> What is a "software patent" or a "patent troll?"
Defining a software patent is not hard. A patent is a software patent if you can infringe it entirely with software. Software can't infringe a patent for a cholesterol drug or a light bulb, so those aren't software patents.
> I look at the tech industry, and see one where product companies rather than hard R&D companies make all the money. The folks at SRI who invented Siri get a small license fee, while Apple makes billions by wrapping it with a candy shell.
There is so much more to a modern smartphone than just Siri that the idea that SRI should be getting any significant fraction of the revenue is a little disingenuous. The "problem" with software is that you can have ten thousand discrete inventions that each individually would have been a revolution in 1995 but the market still expects you to package them all together and sell the device for around $600, and then sell a device with all the same software but less powerful hardware for $150. After the costs of hardware design, manufacturing, marketing, distribution, administration, etc., plus the cost of writing the software to actually implement tens of thousands of "inventions", you would be lucky to give each inventor a penny per device.
But the real problem is you can't accurately price the contribution each inventor makes. There are too many of them that are each worth such a small percentage of the total that spending any kind of significant resources to evaluate their individual value isn't cost effective. But if you can't price patents accurately then you don't have a functioning market, which is fundamentally why we end up with such poor quality patents in software -- if the bad ones go on the pile with the good ones and they all get a similar payout then it's a lot easier to collect rent by filing low quality patents than actually spending money on useful R&D.
> It's an industry where there is tremendous pressure to vertically integrate, because you can only justify R&D on basic technologies if you also control the product that gets sold to end-users.
I don't see what vertical integration has to do with it. Microsoft circa 2002 had every incentive to invest in operating systems R&D, and they did, but they didn't need to sell the hardware or control the market for third party applications to do it.
If you're actually selling software to users you're competing against the previous version of your own software as much as you are against competitors anyway. And the patent can't convince users to upgrade once they already have a version that implements it. If anything software patents are contributing to the vertical integration because a vendor that can establish a monopoly or otherwise strong market position using patents (and any combination of other tactics) can use it to leverage that market position into control of vertical markets.
If you want less vertical integration then what you want is more modular systems, open standards and free software. Which is the opposite of what software patents do.
> Defining a software patent is not hard. A patent is a software patent if you can infringe it entirely with software. Software can't infringe a patent for a cholesterol drug or a light bulb, so those aren't software patents.
Presumably, by "entirely with software" you mean "entirely with software running on a [hardware] comupter." And that's why so many companies oppose eliminating "software patents" even though it seems like a no-brainer to many.[1] We're not just talking about your typical troll patent, but e.g. techniques for processing signals from MRI machines. We're talking about the software behind self-driving cars.
> But if you can't price patents accurately then you don't have a functioning market, which is fundamentally why we end up with such poor quality patents in software -
Your right, this is a huge problem. It would be a lot better to have fewer, stronger patents with well-defined boundaries.
> If you want less vertical integration then what you want is more modular systems, open standards and free software. Which is the opposite of what software patents do.
Your two examples support my point. I think it's underestimated how much we owe the open PC ecosystem to the fact that Microsoft could rely on strong copyright, enabling them to make money selling an OS. If it wasn't practical to protect the OS as a separate product, they would have had a tremendous incentive to get into the PC business.[2]
As for open standards, most of our hardware standards are built on patent pools. They enable companies to cooperate to build standards, while keeping out freeloaders.
[2] We have more open source than ever in the hands of ordinary consumers, but only because software has taken a backseat to hardware and services that can be more easily monetized without the worry of piracy. Everyone uses web browsers, but there's no money in building web browsers. Instead, the two major engines are developed by a cell phone company and an advertising company.
> Presumably, by "entirely with software" you mean "entirely with software running on a [hardware] comupter."
A [hardware] computer that exists in the prior art, yes.
> And that's why so many companies oppose eliminating "software patents" even though it seems like a no-brainer to many. [1] We're not just talking about your typical troll patent, but e.g. techniques for processing signals from MRI machines. We're talking about the software behind self-driving cars.
I don't get the "that's why" at all. If you come up with new techniques for processing signals from MRI machines that it would be practical for a doctor to calculate by hand on paper then you can't patent them but presumably the same people would still like to. The use of a computer isn't why anyone would want to patent such things, it's just a conceit to circumvent the unpatentability of math. Any sufficiently fast general purpose computer of any method of operation whatsoever could stand in as the "hardware" for any software patent, because you're not patenting anything to do with the hardware.
