Hacker News Reader

You still need to fix 2 other problems with software patents, otherwise it won't work.

1) shorten the software patent period to 3 years

2) dramatically decrease the vagueness and triviality of software patents somehow

Otherwise, you end up like in the mobile industry, where something like 250,000 patents exist in any given phone, and I assume most of those are not enforced, but if they were regulated, and manufacturers were forced to pay for them by default, I could easily imagine patents costing more than the product itself. This is why you need to drastically reduce the number of patents as well.

Well, I think my system would actively incentivize the reduction in the number of patents. Fewer patents would lead to higher licensing fee per patent, so you'd have an incentive to only patent the stuff that really matters (since every patent has a base cost to acquire).

Also, I'm not proposing that there be some sort of strict requirement that you must pay the licensing fee for all the patents you use. Obviously cross-licensing agreements and such would still be valid.

What I'm trying to cure is the system the patent trolls currently operate under:

Patent Troll - "I'm going to sue you for $1 mil for infringing my patent unless you pay my licensing fee."

Startup Founder - "How much is your licensing fee?"

Patent Troll - "Only $10,000"

You have to keep in mind that most patent lawsuits never see judgement. They exist only to coerce alleged infringers into paying licensing fees. The problem is that the licensing fees are arbitrary. They are not reflective of any sort of market, but rather reflective of infringement-penalty * chance-of-judgment-against (well, I guess this is a perverse sort of market: a market of jury pools).

Indeed. Also may I say "thank you" to the Supreme Court for their wonderful decision in Caraco v. Novo Nordisk, which more or less ends the usefulness of method-of-use patents for drugs.

I thought 3-5 years would be a good enough length of time (leaning more to 5 years for physical applications). Though companies may still then hoard technology that isn't ready for the market yet, which would slow innovation. Longer patent times are perfectly fine if fees are regulated.

It often takes more than 3 years to commercialize fundamentally new computer algorithms. It is not just development and implementation, there is also an integration and transition period from software designed under the assumption of previous algorithms. 10 years is probably a more reasonable limit for computer algorithms. The R&D cost that goes into computer algorithm patents is often substantial as well.

When people say "software patents" they are usually talking about "business process" patents, not computer algorithm patents, and business process patents should be disallowed generally. They are inherently vague.

>When people say "software patents" they are usually talking about "business process" patents, not computer algorithm patents, and business process patents should be disallowed generally.

You are contradicting yourself. "Business processes" are nothing more than algorithms, themselves. Algorithms are math. Math isn't patentable.

Not at all. What distinguishes "business process" patents from other patents is that any specification strict enough to be on par with traditional patent matter are largely useless; they are uniquely vague in their specification because that is what makes them useful. This allows an enormous amount of scope creep simply based upon the subjectiveness of the interpretation.

Computer algorithms are "math" in the same way that chemical process patents are math. By the criterion upon which you deem computer algorithms to be math, you can prove that all subject matter is math via algorithmic information theory. To make the distinction, you can patent a specific sorting algorithm (legal, there are an infinite number of such algorithms) but you cannot patent the idea of sorting (illegal because it is a mathematical concept). If the argument that computer algorithms are invalid because they are mathematical concepts is going to be made, it needs to be applied consistently which would exclude all currently patentable subject matter.

In fact, computer algorithms have always been considered patentable subject matter around the world because they strictly specify the design of an electronic circuit. Electronic circuit designs are a traditionally patentable subject matter. A mathematical concept has no such specification.

>If the argument that computer algorithms are invalid because they are mathematical concepts is going to be made, it needs to be applied consistently which would exclude all currently patentable subject matter.

I wouldn't argue it that way, but you are correct: nothing should be patentable. The whole patent system is wrong and fucked up, we need to get rid of it before it causes even more irreparable harm to mankind.

At the current rate of change 1 year is more than enough for an iPhone/iPad/kindle/etc... to completely dominate the market in terms of mindshare. Software patents need to go away immediately, and technology patents, if needed at all, need to be a year at most. If your idea is so easy to copy, then it probably wasn't that great of an idea anyway. It's taken 5 years for Google to truly catch up to the iPhone/iPad and be a real competitor.

That is a common misconception. Actually patents were an adaptation of an existing system where royalty would give monopolies away to companies they liked.

No-one did any scientific tests at the time to ensure that they were a good idea in the first place. Now those who do retroactive studies are ignored.

Actually most of the world, that has gotten rid of royalty, seems to be doing its best to bring back at least some of the trappings; for example, http://westhunt.wordpress.com/2012/07/21/presidential-transp...

>Actually patents were an adaptation of an existing system where royalty would give monopolies away to companies they liked.

