Abstract: A data retrieval system using a compression scheme especially for serial transmission, which does not inadvertently generate typical flow control codes, such as ASCII values less than 32. A personal computer with a modem is used to access a remote mainframe database. Retrieved data is compressed according to the scheme and transmitted over telephone lines. Compression is achieved by encoding a compressed byte pair in place of a recurring sequence of data longer than two bytes. The byte pair contains information as to the length and offset into the transmitted packet of a previous occurrence of the recurring sequence, so that it may be copied from previously received bytes. The length and offset are stored in the byte pair such that neither byte has an ASCII value typical of those used for controlling modems.