A medium-sized computer requires considerable justification 

 with particular attention to the number of users and the types 

 of problems encountered. Great care must be exercised in long 

 range plans which incorporate the computer into future research. 

 Considerable effort must be spent in making this type of computer 

 system accessible to the scientist. 



The computer center facility can be justified only if the lonK 

 range plans of several scientific groups can project 5 to 10 years 

 of full-utilization of the facility. Careful thought must be 

 given to the fact that many smaller computers might outperform 

 the giant. Such large computer systems might be better on the 

 shore with only dedicated or medium-sized computers participating 

 in the "at sea" phase of operation. 



We must be very careful in allocation of our seagoing 

 scientific manpower. Survey jobs are probably more effectively 

 analyzed on shore with a seagoing computer performing data 

 manipulation and consistency checks. The temptation in this area 

 is to obtain a computer center facility and do all the processing 

 at sea. This would be a mistake, since survey experiments can be 

 well planned in advance and data checks can be made with simple 

 computers on a routine basis. Our costly computer science 

 specialists at sea should be allocated to problems requiring 

 creativity and careful co-ordination of computer and experiment. 

 Most creative people will simply not spend significant fractions 

 of their life at sea. 



The decision on whether to do on-line or off-line processing 

 hinges on the requirements of a particular experiment. Program- 

 ming languages should be developed which allow the maximum 

 flexibility to the scientist. The scientist using an on-line 

 facility should be able to bring cables from his sensor apparatus 

 to the computer and proceed from there to data acquisition and 

 analysis via simple programming steps. This programming is pre- 

 ferably written in a secondary language such as FORTRAN or ALGOL. 

 Thus the scientist wishing to make changes in his experiment does 

 not have to resort to rewiring electronics but to altering a 

 computer program. 



The medium-sized computer is best suited to this flexible on- 

 line operation because it can combine the interface capability of 

 the dedicated computer with a more easily operated compiler. The 

 medium-sized computer generally has punched card input-output and 

 a printer to facilitate programming. There are many who feel 

 that eliminating paper tapes and typewriters justifies the 

 additional cost of the medium-sized computer. 



38 



