212 
  
  
  
4  Ggecond Level Application Systems (A2) 
Second level application systems provide grist for the paper 
mill, more of which should be carried out, considering our 
relative innocence in employing digital image analysis 
techniques on remotelu-sensed imagery. Here is uhere highly 
computer-literate analusts wish to specify new algorithms 
for rectification, enhancement and especially for 
classification, as well as improving existing algorithms. 
Programmers are hired to design code, test, debug and 
struggle with mysterious things such as FORTRAN, JL, 
RSX-11M. TASKLINKER, VMS. DMAs and system crashes. Who has 
time to consider mundane things such as the user interface, 
when one can so directiy challenge the MACHINE? 
The AZ user (programmer) interface” Things are looking even 
worse. There "is9n't one. I have met several programmers 
that use major. commercially available image processing 
systems. When they add new processing capabilities to the 
system that their employer has purchased they all use the 
same basic technique —— start over! 
This technique is fool-proof. You don’t have to struggle 
through reams of misspelled documentation. You can avoid 
the originator ’s design flaws, and generate some new ones of 
your own. You are as free as Robinson Crusoe. You can keep 
your Job as long as you like. | And you can still blame your 
problems on the original supplier anyway. 
J. Systems Design for Understandability - Pl thru A2 
All data analysis systems, in particular the ones we 
describe here, require design; implementation and 
maintenance. If anything goes wrong during any of these 
three phases, the system will see a very short life cycle, 
and the end-user would probably have spent fewer dollars if 
he had used more traditional techniques. 
Furthermore, it is better to say "I don't know” than to 
quote results that rest on more assumptions than anu one 
person can possibly verify. What I am saying here is that 
one must have a very confident grasp of what these systems 
{especially Al and AZ) are actually doing to the data in 
order to specify a processing sequence and provide a lucid 
report of subsequent results. 
Here, then, is the point of this paper and the goal of the 
system design that I am about to describe: 
The user interface is not only critically important, but 
it extends well beyond slick graphics, convenient cursor 
movement, and analust's-console understandabilitu,