> This is intrinsically unfair to the patent-owners with large portfolios, because no matter how many patents are being infringed, they only get to assert a few,
I agree with that. That is one of the symptoms of the disease. It is preposterous for every competitor to be infringing thousands of patents held by every other competitor. That is a pathological outcome that doesn't commonly occur in non-software industries. But we have to address the disease, not the symptom. How do we get to a world where the large majority of patent infringement in the software industry is a result of intentional infringement rather than independent invention or a need for compatibility with existing software? The easy solution is for software to be unpatentable. If you don't like that, what's your alternative? (And if the solution involves improving software patent quality, please explain how and also what to do about all the low quality patents already issued.)
> and even if they prevail, damages are limited to the patents asserted.
The other side of this coin is that juries are extremely arbitrary when evaluating the value of a patent, and in the case of software patents will often vastly overestimate their value. There are easily thousands of features the lack of which would materially reduce market demand for a device, and each of those features may be covered by multiple (perhaps hundreds or more) patents. The value of the individual patent is therefore necessarily quite small (and it has no value without being part of a device requiring a large conglomeration of such features) -- it can't possibly be right that the value of one of five hundred patents covering a specific feature is the full hold-up value of including that feature, because that would make the accumulated cost of all the patent licenses vastly in excess of what customers pay or are willing to pay for the device as a whole.
As I think about it, this feels close to the heart of the problem with software patents: The patents each cover a small, independent but complementary incremental improvement that all come together to build common software. The value of the average such patent, i.e. what a practitioner of the invention can reasonably afford and be willing to pay if they were all licensed ex ante, is so much lower than what a defendant has to risk a jury awarding, or the hold-up value of an injunction on a shipped product, that it creates a profitable patent trolling market for entities willing to be loathed in exchange for the significant difference between the market value of a patent when licensed or purchased ex ante and the hold-up value of the same patent to vendors who are unknowingly infringing.
The trouble is the reasonable value of the average software patent is less than the transaction cost of negotiating a license for it. So they either get [cross-]licensed in bulk or (generally obliviously) infringed. But bulk licensing patents is a huge fail because it makes the quality of the individual patents irrelevant, which creates demand for large quantities of low quality patents -- which makes the problem even worse, because the more patents have to be evaluated the less resources you can afford to dedicate to evaluating each one. Meanwhile it devalues and deincentivizes high quality patents that require significant R&D to receive because they go on the same stack with the junk patents.
But how can you fix it? The fundamental problem is that the transaction costs of licensing a typical software patent eat most or all of the value of the patent. Trolling with bad patents is profitable because it foists transaction costs and uncertainty onto the defendant in excess of the value of the patent, so that paying more than the patent is worth but less than it costs to prove that in court becomes cost effective. Recalibrating damage awards and making it more difficult for plaintiffs to unreasonably foist litigation costs onto defendants may reduce the trolling problem, but if effective it would eat a lot of the value of most software patents, often the entire value, because the consequence of the value of the patent not exceeding (or only barely exceeding) the cost of negotiating a license would be born by the patent holders rather than the prospective licensees.
Which seems like a necessary outcome in any event. If the perverse incentives created by bulk licensing are to be avoided then the absolute number of software patents in play has to go down significantly, so that those remaining are the few high quality patents that justify licensing individually. I don't see any obvious way around that.
[note: responding in two comments because it said "that comment is too long"]
I agree with that. That is one of the symptoms of the disease. It is preposterous for every competitor to be infringing thousands of patents held by every other competitor. That is a pathological outcome that doesn't commonly occur in non-software industries. But we have to address the disease, not the symptom. How do we get to a world where the large majority of patent infringement in the software industry is a result of intentional infringement rather than independent invention or a need for compatibility with existing software? The easy solution is for software to be unpatentable. If you don't like that, what's your alternative? (And if the solution involves improving software patent quality, please explain how and also what to do about all the low quality patents already issued.)
> and even if they prevail, damages are limited to the patents asserted.
The other side of this coin is that juries are extremely arbitrary when evaluating the value of a patent, and in the case of software patents will often vastly overestimate their value. There are easily thousands of features the lack of which would materially reduce market demand for a device, and each of those features may be covered by multiple (perhaps hundreds or more) patents. The value of the individual patent is therefore necessarily quite small (and it has no value without being part of a device requiring a large conglomeration of such features) -- it can't possibly be right that the value of one of five hundred patents covering a specific feature is the full hold-up value of including that feature, because that would make the accumulated cost of all the patent licenses vastly in excess of what customers pay or are willing to pay for the device as a whole.
As I think about it, this feels close to the heart of the problem with software patents: The patents each cover a small, independent but complementary incremental improvement that all come together to build common software. The value of the average such patent, i.e. what a practitioner of the invention can reasonably afford and be willing to pay if they were all licensed ex ante, is so much lower than what a defendant has to risk a jury awarding, or the hold-up value of an injunction on a shipped product, that it creates a profitable patent trolling market for entities willing to be loathed in exchange for the significant difference between the market value of a patent when licensed or purchased ex ante and the hold-up value of the same patent to vendors who are unknowingly infringing.
The trouble is the reasonable value of the average software patent is less than the transaction cost of negotiating a license for it. So they either get [cross-]licensed in bulk or (generally obliviously) infringed. But bulk licensing patents is a huge fail because it makes the quality of the individual patents irrelevant, which creates demand for large quantities of low quality patents -- which makes the problem even worse, because the more patents have to be evaluated the less resources you can afford to dedicate to evaluating each one. Meanwhile it devalues and deincentivizes high quality patents that require significant R&D to receive because they go on the same stack with the junk patents.
But how can you fix it? The fundamental problem is that the transaction costs of licensing a typical software patent eat most or all of the value of the patent. Trolling with bad patents is profitable because it foists transaction costs and uncertainty onto the defendant in excess of the value of the patent, so that paying more than the patent is worth but less than it costs to prove that in court becomes cost effective. Recalibrating damage awards and making it more difficult for plaintiffs to unreasonably foist litigation costs onto defendants may reduce the trolling problem, but if effective it would eat a lot of the value of most software patents, often the entire value, because the consequence of the value of the patent not exceeding (or only barely exceeding) the cost of negotiating a license would be born by the patent holders rather than the prospective licensees.
Which seems like a necessary outcome in any event. If the perverse incentives created by bulk licensing are to be avoided then the absolute number of software patents in play has to go down significantly, so that those remaining are the few high quality patents that justify licensing individually. I don't see any obvious way around that.
[note: responding in two comments because it said "that comment is too long"]