Section 1.6 describes the advantages of a so-called "self-describing architecture," but does not define that term. The reference cited [9] does not mention that term. Metadata can indeed describe some aspects of data - but usually limited to the W's (what, where, who, when, ...) plus, in some cases, a data dictionary of some type. But even this is only partly self-describing, and prior knowledge is needed to interpret the metadata (in other words, the metadata doesn't describe itself). So self-description is never total, and metadata only captures some limited self-describing aspects. 

The authors should define what they mean by self-describing, and perhaps, use more direct language, such as metadata.