Each record in the data should include an identifier. This identifier should be persistent across revisions to the data set so that external references to individual record can follow updates. The identifier can be a globally unique URI identifier following Semantic Web best practices, for instance.
The data format should be documented so that those familiar with the domain of the data set can understand it. All columns, tags, and abbreviations should be described. However, XML schema or the like are not necessary.