the merits of each case. It's hard to give hard and fast rules.

That said, it's hard to say that any other jurisdiction has a far simpler or better test for software implemented inventions. It's just a hard question.

[_rpd]

To put it mildly, the courts are still working it out. For example ...

Federal Circuit: Software and Data Structures Are Not Inherently Abstract

http://patentlyo.com/patent/2016/05/structures-inherently-ab...

[sdabdoub]

One of the main issues is, despite that decision, the USPTO is understaffed and underfunded, so stuff will still get patented that should have no grounds for patentability.

[sosborn]

That doesn't make sense to me. If they are understaffed and underfunded then you would expect the number of granted patents to decrease.

[philipkglass]

I think the reasoning is that understaffing means each examiner can't spend as much time looking for prior art, so things might be patented that should have been rejected by a lengthier prior art search.

[pbhjpbhj]

The USPTO would still have an obligation to process the patent applications though. So now each patent has an average of 4h20m spent on it instead of 4h40m (some are hundreds of pages of complex concepts that potentially no one in the world has seen published before).

The examiner then would be under pressure to glance over a lot of applications and just let the courts sort it out later.