[azakai]

I don't think this is so difficult. Something has to be nontrivial to practitioners in the relevant field, so just survey a few random software engineers - not some clerks at the patent office.

Pay them for their time, obviously, and that cost should come from the patent application fee.

If this were done the vast majority of software patents existing today would not have been granted. Which is probably why the system wouldn't do it.

[JumpCrisscross]

You just described a patent lawsuit - both sides bring experts who vouch for their perspective in light of a judge who moderates and then rules on the debate from a general legal perspective.

[sses]

It's significantly different from a patent trial. In this scheme the patent is not granted - and can't meaningfully affect the market - until it has passed some level of technical scrutiny by a (hopefully) neutral third party. In a trial the patent has already been granted, it's already affecting the work done by engineers, licensing fees may have been paid, and so on. And the evaluation by experts in the trial seems less likely to be neutral.

[backprojection]

1. I think the obviousness of an idea might change throughout time. 2. What if one industry is dominated by two companies, in the sense that all the experts are employed by one or the other, then how could they be unbiassed?

[drcube]

Simple. Monopolies and duopolies shouldn't be awarded patents, which are just monopolies themselves. Why award monopolies to monopolies?

[rcthompson]

First you have to prove in court that they are a monopoly.

[taligent]

Nothing wrong with being a monopoly so long as you don't use that monopoly to restrict competition.

[tikhonj]

There's nothing illegal with being a monopoly. That doesn't mean it isn't wrong. I think in most cases (with exceptions like utilities), not having monopolies leads to a healthier market than having them.

[CamperBob2]

1. I think the obviousness of an idea might change throughout time.

Sounds like an excellent reason not to allow someone to own the idea for 20 years.

[darkarmani]

Yes, or changing it from a binary system to a continuum that maxes out at 20 years. Software patent? 5 years. New drug - 20 years. New algorithm (can't patent math of course) 10 years.

[vibrunazo]

I really don't think it would be that simple to make sure that process in unbiased. And that the additional costs wouldn't make patenting exclusive to billionaire companies.

But I very sincerely hope you're right :)