Bernhard Eversberg
Wed, 01 Sep 2010 07:08:43 -0700
Marjorie Bloss wrote:
Regarding vendor preparation for RDA, it would have been difficult to develop a system to accommodate RDA out of the box when RDA wasn't available until June 23rd of this year. However it would indeed have been possible for ILS vendors to redesign their systems to support relational databases perhaps even using the FRBR/FRAD conceptual models as their basis.
"Relational databases" is an all too common misunderstanding here. The term does not indicate that these engines are particularly well equipped to map relations as FRBR has them in mind. They are not. "Relational" stems from mathematical concepts of set theory, not more. A "relation", in the technical sense of the term "relational database", means nothing but a "subset" of a table. It does not denote the idea RDA (or FRBR) has in mind when saying "relation" and meaning a link between entities. Or, an RDBS is not based on the entity-relationship idea. What we need are *object oriented* databases. These have less difficulties with mapping entities, and relations between them, as RDA understands these matters. However, there's no standardization of object oriented database structures or concepts. For RDBS, there is SQL as a near-universal language for queries and operations on data, but there is no such thing for the OO world. This means it would be more difficult to construct object oriented database schemas that would be portable between systems. In reality, RDBS are very often the lower level engines in library applications, but the R in them is not actually used to a large extent. Rather, there is always a very substantial layer of software above it, and that layer is then mostly strictly proprietary. In short, it doesn't make a lot of sense to invoke "relational databases" when talking about RDA implementation. Just "databases" is enough, or if RDA "relations" are to be emphasized, one may talk about "entity-relationship models for databases". B.Eversberg