It doesn't do anything to say that math done by a computer should be patentable. If you can do math you can do it with a computer. What you're really getting at is that you think some math should be patentable but not all math, without articulating any kind of reasoned dividing line between the two. Which is how we get here:
> Your right, this is a huge problem. It would be a lot better to have fewer, stronger patents with well-defined boundaries.
Certainly it would. But how?
The problem isn't a regulatory one, it's an economic one. Even if you exclude all of the existing software patents that should not have been issued because the claims are excessively broad or vague or claim material in the prior art, the remaining patents still don't all have the same value, but there would still be so many that they couldn't be accurately priced cost-effectively at scale. And having to pay a hundred times what a patent is worth because valuing it is too expensive has essentially the same consequences as having to pay for a patent that shouldn't have been issued because invalidating it is too expensive. It provides the same incentive and ability to file low value/low quality patents and then use them to collect undue rents from everyone.
And even that's assuming some hand wavy magic solution to eliminating all the improvidently granted software patents, which it isn't at all clear how to do cost-effectively in practice.
The choice we seem to have is between no software patents and prolific low quality software patents. Given that choice I think no software patents wins.
> I think it's underestimated how much we owe the open PC ecosystem to the fact that Microsoft could rely on strong copyright, enabling them to make money selling an OS.
Which is why software patents are unnecessary. You can do well enough with just copyright, not least because the cost of writing and testing software is generally much more expensive than coming up with the idea of what software to write. Copyright protects the thing that required most of the resources to create. Someone who wants to copy it has to duplicate all of that effort, which takes time and inhibits freeloading, and in the meantime the original creator enjoys first to market advantage and builds network effects.
Meanwhile someone who does copy the general idea is forced to present their own take on it, which in practice means making their own improvements and advancing the status quo. Android obviously copied much from the iPhone, but they also made improvements that Apple then copied back. Then Apple has to make further improvements of their own instead of resting on their laurels if they want to stay in the game. Copyright for software inhibits freeloading just enough without eliminating competition, so you get the best of both worlds, and hence the rapid rate of progress we've seen.
Having patents on top of copyright throws all of that out of balance. It gives too much advantage to incumbents and large entities. Just copyright and the network effects inherent to software have more than enough propensity to create monopolists as long as the monopolist doesn't fall asleep and allow a new competitor to surpass them. Software patents allow the incumbents to prevent even that by maintaining a patent thicket around the market to wall out non-incumbent players.
Defining a software patent is not hard. A patent is a software patent if you can infringe it entirely with software. Software can't infringe a patent for a cholesterol drug or a light bulb, so those aren't software patents.
> I look at the tech industry, and see one where product companies rather than hard R&D companies make all the money. The folks at SRI who invented Siri get a small license fee, while Apple makes billions by wrapping it with a candy shell.
There is so much more to a modern smartphone than just Siri that the idea that SRI should be getting any significant fraction of the revenue is a little disingenuous. The "problem" with software is that you can have ten thousand discrete inventions that each individually would have been a revolution in 1995 but the market still expects you to package them all together and sell the device for around $600, and then sell a device with all the same software but less powerful hardware for $150. After the costs of hardware design, manufacturing, marketing, distribution, administration, etc., plus the cost of writing the software to actually implement tens of thousands of "inventions", you would be lucky to give each inventor a penny per device.
But the real problem is you can't accurately price the contribution each inventor makes. There are too many of them that are each worth such a small percentage of the total that spending any kind of significant resources to evaluate their individual value isn't cost effective. But if you can't price patents accurately then you don't have a functioning market, which is fundamentally why we end up with such poor quality patents in software -- if the bad ones go on the pile with the good ones and they all get a similar payout then it's a lot easier to collect rent by filing low quality patents than actually spending money on useful R&D.
> It's an industry where there is tremendous pressure to vertically integrate, because you can only justify R&D on basic technologies if you also control the product that gets sold to end-users.
I don't see what vertical integration has to do with it. Microsoft circa 2002 had every incentive to invest in operating systems R&D, and they did, but they didn't need to sell the hardware or control the market for third party applications to do it.
If you're actually selling software to users you're competing against the previous version of your own software as much as you are against competitors anyway. And the patent can't convince users to upgrade once they already have a version that implements it. If anything software patents are contributing to the vertical integration because a vendor that can establish a monopoly or otherwise strong market position using patents (and any combination of other tactics) can use it to leverage that market position into control of vertical markets.
If you want less vertical integration then what you want is more modular systems, open standards and free software. Which is the opposite of what software patents do.