Software patens ever work?

[The Happy Friar]

Inspired by Admin's post in shoutbox: could software patents ever work to the benefit of creators and consumers the same way a patent on a physical object does?

The way a patent is supposed to work is that someone invents something (a process, physical object, etc). and files a patent to protect their control over the something they invented for a set period of time (~19 years but it can varie).  Example: I invent a steering wheel.  I have to write up what it does, how it does things, any other patent it's based on and give a working diagram of how it works.  You can plainly see what I invented and from the information I provided create your own.  Another example: I invent a process to turn lead in to gold.  I write up my chemical formulas, what's in it, what it does and what the results should be.  You can plainly see what I invented and create your own.

The tradeoff for the protection is that I hand over all my information so that anyone can look it up at any time and when the patent is expired anyone can use the information I came up with to create the same thing.

Now comes software patents: I believe they're under the process category.  You can't have a blow up diagram like the steering wheel.  Instead they say what they do, how they do it and give you some give diagrams on how it works but you're lacking the important thing you gave up in the previous two examples: the actual way to duplicate what was created.  If I made a steering wheel out of a piece of metal that I used a heat source to bend, welded a crossbar on it and said that was my steering wheel, you could make one that was oval and didn't have a crossbar but instead was filled in with steel, those wouldn't be the same thing, both could be patented and both wouldn't "step on the toes" of the other.

With a software patent you give abstract terms that are compatible across various programming languages.  You're not specific.  It's broad enough where two different ways to handle something can be under the same patent due to the abstract terms (think of patenting a loop: you say "it goes back and forth between set points and runs specific code".  Well, that covers for/while in most languages and goto in basic, two completely different things but you don't need to be specific enough to say what you're specifically patenting).

My solution is to stop treating software patents like a process and treat them like a physical item: you must post your code you're patenting for all to see, no more "ha! you are getting sued because you did X Y & Z", it would have to be a duplicate or close enough that they would be considered identical in design. 

But wait, you don't want to give out your code, right (assuming closed source here)?  Because then someone could steal it!  Well, that already happens, but also look again at physical patents: I can take apart a physical object or I can break down a physical process.  It's ILLEGAL to do that to code without express permission.  There's the "right of first sale" law and there's already laws going on the books that car manufactures can't withhold car computer information from non-dealer fix-it shops, do the same for software.  Make it a "right" to decompile just like it's my "right" to take apart my drill and see how it works or watch how an assembly line puts something together.   The reason patents exist is because anyone could duplicate something w/o permission so a patent was designed to stop that and give back to the public after the inventor got a reward for their work.

But wait, then my company will loose lots of valuable IP!  Ok, then lets go the other way: food recipes aren't allowed patents.  NOBODY knows Coke's recipe except the people who make it and it hasn't been stolen.  Why?  Since it's not patentable it doesn't NEED to be shared and contracts can be formed to stop employees from using that information because it's not in the public (like patents).    It's the other trade off: no legal protection if someone who gets the info (ie Pepsi), just legal protection from someone giving it away (CEO under contract to not share the info).

Software doesn't need to be treated any differently then any other physical good out there, it's no different then what has been around for centuries. 

From the link posted in the shoutbox: Anyone who thinks they have a unique idea that they want to “own” and milk for money can do so—but first they have to track down and appropriately compensate all the people who made possible the compilers, algorithms, programming courses, books, hardware, and so forth that put them in a position to have their brainstorm.

That doesn't happen with physical patents unless you use something already patented (ie add something to an alarm clock, for example).  Also, several of those things listed are already compensated: the purchase of the compiler, the courses, books, hardware & so on.  Just like if I used an arc welder to make my steering wheel, I already compensated the creator of the welder.   Patents already cover some of this too: you have X years (I think 2) that your patentable item can be in the public until it becomes public domain.   If you don't apply for a patent in those X years you never get one.  So once again, if you don't want to patent then don't share the details, if you do, share them for all to see.

[kren.Z]

I don't know much about patents, but I think the technology sector works differently. Companies/people usually develop technology to better serve a core business, rather than develop technology as an end in itself.

You can't refute that Google is one of the best technology and engineering companies on the planet, but they make their money by serving you ads.

MapReduce (http://en.wikipedia.org/wiki/MapReduce) and the GFS (http://en.wikipedia.org/wiki/Google_File_System) is a software framework developed internally by Google to support its own distributed computing needs. Hadoop (http://en.wikipedia.org/wiki/Apache_Hadoop) was derived from the research papers. This is extremely important and useful technology. Tons of companies need to build their own infrastructure for distributed computing, not necessarily for search

All of this was built to serve you ads more efficiently. AdSense is patented though (http://appft1.uspto.gov/netacgi/nph-Parser?Sect1=PTO2&Sect2=HITOFF&p=1&u=%2Fnetahtml%2FPTO%2Fsearch-bool.html&r=1&f=G&l=50&co1=AND&d=PG01&s1=20040059708&OS=20040059708&RS=20040059708)


...but you get gems like this from google. http://google-opensource.blogspot.com/2012/05/introducing-ceres-solver-nonlinear.html . THIS technology has implications in a WIDE RANGE of areas.


