Weekly edition Kernel Security Distributions Contact Us Search Archives Calendar Subscribe Write for LWN LWN.net FAQ Sponsors

# Study for US Congress outlines options against patent trolls (The H)

## Study for US Congress outlines options against patent trolls (The H)

Posted Sep 7, 2012 19:53 UTC (Fri) by Wol (guest, #4433)
Parent article: Study for US Congress outlines options against patent trolls (The H)

Just accept that a simple definition of software is "if it's a big number supplied on a transfer medium, then it has to be maths".

Maths isn't patentable.

End of problem.

Of course, you then need to clarify that putting that number on to a general purpose computer doesn't make a new patentable machine, but that's just common sense (of course, common sense doesn't seem to be much in evidence here :-(

Cheers,
Wol

Study for US Congress outlines options against patent trolls (The H)

Posted Sep 7, 2012 20:43 UTC (Fri) by paulj (subscriber, #341) [Link]

Does that make instructions for how to build some machine maths?

Study for US Congress outlines options against patent trolls (The H)

Posted Sep 11, 2012 13:13 UTC (Tue) by Wol (guest, #4433) [Link]

What if it does? The *instructions* can be maths, they're protectable by copyright, so who cares?

The *machine* remains patentable by copyright.

This is exactly the confusion the "let's patent software" crowd are using to try and get software patents.

Cheers,
Wol

Study for US Congress outlines options against patent trolls (The H)

Posted Sep 11, 2012 13:26 UTC (Tue) by paulj (subscriber, #341) [Link]

Not sure what you mean by "patentable by copyright". Given the previous sentence, I'm assuming "protectable by patents" is what you had intended there. In which case:

So then the instructions are not patentable (but copyrightable), while the instructions + a machine is patentable. And guess what, this is *exactly* how software patents work in many parts of Europe, the patent claims are made for the combination of some set of instructions and a machine (e.g. a general purpose computer).

Study for US Congress outlines options against patent trolls (The H)

Posted Sep 12, 2012 23:22 UTC (Wed) by Wol (guest, #4433) [Link]

Which is why my original comment said "of course adding this big number to a general purpose machine does not make a new machine".

Following a list of instructions "with a machine" isn't patentable. Doesn't stop the patent offices issuing patents, though :-(

Cheers,
Wol

Study for US Congress outlines options against patent trolls (The H)

Posted Sep 7, 2012 22:40 UTC (Fri) by yokem_55 (subscriber, #10498) [Link]

What about implementing an algorithm in a hardware, fixed function ASIC? Then what happens when you virtualize that ASIC? Where exactly is the line between patentable hardware and unpatentable math? I find software patents as repugnant as anyone else around here, but blindly equating software to unpatentable math oversimplifies the situation...

Study for US Congress outlines options against patent trolls (The H)

Posted Sep 7, 2012 23:07 UTC (Fri) by dlang (✭ supporter ✭, #313) [Link]

Math algorithms are not supposed to be patented.

what algorithms are you implementing in hardware? do they boil down to doing a series of calculations? (which they will for digital logic). If so the ASIC should not be able to be patented.

you could copyright it, but there had better be some creativity in what you do, not just mechanical layout based on high-level description (although any non-trivial high-level description can be copyrighted)

It doesn't matter if the implemenation is in circuits or in software.

just like combining words into a book is copyrightable, but not patentable.

If you create a new type of chip feature (i.e. a flash memory cell, phase change memory cell, new type of transister, gate, etc), that feature is patentable, but not the functionality of the ASIC or the algorithm that it implements.

ASIC simulation and patents

Posted Sep 19, 2012 0:48 UTC (Wed) by Max.Hyre (subscriber, #1054) [Link]

I think the HW engineer is (or should be) simply out of luck on that one. I offered elsewhere a way to distinguish SW for purposes of excluding it from patent system, and mentioned that exact case. Simplifying, I think if the object under discussion (not instructions for it, but the object itself) can be sent over a data link, it shouldn't be patentable.

