Surrogate subsets: a free space management strategy for the index of a text retrieval system

This paper presents a new data structure and an associated strategy to be utilized by indexing facilities for text retrieval systems. The paper starts by reviewing some of the goals that may be considered when designing such an index and continues with a small survey of various current strategies. It then presents an indexing strategy referred to as surrogate subsets discussing its appropriateness in the light of the specified goals. Various design issues and implementation details are discussed. Our strategy requires that a surrogate file be divided into a large number of subsets separated by free space which will allow the index to expand when new material is appended to the database. Experimental results report on the utilization of free space when the database is enlarged.