An analysis and improvement of the kermit file transfer protocol by data reduction

The composition of over 200 files typical of those transferred between computers in a university environment is determined. The files are analyzed on a byte basis and frequencies of repeated characters and of strings of non graphic characters are determined. Two operators which result in less protocol overhead for strings of non graphic characters are developed and shown to operate properly. The relative savings in protocol overhead from these mechanisms is determined by file type: source, list ing, object, text, and Huffman coded. Savings ranging to almost 50 percent would result from the combination of the non-graphic string operators and a repeat sequence opera tor. Most commonly, however, savings is in the range of 10 to 15 percent.