produce contour maps on an incremental plotter or on the regular line 
printer. The line printer routines proved particularly useful since 
they were practically machine and output device independent, were able 
to produce maps concurrently with the rest of the job, and were econom- 
ical. 
  
Economy of computing was considered in terms of time (or speed) of com- 
putation and storage requirements. To some degree time and storage are 
interchangeable since greater speed can sometimes be obtained at the 
cost of larger storage demands. The programming methods and languages 
used may also affect time and storage requirements; thus the economy and 
machine independence goals are somewhat contradictory. Since GCARS I 
required matrix representation of models and performed minimum path 
analysis on these matrices, storage requirements were largely associated 
with the matrix size while computation times depended on both the matrix 
sizes and the efficiency of the minimum path algorithm. 
The minimum path algorithm portions of GCARS I System were adapted from 
Martin's FORTRAN coding of the British Road Research Laboratory algo- 
rithm (3). The adaptations improved the efficiency of the algorithm by 
a factor of at least four. The GCARS I system was designed to analyze 
matrices up to 2500 points and generate five alternatives within five 
minutes of computer time while using only moderate core storage (around 
165K bytes on an IBM/360). Thus an analysis would cost somewhat less 
than $20.00 on many installations. 
The goal of flexibility (Goal 4) appeared fairly easy to satisfy if one 
assumed that all factors could be measured on some type of value scale. 
The term "costs" and "cost models" in the description of the basic sys- 
tem concept were placed in quotes to indicate that a much broader concept 
than pure monetary values was intended. At the time of GCARS I develop- 
ment there was considerable discussion concerning the measurement of 
value (1, 4, 5, 6 and 7). O'Flynn (7) discussed the problem at some 
length and concluded that "the most suitable approach is to outline the 
precise physical magnitudes of the non-market outputs". The initial 
GCARS System accepted this logic. 
The sensitivity (Goal 5) of each analysis was measured in the initial 
GCARS System by comparing the path totals of each alternative to the 
first choice path total. A series of ratios were thus obtained and 
displayed along with the lengths of each alternative as shown in 
Figure 3. The engineer in charge of the study could use these ratios 
to measure the sensitivity of the corridors selected for any particular 
factor or factor combination since rapidly increasing ratios indicated 
a single, narrow, well-defined optimum band. However, no equivalent 
quantitative figure was developed to compare the routes generated for 
different factor combinations. The supplied sensitivity measures de- 
liberately did not take into account any "route-dependent" factors, 
such as maintenance or operating costs, since after careful consider- 
ation it was concluded that the engineer was in the best position to 
make such judgment evaluations. 
The GCARS I system design did not specifically consider the compatibility 
goal (Goal 6). However since it was proposed to use GCARS as a supple-