Patents are an artificial system forced upon us with the hope that they accelerate innovation. Just as you said, the hope there is that it encourages someone to take the risk of making something new with the hope of massive exclusive rewards.

The problem with that hope is that it ignores how innovation happens. There is no reclusive mad scientist working in some dank basement on the next-best-great-thing. That's childish.

Innovation happens through collaboration. You take someone else's work and you build upon it. Bit by bit, people build the next-best-thing.

The current patent system tries to take this into account by creating a lifespan of 20 years or so. But for those 20 years ... you are stifling the potential innovation that can happen by incorporating that new-best-thing into the next-best-thing.

This is even worse for software! A new paradigm arises (Internet, AJAX, Geolocation, etc.), a new set of algorithms become necessary to make things work. These paradigm shifts take years, not decades. As a consequence, innovation is limited to those who created the initial algorithms, since they're the only ones who can build on the most basic work.

Imagine Google created the algorithm/API for getting geolocation from the browser and built it into Chrome, then patented it. Nobody else would be able to do it. Obvious invention you say? Not as obvious as Amazon's one-click-to-buy button. The worst part in this example is simply that while Google may have invented it, there is an almost certain chance that Mozilla/Apple/Microsoft would have been able to make it better within weeks. We are missing out on that kind of innovation for patented software.

Today, Startup can sue, alleging patent violation, but most startups don't have the cash on hand for an extended patent battle anyways.

That's a very interesting point(s). Again, arguing from a devil's advocate POV. People who are generally against software patents commonly make the claim that it stifles the innovation of the single person, sitting in the their dank basement working on the next-best-great-thing (see other comments in this topic for examples, including a rather brilliant quote by Carmack).

In terms of collaboration, sure, patentable ideas/software often exist due to collaborative team efforts, just like patentable physical widgets like industrial egg fryers. But that team is generally paid for by one company and the patent is then assigned to that company.

There really appears to be a few major issues that I see with the present system:

1) Patents are transferable

2) Patents aren't researched enough against prior art before reward

3) Patent length is too long

4) Fighting bad patents is expensive and overly hard, even when armed with tons of prior art

5) Patents don't necessarily need an extant physical or implemented example to be granted.

6) Patents can be granted without the intention of implementation.

I think for real patent reform to work, these 6 points need to be addressed to keep the concept of patents (risk in exchange for exclusivity) a workable model:

1) Patents shouldn't be transferable -- this eliminates patent trolls

2) The Patent Office needs to be properly funded and given enough time to properly research patents before grant.

3) Patents (like copyright) needs to be shorter and dependent on the effort required to create the patent in the first place. For example, software patents should probably only survive 12-18 months. That gives the owner a risk/exclusivity that's long enough and if the idea proves valuable, opens it up for others to use (and improve the economy as a whole) within a reasonable time. In other words, this recognizes that risk doesn't just end at the grant of a patent. Actually doing something with the patent is important.

4) Appealing bad patents should be relatively simple and swift.

5) You should actually be able to demonstrate something that's patented.

6) If you don't use the patent in a business use-case, you lose it. A bit like Trademark.

Just some random ideas.

6) If you don't use the patent in a business use-case, you lose it. A bit like Trademark.

Suppose somebody invents a device but lacks the facilities to mass-produce them (building by hand is probably too expensive to make for a viable product). Rule (1) bars the inventor from selling the patent to somebody who has these facilities, so it appears necessary to arrange a licensing deal with a manufacturer (I expect selling a license to manufacture is still allowed and would count as a business use of the patent). How long does the inventor have to make such a deal before those who do have the right production facilities can simply ignore the patent and produce/sell the invention without compensating the inventor?

Licensing of a patent by the original assignee for specific use of for immediate manufacture makes sense to me.

b) patented it themselves first

It seems to be fairly common that somebody already did (a) but chose to forgo (b).

spent $130 million dollars just in the R&D phase for what amounts to…. Obvious? Perhaps. But nobody had gone through the trouble to actually do it, and spending that kind of overhead was a significant risk.

How much of that was it actually necessary to spend just to come up with the concept? Evaluation of the available COTS components, writing the perl scripts and cron jobs, management overhead, etc. are costs a competitor would have to incur as well, which makes them not part of the cost a patent is meant to compensate for. The fact that nobody had done it before does not necessarily mean nobody was clever enough -- just that nobody who was clever enough had had the problem this product was meant to solve, and being a solution to a novel problem does not make it patentably non-obvious.

In other words, if the ideas are obvious, a piece of software that does this should be buildable.

I don't think a computer program is likely to cover everything that would count as obvious.

"A person of ordinary skill is also a person of ordinary creativity, not an automaton." - SCOTUS opinion on KSR v. Teleflex

This is probably about as long as it would take for a competitor to write the necessary code and do the necessary testing anyway, so it seems that the reward of being first to market with good code is enough compensation for the risk involved.

If we really want to fight software patents of obvious ideas, or permutations of obvious ideas, perhaps somebody should write a tool that simply combine every obvious idea in software into some greater system, then documents it all thus creating prior art that stops all of these so called bad patents...and then patent the "prior art permuter". In other words, if the ideas are obvious, a piece of software that does this should be buildable. I call this the "obviousness of software patents theorem".

I was thinking about a similar system, wherein hardware and software components are treated like building blocks. You throw some (or all) known blocks into a bag, and randomly draw n blocks and come up with a way they can be used together. You then start applying random domains to the same building blocks (e.g. security, gaming, manufacturing, automobiles) to prove that moving an existing invention to a new domain should be obvious.