The latest arrival on the data-aggregation scene is infochimps.org, by Philip Kromer at the University of Texas in Austin. Infochimps is most similar to Numbrary
This site is clean and well organized. Tagging of datasets is quite thorough. A unique feature to infochimps is recognition of data fields across sources, so you could find, for instance, all sources that have figures for “Personal Income”. So far I’ve been unable to find any fields that actually appear in multiple sources, but this is presumably just a matter of adding metadata to datasets in the repository.
There’s no online viewing of dataset contents, but every dataset is available in (compressed) CSV, YAML and Excel formats. All the reformatting of the source data has been done by infochimps, saving users from repeating this chore. Like Numbrary, infochimps provides links to the source documents that were used to construct the datasets on the site.
So far infochimps has accumulated almost 1,500 datasets, and many more are promised on their blog.