Chapter 30 ½ The SYMBOL Computer 507

we estimate approximately a six percent saving in the manufacturer s support expense. This is illustrated in Fig. 6.

Good service is a must in a large system. The SYMBOL hardware has been engineered for good reliability and at the same time easy maintenance. We do not anticipate any added expense for SYMBOL hardware maintenance over conventional systems with equivalent storage and logic circuit counts. Our experience on the SYMBOL model has verified this belief.

The previous material has split the computing dollar up in parts and has described how major savings can be realized with a "total systems" approach. The SYMBOL techniques described herein together with good time-sharing, conversation mode practice can reduce computing costs up to 50 percent. Referring to Fig. 7, one may visualize how the savings in the whole computing pie add up.
 

Conclusion¹

The traditional boundary between hardware and software has been weakened during the past ten years and is due for a significant shift beyond the token improvements. It is believed that in SYMBOL a major step towards significantly more capable hardware has been attained.

The SYMBOL system is now entering an extensive evaluation phase where the system's strengths and weaknesses will become more apparent through actual day to day usage. The developers of

the system have gained much insight into the merits of each of the approaches taken. The overall approach to memory management is considered a breakthrough. The moving of data attributes from instructions to the data is considered fundamental.

No claim is made that the SYMBOL system has been balanced for optimum performance and use of hardware. Certain critical areas of memory management and system supervision are felt to be 10 to 100 times more efficient than conventional means. Certain aspects of structure referencing are a major advance over software list processors but fall short of being competitive for some types of large array referencing. Many of the weaknesses in this first SYMBOL model were solved by the designers too late to be factored into the actual hardware. Many other aspects of the system such as the paging and system supervisor algorithms can be evaluated after significant usage experience.

The computing professionals have debated for many years the questions: Can a compiler be developed in hardware? Can the heart of system supervision be committed to hardware? Can data space management be taken over by hardware? Can hardware be designed to take over major software functions? Can complex hardware be debugged? These and many other questions have been positively answered with the running SYMBOL system. The most significant part of the entire project is that the concepts were reduced to full scale, operating hardware.
 

References

Chesley and Smith [1971]; Rice [1967].

¹This conclusion is taken from the paper by Smith et al. that makes up the first part of this chapter.