This highly readable, clear, and accessible book by Richard Gartner demonstrates that the term “metadata” is neither--on the one hand--an exotic buzzword, nor--on the other hand--a reference to a resource description framework (RDF) triple. The author properly uses abstraction (“suppression of irrelevant detail” [1]) to present the essential characteristics of the concept of metadata that has been with us for millennia (without using the term) and has been adopted and used in computing since the late 1960s/early 1970s.

Throughout the book, Gartner stresses that metadata are designed by humans for a particular purpose and that therefore “there is nothing objective about metadata: it always makes a statement about the world, and this statement is subjective in what it includes, what it omits, where it draws its boundaries, and in terms it uses to describe it.” In other words, different approaches to metadata mean different abstraction levels and different viewpoints [1]. The book includes several interesting detailed examples illustrating this subjectivity, especially “relationship of metadata to ideology, consciously and unconsciously,” such as the change of Library of Congress subject headings (terminology as ideology), the fourth edition (dated 1999) of the Chinese library classification that “inherits a Marxist orientation from its earlier editions,” and sinister uses of classification schemes by eugenicists and by Nazis. Gartner does not, however, refer to Orwell’s use of language (Newspeak) as arguably the ultimate metadata.

The need for the proper structuring of metadata and therefore the essential role of (explicitly or implicitly specified) relationships in doing just that is also emphasized throughout, again with numerous examples, starting with various library catalogs. Here, the author notes that the rules (cataloging practices) must be explicit and consistent, using machine-readable cataloging (MARC) developed in the late 1960s by Henriette Avram as a detailed example. At the same time, Gartner emphasizes that fixed hierarchical schemas (like the Dewey Decimal Classification) are very restrictive, setting frameworks that “may blatantly support ideological agendas” and limiting search abilities. Flexible units of topics, “which could be combined flexibly to form more complex combined subjects” (“facets” invented by S.R. Ranganathan in 1933), are a huge step forward. We may say that faceted classification represents abstraction viewpoints showing mutually independent hierarchies, but unfortunately they are not particularly popular in some information technology (IT) environments because many popular tools do not support them. Gartner stresses that a single hierarchy “can build statues which set out our knowledge in stone instead of making it malleable and able to grow.” I think that the problem here is not only in dealing with new knowledge “unknowable at present” (Ranganathan), but also in ossifying a single “correct” or “proper” viewpoint, as discussed elsewhere in the book.

Gartner properly observes that “finding [the] context is vital to make sense of the semantics of a field” and refers to the need to specify relationship semantics in a precise and explicit manner: “the words ‘broader,’ ‘narrower,’ and ‘related’ do not tell us a great deal about the semantics of relationships between terms.” Note in this context that a facet, as described by Ranganathan, is an atomic concept from which more complex ones can be created by combination--in good agreement with the definition of composition based on property determination [1,2,3]. Note also that in the online world (unlike, for example, in a library card catalog), the sequencing of the components of a compound subject is not important, and therefore choosing and manipulating facets (components) is easy. As one of the results, a specific item may and will have different codes in different classification schemes. Gartner further observes that defining the “semantics of these links” (that is, relationships) is an important part of defining an ontology; however, this was first “propounded” not by Tom Gruber in 1990, but rather by Wand and Weber in 1989 [2] based on Bunge’s exact philosophy [3]. While specification of relationship semantics is essential for business domain modeling [2,4,5], it regretfully is also not particularly popular in some IT environments, and for the same reason of inadequate tool support.

Gartner notes that divisions between communities and within them led to different metadata standards (over 100 as of 2010), making computer-based information processing even more challenging. In this context, we may point out that implementation-based approaches and tool-based restrictions (such as “binary relationships only,” “no mutually independent hierarchies,” and so on) should not stand in the way of understanding and specification [1,2,3,4,5]; therefore, for example, the semantic web should be a mechanism to make it possible to solve ontological problems, with ontologies (basics of the business domain) developed by domain subject matter experts.

Summing up, the book presents an enjoyable bird’s-eye view of metadata and related concepts, with outstanding examples accessible to nonexperts. In the last chapter, Gartner reminds us that metadata “is [sic] inherently ideological, ... fundamentally an expression of worldviews and no more than that.” I highly recommend the book.