[jandrewrogers]

The reason for creating a patent system is to encourage public disclosure of inventions. Patents came out of the enormous problems that resulted in the widespread use of trade secrets, which actively hindered technological development in the past. This essentially forces everyone to redundantly do a lot of R&D that has already been done under non-disclosure rather than incorporating ideas that have been publicly disclosed and using that as a starting point for further innovation.

We already see this in a number of areas of computer science, where the algorithm state-of-the-art at private companies is a decade ahead of the public academic literature. This basically has the academics in many computer science domains essentially trying to reinvent things that have been known for years as trade secrets at large companies rather than doing genuinely original research.

I do not think there is an obviously good solution, since algorithm patents are difficult to enforce in any case, but one of the trends resulting from a lack of algorithm patent enforceability is that almost all cutting edge computer science R&D is now done and kept as a trade secret by companies because it has a material commercial advantage.

[bjt]

We don't need a patent system in order for openness to beat closedness in the software world. Linux and lots of other open source projects are winning their markets without relying on patents at all.

No one forces you to choose a patent over keeping your invention secret. Trade secrets still exist everywhere. And that's OK.

You say it's a problem that companies are a decade ahead of academia in some CS areas. Given that companies can choose what to make public and what to keep secret, what's the alternative? It's the ability to have some exclusive time in the market that incentivizes the companies to do that research in the first place. I'd rather have them ahead by 10 years because of trade secrets than getting 20 years of exclusivity through a patent.

[jandrewrogers]

Linux leverages ubiquitous disclosure in computer science from decades past. Open source software is largely built on a computer science technology base that, while sufficient, does not reflect many qualitative computer science advances that have been made in the last decade. For software that routinely leverage computer science advances (e.g. databases and parallel systems) you already see a divergence in capability and performance between platforms that are not open source and the open source world because of this lack of knowledge transfer.

Graph analysis platforms are a great example of this. Every open source graph analysis platform is a non-scalable toy that is useless for most real-world use cases compared to private systems that were being quietly deployed five years ago. Why the discrepancy? Because the unpublished algorithms and data structures used by some closed source systems are several orders of magnitude more scalable than is possible with the best existing public algorithms. And this is not the only example with which I am familiar.

The problem open source software has if it is based on increasingly obsolete computer science is that it will not be economically competitive with closed source that is much more scalable and/or much faster and/or has much higher throughput per watt. A combination of materially better capability and efficiency has an enormous impact on CapEx and OpEx, and people care about those kinds of things when it starts adding up to large amounts of money.

[hackinthebochs]

> I'd rather have them ahead by 10 years because of trade secrets than getting 20 years of exclusivity through a patent

The difference is that had the invention been patented, the academic world could be advancing and building on the idea. Trade secrets effectively freeze the progress of an idea. Patents allow an idea to be advanced, even if no company could commercially benefit from the idea in the intervening time. Are we better off as a society with Watson's inner workings kept as a trade secret?

[bjt]

We're not forced to choose. Unlike a patent, a trade secret leaves anyone else free to invent and use the same idea.

[hackinthebochs]

>Unlike a patent, a trade secret leaves anyone else free to invent and use the same idea.

What is the probability that such an idea is independently recreated? Hard to say, but its likely inversely related to how ground-breaking the invention is, which are the ones that society benefits the most from them being made public.