I am quite vocal about disliking attribution licences; I was asked by Maria Ohlsson why that was on Twitter and I decided that it wasn’t possible to write meaningfully what I think in 140 characters, so here’s a blog post.
I take the position that there is a good licence for data and that is the Open Data Commons Public Domain Dedication Licence (ODC-PDDL); it’s a licence that is good for a number of reasons, not least because it has been around for a long time and it allows common practice to continue. What it doesn’t do is needlessly force potential data users into needlessly attributing, although if attribution is necessary, then you still must do it because of the clause regarding “community norms”. The ODC-PDDL is a good licence for other reasons too; it allows you to present your data as a truly free data set, and still responds to those jurisdictions where this is not possible (e.g. Sweden, where you can’t give away your moral rights). Using the ODC-PDDL isn’t the same as cutting your data loose — you do that when you publish your data and forget about it, which is a governance issue rather than a licencing issue.
So, what’s the problem with attribution then?
- Attribution licences aim to create a situation where you need to read the licence — the specific rights in a particular case depend on the licence. You are probably not a lawyer and if you’re like me, you probably don’t enjoy spending time reading licence agreements.
- The desire to control how and when attribution is done causes attribution licences to proliferate; we have a full array of CC-X licences, Open Government licences, etc. All of which add little that makes the data more usable and a whole array of largely inconsequential changes. 
- Attribution actually harms the usefulness of your data — in situations where attribution damages the final product (and here it is often unclear whether you need to attribute at the level of say interface, which ought to tell you something about attribution licencing), in situations where the data could reasonably used without harming the licensor, the licencee abandons using the data because of the licence (rather than contacting the licensor).
- Similarly, knowing when to attribute in the data — if I snag two or three rows from a data base that I use to annotate a class in a data set, what should my attribution practice be? Back to trying to understand that lawyerese…
- How are you going to police attribution? You aren’t/can’t, so drop attribution. Unless you have lots of time on your hands and sophisticated tools at your disposal you can’t police attribution easily; if you can’t police, then don’t imply that you’re going to because making your data useful is a higher aim than ensuring you get attributed in the few cases where you otherwise wouldn’t.
- There was already a data licence that worked in the real world when CC-X and other attempts at creating an attribution-friendly licence came along; in ODC-PDDL, attribution was mandated by community norms, the way it should be. To my mind, the needs that motivated attribution licencing largely stem from a lack of understanding of what open data is.
- The strongest argument of all against attribution: what benefit does mandating a community norm bring? It excludes use outside that community. My commercial partner will look askance at your attribution licence because they’re in commerce, not an academia.
Further reading: http://knoesis.wright.edu/pascal/pub/nomoneylod.pdf