Thursday, December 08, 2011

Open Science and Non-Commercial licenses (a personal reflection to the Oscar/RSC controversy)

Peter has started a new line of discussion in his blog, referring to a correspondence with representatives from RSC last year, about an annotated literature corpus to (re)train the Oscar3/4 text miner. There are very many sides, and after I reread this post for a second time, I was still not 100% happy about all words: I can only try to express the complexity of the matter and how it started, but do hope to be clear that non-commercial licenses are not useful in Open Science.

I have taken part in parts of the correspondence Peter refers to, and I would not have written up things as Peter wrote up his impression of the outcome of that discussion, and at some point I seem to no longer have been included in the email correspondence, as I at least did not know the final outcome (see below), and cannot fully comment on the accuracy of Peter's coverage of that correspondence, but my impression on the outcome, as limited as it was, is not that far away from what Peter wrote up: Oscar4 needs training (doi:10.1186/1758-2946-3-41), and the RSC was unwilling to contribute the full text training corpus to the project without a non-commercial (NC) clause (and I explain below why I think this is bad). Oscar without a training corpus is useless; Oscar with a NC-licences training course is not Open Source (see below). As detailed below, the corpus at sentence level is NC-free licensed, and a lot of training can be done that way. Sufficient?

Peter wrote:

"I pointed out very clearly that CC-NC would mean we couldn’t redistribute the corpus as a training resource (and that this was essential since others would wish to recalibrate OSCAR). Yes, they understood the implications. No they wouldn’t change. They realised the problems it would cause downstream. So we cannot redistribute the corpus with OSCAR3. The science of textmining suffers again."
I do not know if it is factually correct that the RSC would not change (below we read they attempted), or whether the organisation really understood the problems. But, it certainly is a fact that we cannot redistribute Oscar4 as an Open Science project with a NC-licensed clause.

And, I want to add and stress here, that blog posts sometimes are just like press releases: things have the highest impact if written down in a black-and-white fashion; and getting things factually wrong happens to all of us now and then.

One of the outcomes I learned about this week, is that the RSC released the corpus in some form without the NC-clause. The full text paper corpus remained the NC clause of the CC license, but there is also a version where all sentences are released, and this has a CC license without the NC clause. I think this is not optimal, but still very much appreciate the gesture the RSC is making here, and would to kindly thank them for that! And do I want to make that clear too (thanx to Cameron for phrasing it so well in his comment), it is the principle freedom for the RSC to decide what they want to do, and I fully respect that.

Well, with that out of the way, and I wanted to say something about it, having been involved in the discussion, and feeling a bit in between Peter and the RSC here, appreciating both their view points, and having a third one myself, let's focus on this non-commercial clause a bit more.

Of we enlarge our scope a bit, away from written material, to Open Science, it is clear that the non-commercial clause is bad. In the Open Source world, organisations like the Debian project clearly state that non-commercial clauses violate basic freedoms. From an Open Standard point of perspective, this is pretty much the same. The reason, whether you like it or not, we live in a commercial world. Society expects us to me commercial, and any serious business is legally required to make making profit a company goal. Now, this effectively means that any science made available as non-commercial is not Open: you are effectively not giving people the freedom they need to advance science.

In short, a CC license with the NC clause is in fact quite like "yes, we love to be Open, but we are too scared". Now really, I understand this scare. I am a scientist, post-hopping around Europe, not tenured, and not being an experimental scientist, unlikely to become one. Don't tell me about risk and scare of making things Open. Yet, I did, and it payed of (not enough yet; still looking for a fixed academic position, as I already indicated). But in the more than 15 years I have been working now in Open Science, I have yet to find a compelling (or any) argument to back up this fear: the perceived risk of the NC clause has so far not proved any different than a fear of ghosts.

On the other hand, if I would not have been involved in Open Science, I would not have worked for the top European institutes I have been working in the past ten years.

So, what are the arguments for using the NC clause? The fear I understand, but arguments I do not see that support that a NC clause is useful in an Open Science setting.

Further reading: