On retrieval from a small version of a large data base
Adrian Walker · Very Large Data Bases · 1980
A person in the higher levels of a hierarchical organization may wish not to use a fully detailed data base, but rather an abstracted data base, perhaps dropping into detail in only a few areas. We compare answers from queries put to an abstracted data base with answers obtained by querying a full data base and then abstracting the result of the query. We show that, for some common relational retrievals, querying an abstracted data base always yields the correct information, plus, in some cases, some incorrect information, and we give simple conditions on the abstraction which ensure that only the correct information is fetched. For the cases in which the conditions may not hold, we suggest an ordering, called succinctness, for comparing the quality of different abstractions.