I've heard reports that the journals _Science_ and _Nature_ want the metadata on forthcoming articles to be embargoed. In other words, they are saying that they do not want a repository version to become discoverable before the work is published, even if the article itself is not available through the repository. I think this is really unhelpful.