[JoshTriplett]

> I know there are lots of anti-patent folks on HN, but if anything deserves patent protection, digital fountains certainly do.

I'd suggest that the usefulness of algorithms like fountain codes (and other innovative algorithms like RSA and many compression techniques) leads to the opposite conclusion: such algorithms prove so broadly useful that patents on them have even more of a negative effect than usual.

[zvrba]

How does it have negative effect? Consider three cases: 1) an algorithm is designed and patented, and the inventor has the right to use or license it. 2) an algorithm is designed and never published, i.e., held secret for proprietary use of the inventor. 3) an algorithm is designed and put into public domain.

With 3), the inventor never gets a chance to commercialize his invention: as soon as he publishes the algorithm, he will be destroyed (or assimialted) by large players. With 2), we never know how many cool, useful algorithms are rotting away in commercial products. As far as the society is concerned, the algorithm in question does not exist. With 1), the inventor benefits, AND also the society (in the long-term).

IMO, the patent system should be reformed so that having a large portfolio of unused patents becomes a liability. For example, the patent holder would have to periodically (e.g., every 2 years) apply [with fee, of course] for patent renewal. For this, the holder must be able to document that he is actively exploiting the patent (1). There should also probably be a minimum limit for the required income from patent exploitation. Failing to do so, the patent would automatically expire. The patent office should publish an on-line database of expired patents.

There should also be restrictions on what can be patented: for example, work funded by government funds must not be patentable: the inventors' (in this case, mostly researchers at universities) day job is by definition designing things that will benefit society, are paid for that by the society, and therefore have no further rights to patent protection. (This is, IMO, fair since research is [should be] free, so the society has no firm guarantees that the researchers will produce something valuable, or anything at all.)

(1) By exploitation I mean either making a device covered by the patent, or licensing it to somebody who making such device.

[JoshTriplett]

You seem to suggest that algorithms won't get published if their authors can't patent them and commercialize them. Given that the vast majority of algorithms don't get patented, and that the vast majority of patents never get licensed except in bulk cross-licensing deals, this seems almost entirely untrue.

Aside from that, I'd point out that in the rare case when an algorithm does get published as a patent and licensed, that makes it entirely unusable in Free and Open Source Software as well as in open standards, for a period of ~20 years. At least when the algorithm remains unpublished, someone can re-discover it (or something more useful that would build on it); patents force willful ignorance of an entire class of algorithms, as well as anything in the general vicinity of those algorithms.