Study for US Congress outlines options against patent trolls (The H)

Posted Sep 9, 2012 0:45 UTC (Sun) by ssmith32 (subscriber, #72404) [Link]

I once was asked in an interview how to calculate the square root of a number. I said Newton's method. He said "no, that's maths, how do you solve it with a computer?" If an engineer at a large "hot" tech firm has that much trouble with the concept, I think it will be a while before people understand that algorithms are math. (fyi Newton's method is like a four line for loop...)

Study for US Congress outlines options against patent trolls (The H)

Posted Sep 10, 2012 18:54 UTC (Mon) by k8to (subscriber, #15413) [Link]

Of course, they were probably trying to evaluate if you knew how to use the standard math packages.

Study for US Congress outlines options against patent trolls (The H)

Posted Sep 10, 2012 21:36 UTC (Mon) by ssmith32 (subscriber, #72404) [Link]

lol. I wish! No, this was a standard tech interview - doing things the easy way was explicitly off limits.. he wanted me to do a binary search for the square root.

Study for US Congress outlines options against patent trolls (The H)

Posted Sep 10, 2012 22:24 UTC (Mon) by k8to (subscriber, #15413) [Link]

oh.

:-(

Well, at least a domain-bounded bisection search is traversing a monotonically increasing continuous function. So it works even if it's pointlessly slow.

Still, folks who want you to implement the wrong algorithm. Uh, bad sign.

Study for US Congress outlines options against patent trolls (The H)

Posted Sep 11, 2012 8:08 UTC (Tue) by ssmith32 (subscriber, #72404) [Link]

Well, I shouldn't judge too harshly*: I'm no math genius, and have certainly written my share of sub-optimal (but clear and finishes in time!) code :)

The troubling part for me is the weird insistence, that there was a math way and an engineering way, isn't really that weird: I've ran into in other places as well. And people feel surprisingly strongly about it.

I admit that professional training may affect how you approach a problem, but when you start approaching the distinction between the act of coding and mathematical algorithms as some sort of fundamental difference...hmm.. So it's an uphill journey.

*Although I did have a little chuckle a year or so later when it turned out whoever managed their IPO had trouble with the maths as well :D

Study for US Congress outlines options against patent trolls (The H)

Posted Sep 11, 2012 11:21 UTC (Tue) by ekj (guest, #1524) [Link]

Sometimes that's true: that there's a math way and an engineering-way and the two are fundamentally different.

For example, simply assuming that all 256 bit (genuinely!) random bitstrings will be different from each other is sound engineering, but invalid math. (the engineeer might worry if the bitstrings are *really* genuinely random, though)

If there's 1:2^80 odds that a cosmic ray will strike your hardware and flip a crucial bit, but a 1:2^128 odds that two random strings will collide by accident, the engineer worries most about the first possibility, while the mathematician worries about the second possibility.

Study for US Congress outlines options against patent trolls (The H)

Posted Sep 9, 2012 7:41 UTC (Sun) by Seegras (subscriber, #20463) [Link]

> Of course, you then need to clarify that putting that number on
> to a general purpose computer doesn't make a new patentable machine,

Oh, the famous "Moby Dick Support Device" http://seegras.discordia.ch/Blog/the-moby-dick-support-de...

I still can't understand the logic behind that (and the fact that courts and lawyers use a definition of "algorithm" in patent lawsuits which is completely at odds with reality -- I mean you can't just legislate Pi to be exactly 3, can't you?).

Study for US Congress outlines options against patent trolls (The H)

Posted Sep 9, 2012 22:43 UTC (Sun) by shmget (subscriber, #58347) [Link]

"I mean you can't just legislate Pi to be exactly 3, can't you?)."

In the US ?, sure you can try.. that has been attempted before:
http://en.wikipedia.org/wiki/Indiana_Pi_Bill