Mechanized retrieval of texts has developed in terms of rapid item-by-item scanning and selection. If texts are requested by content or relevance only, the major problems are indexing and specifying. In terms of mechanical translation, these are the most profitable targets for mechanization. Because all retrieval systems must ultimately produce legible documents, searching and output speeds inevitably differ (mismatch). Searches should be made for blocks of requests at a time, and multi-level access, based on some strategy of search, is always needed, however rapid the item-by-item scanning. Rational terminology and library principles are necessary in all filing systems, stores vocabularies, program libraries, etc., whether mechanized or not. Automata for library-type activities have to simulate the libary users, as well as the library organization.