The objective of the present paper is to find methods necessary to compose a general text processing system by analysing an SLAC-developed system, `FORMAT' and running test programs fit for testing as many feature in `FORMAT' as possible.
The explanations given included a FORMAT's general flow which varies significantly according to each feature, and concrete internal paths and methods of handling FORMAT's facilities, e.g., ones for the frequency dicitionary and locating text words in the input streams.