Do you have a historic reference for that? According to Wikipedia, the earliest known patent law was in Greece, 500 BC, and modern patents are said to have originated with a statute in Venice in 1474, but no mention is made of this being an adaptation of royalty favoring companies with monopolies (which obviously was a practice that extended long after 1474).

I can back this up. I took a course last year from a guy who's one an expert on the history of IP law (http://www.adrianjohns.com/). The patent system in its modern form grew out of the statute of monopolies (https://en.wikipedia.org/wiki/Statute_of_Monopolies), which was intended to be used only in cases of new innovation, but royalty abused it by granting monopolies on common goods to favored merchants (who paid exorbitant fees for the privilege).

A single product likely infringes on thousands of patents (a possibly dubious Google/RPX estimate was 250k patents for a smartphone), in which even $0.01/patent becomes unsustainable. Especially in software, chances are that you didn't learn about the ideas in the patent by reading the patent or observing products that use the patent, but rather on your own while solving a problem.

I guess I should clarify: there are different sorts of licenses. Some are per device or product sold. I was referring to an unlimited use license. The idea being (with the formula) that if your idea is good enough that every one of your competitors wants to use it, then they will effectively pay you for the full cost of discovery/development (at industry average R&D costs).

So at $10 per patent, you'd pay $2.5 mil as a one-time fee. I'm curious how that would stack up to current litigation and other legal costs to fight/prevent patent litigation.

If an unlimited use license is $10, it's not worth the patent owner's time to write a letter asking for the license fee.

My impression is that many of the ideas covered by software patents are independently "discovered" by each entity that happens to use them. The patent is now inhibiting innovation by dwelling on an old idea that is apparently not so novel after all.

> If an unlimited use license is $10, it's not worth the patent owner's time to write a letter asking for the license fee.

I guess that's kind of the point.

Patents only inhibit innovation when they can be used as a cudgel to force inventors into paying exorbitant fees. If fees are inversely proportional to the proliferation of patents in an area, the natural incentive is to only patent that which is important.

> Originally, patents were supposed to encourage idea sharing.

The original idea of Marxist communism was very humanitarian in its intentions. We know how dictatorial it turned out when it was implemented. Good political theories that turn bad typically didn't take all parameters of reality into consideration. They're utopias.

Well no, the "class struggle" and "armed takeover of power by the proletariat" is the core of Marxism, not quite so peaceful. But this is irrelevant to the topic.

You don't have to be a pacifist or a doormat to be a humanitarian.

Arguably the US was much closer to Communism than Russia in the 50-60's. Unions effectively controlled the means of production through conflict 'strikes + other means' and redistributed more than the fair market value of their labor to the workers over the long term. Russia on the other hand was a de facto feudal system, with the party sharing power and privilege to a few at the expense of the peasants.

http://en.wikipedia.org/wiki/West_Virginia_Coal_Wars

Not to mention the 91% top income tax bracket.

Patents were initially devised to let inventors to recoup their R&D expenses by giving them exclusivity on the market. And this - the R&D expense - should really be a defining factor in whether or not something is patentable.

This way a 5 minute brain fart by a recent graduate won't be patentable, but a clever optimization technique that stems from several man-month of testing on a cluster of Crays will be.

That would tax an efficient use of resources. Its the innovative idea in the patent that should matter, not how much money you spent on figuring in out. Some might take less time and resources to figure out the same thing and basing everything on R&D expenses makes it completely unfair to them.

But why grant software patents at all?

Software is already protected with copyrights as literary works. Copyright protection is not as broad as patent protection. It doesn't prevent competitors from writing a program with similar features but different source code, but it does protect against competitors copying the source code.

All patents are granted copyright protection in this sense. After all, they are filed on paper and both the process and implementation are reducible to such. To give an example, chemical process patents are among the oldest and most venerable kinds of patents in existence and are essentially equivalent to computer algorithm patents (except involving molecules instead of bits).

A new algorithm for efficiently producing a particular chemical is protected by patents. The implementation of that algorithm, such as in a chemical manufacturing plant, is protected by copyright. There are an infinite number of ways the abstract process can be reduced to practice, all of which are copyrightable.

In practice, the implementation copyright and the process patent are separately valuable, are often sold separately, and protect different people. The patent protects the effort of the individual who developed the chemical process. The copyright protects the effort of the engineer who designed a way to efficiently uses that chemical process.

> That's probably all it's worth.

Although probably not, unless the parent has all the code needed to implement whatever the patent covers in the language I'm using and correctly following the conventions and patterns of the language and platform.

That could be worth $10.

Would you pay somebody $10 for a blogpost like that?

There's a bunch of those around. Most don't even have ads or any revenue stream other than infamy.