While I will push for open access of all materials, I certainly don’t plan on being mean about it; I think you can work on them opening their data (except where there are privacy concerns) without alienating them. Also, I tentatively support limited embargoes, where you have the dataset for up to a year before releasing it. It’s not the ideal solution, but it’ll protect people who think their biggest advantage is the fact that they have data that other people don’t.

I should be braver about this. In my own work with my advisor, we’re all open-code and open-data, all the time. With the Complex Systems folks I’ve been working with, the code isn’t public and I think we’re not allowed to redistribute the data: we pull a large fraction of the traffic from Twitter, as it happens. And I don’t know how much I’d be comfortable demanding, on future papers with them, that they only be in open-access places…