And what do you do after you read a patent? Clearly, not building useful products that use those ideas, because a patent expressly prohibits you from doing that.

The part of the argument you're missing is that patent were intended to "promote the progress of science and useful arts" and, as they're practiced today, are doing the exact opposite: delay the progress and useful arts.

In software, if I can't use an idea for 20 years, it's useless. I would rather take my chances, think about the problem and come up with my own solution. Unfortunately, because of onslaught of patents, that solution is more and more likely to have been patented by someone else.

As a practitioner of software arts and science, I not only don't have the time to sift through thousands of patents, but I also risk potentially fatal damages by independently coming up with ideas of my own.

That is the cost of patents and it's becoming unbearable.

I agree completely. One of the key problems with software patents is the very idea that an algorithm is deemed worthy of patent protection simply because it is not obvious to the "average practitioner" of the art. But what skills/level of creativity constitute "average"? What if you have a company, like Google, whose core hiring principle is to find people who have above average skill and creativity? This seems to set up a perverse disincentive to hire smart people because the smarter they are, the more likely they are to rediscover on their own an idea that is already patented and thus violate it.

The fact is that if the idea patented truly is the best way of doing something then a lot of smart and above average developers are likely to unknowingly bump into that idea out of necessity. And this seems to be a problem that is unique to software patents precisely because of the closeness of software to mathematics itself. Some algorithms are provably optimal for a particular problem which means that everybody smart enough that faces that problem will solve it the same way. If you're smart enough to come up with the best way of doing something on your own but can not use it because it was patented then you're forced into a wasteful outlay of creativity just trying to work around the patent. I don't see this as spurring innovation or at least not worthwhile innovation.

The references section is actually a big problem with patents. Having worked with attorneys in the past on patents, they actually want you to find all prior art and to put it in the references of the patent. Why? Because patents are rarely invalidated due to 3rd party prior art listed in references -- even if it really would invalidate the patent looked at objectively. But it effectively takes that prior art off the table for future litigation.

I'd be surprised if Google didn't list Kleinberg's work as it was really well known in academic circles at the time. But if Google ever sued you with PageRank, and you tried to say, "Look -- prior art in Kleinberg!" They'd just say, "The USPTO already examined it and determined it isn't prior art. Sorry." You could almost certainly do a better job than the examiner in looking at the evidence, but the burden is proof is so high for you at this point that unless Larry Page wrote in his code, "This is a copy of Kleinberg's work that we slipped past the UPSTO" you're not likely to invalidate it.

I don't buy that at all. "Look -- prior art in Kleinberg!" is your opinion, not some kind of objective truth that was nefariously "slipped past" the patent office. The differences between PageRank and that prior work are crucial--they are what the patent actually covers.