A recently conducted a survey asked people in research about their readiness to make data freely available. The results were not very uplifting: about 10% of the respondents would share openly, some with a password (30%), the majority won't share at all. We believe that publishing open scientific data should be tied to the notion of a journal/publication because people sitting on valuable datasets will rarely make then available unless they feel it's in the form of a "real", citable publication. We propose the creation of an online journal by the OKFN that allows people to publish articles describing their open datasets, along with methodology for creation, intended applications, or other information useful for researchers who would potentially reuse the data. In most research articles, the dataset is provided as supplemental material, if at all. Here it would be at the core of the publication, while methodology, applications, etc. are supplementary. It could be an all-purpose OKFN data journal with discipline specific tracks/committees. (The text for this idea was built from e-mails exchanged in the context of this thread at the open-linguistics list: http://lists.okfn.org/pipermail/open-linguistics/2011-May/000078.html)