C and UNIX were born in AT&T's research labs. Why is linus not a billionaire despite the linux kernel being used EVERYWHERE (
http://www.youtube.com/watch?v=yVpbFMhOAwE)?




..i don't know just rambling...

[quadz]

With a software patent you give abstract terms that are compatible across various programming languages.  You're not specific.  It's broad enough where two different ways to handle something can be under the same patent due to the abstract terms (think of patenting a loop: you say "it goes back and forth between set points and runs specific code".  Well, that covers for/while in most languages and goto in basic, two completely different things but you don't need to be specific enough to say what you're specifically patenting).

My solution is to stop treating software patents like a process and treat them like a physical item: you must post your code you're patenting for all to see, no more "ha! you are getting sued because you did X Y & Z", it would have to be a duplicate or close enough that they would be considered identical in design. 

We do have a system that works that way, however: Copyright.  I'll explain in more detail, below.  First a little bit of background:

A company I've coded for was recently successfully sued by a patent troll.  Which is to say, sued by a nest of lawyers in the shape of an entity that never invented anything, and which neither sells nor licenses any product, but which merely has intellectual property in the form of software patents it has acquired.

So, thus far, I'm with you in the sense that indeed, they had no 'physical' item they were protecting.

What they sued over was a subsystem we had developed over the course of many years.  Our system started out simple, and we'd add capability to it every so often, as needed.

Eventually--even though at each step along the way, all we had done was make simple obvious improvements to our own system--at some point they decided that the totality of our subsystem's conceptual process infringed on their patent.  They wanted royalties covering the last few years of our software sales.

When in fact, one of the stipulations of the eventual out-of-court settlement was that they acknowledged that while we would now be a licensee of their patented process, our software is not even using the process covered by the license!  The whole thing was a surprisingly naked form of extortion: Pay up, because even though you're not really using our process, what you're doing is close enough that we could convince a judge and jury in a particular patent-lawsuit-friendly county of TX that you are infringing, and it's going to cost you a lot more if we go that route.

So now we have an expensive license to a process we don't even really use.  (Not to mention the entire subsystem in question is not even central to the functioning of our product, but is essentially a component of the install procedure!)

But, back to the idea of a physical product.

It does indeed sound like a step in the right direction to require that patent trolls of the sort that bullied us actually have some physical implementation of their process that they're licensing.

Trouble is, it's actually relatively difficult to implement a useful general purpose software library.

And whether or not their implementation is indeed useful is of prime importance.  Because it's really easy to write up some b.s. pseudo-code implementation of a process--this is essentially what the patent application is; and even if it were required to be implemented in "real" code, it's still generally immensely easier to write if it merely has to work as a proof of concept but not be production-worthy code.

So if their implementation isn't required to be useful, then it might as well just be pseudo-code for all the good it would do the licensees who will need to do the hard work of writing the production quality implementation themselves.

OK, so let's say we require the patent holder's implementation must be useful for it to be protected.  But how do we determine whether their implementation of their process is actually useful to anyone?

It turns out we already have a solution.  They copyright their implementation, and sell licenses (whether in source or binary form; their choice.)  And if it's useful, developers purchase licenses!  But if the offered library isn't useful, developers purchase no license and write their own implementation instead (which is indeed where the hard work is.)

Because unlike with software patents, software copyright is actually protecting the part where the heavy lifting occurs: the production of a useful implementation of an idea.

So it seems to me if one tries to improve the software patent situation by adding the requirement for a 'physical' implementation of the process, then the utility of said implementation must be a factor; and it seems that to gauge or ensure this in any meaningful way, we end up reinventing copyright style licensing.


Regards,

quadz

[quadz]

C and UNIX were born in AT&T's research labs. Why is linus not a billionaire despite the linux kernel being used EVERYWHERE

A related question might be, why did the commercial Unix offerings available at the time Linux was conceived, such as Xenix and Minix, not become popular?

It would seem the set of all software is greater than the set of software that can be successfully monetized.

[kren.Z]

you're right. I don't know much about the mainframe era and i take it UNIX wasn't cheap. Some cool and useful tech came out of Bell Labs though.

[QwazyWabbit]

UNIX came out of Bell Labs but they weren't a software company. UNIX was for running telephone switches. They had a narrow focus on their tools so they missed the market completely. It also lacked the pretty GUI and still suffers from cryptic commands to do common functions.

IIRC, in 1997, SCO UNIX for the 80186/80286 was $495 in a huge box set of some 5 boxes and manuals when MS-DOS was $135 in a single box. A huge blunder for SCO, they could have swept the desktop clean of Microsoft forever if they hadn't been so damned proud of a package they didn't even write themselves.

Several of my friends had HP9000's running HP's version of UNIX III called HP/UX.