Bowin 



act cautiously in their operation during rough weather. The 

 necessity of securing the primary recording units of a system is 

 undesirable. We are hoping that We will be able to operate the 

 magnetic tape drives of System III through all weather conditions 

 encountered. At present, however, we do not know if this will be 

 in fact the case.' 



We h'ave found that all inputs that are entered manually by an 

 operator, particularly routine manual inputs are unsatisfactory. 

 All human beings will, with varying frequency, either enter 

 erroneous values, turn the wrong switch, push the wrong button, or 

 otherwise make mistakes. If the operator is to make an entry at a 

 particular time, this is another potential for error. Mistakes will 

 also occur even if the values are entered by the person who will 

 later make use of the data, and therefore by someone who will have 

 the most conscientious attitude towards the task. Such errors 

 create an enormous amount of wasted time and difficulties in the 

 correction and analysis of the data. For example, we are finding 

 that the correction of water depth information that was usually 

 entered manually in Systems I and II, are taking more time to 

 correct than was spent in gathering the data. In System III we are 

 dispensing with all routine manual inputs. Those values that are 

 needed and their entry is not yet automated, will be digitized later 

 ashore using conventional digitizers. Because, as mentioned 

 previously, time and date are the means by which we identify 

 individual data points, we have obtained a calendar clock to provide 

 that information to the system, thereby eliminating the necessity of 

 relying upon a person to enter such information. Should the clock 

 or other inputs fail, we will have provision, of course, for manual 

 input as a backup measure. 



The program for Systems I and II was written in an assembly 

 language which at the time was the only language capable of 

 sampling analog input channels, BCD input channels, branch 

 indicators, processing interrupts, and activating contact closures. 

 During the development of the Systems I and II, the real-time 

 program grew piecemeal as ideas for improvement and expansion 

 of the computers capability developed. A complex interleaving and 

 interconnecting of various parts of the program is a most 

 undesirable way for the computer program to grow. In the latter 

 stages of our utilization of System II we had reached the state where 

 even the programmer who wrote the real-time program was afraid 

 to make modifications or additions for fear it would jeopardize the 

 operation of the system. One still unexplained mystery that may 

 in part have resulted from program interaction was the inability 



260 



