to be a separate program module. Using terminology of Ada, 
that kind of module is called a package. Typically, the 
package consists of a definition part and an implementation 
part. In this context, the definition part can be 
considered to be the data base schema, the user’s view of 
the data. And the implementation part corresponds to the 
physical implementation of data base. 
This approach requires extra programming effort. However, 
taking the current computers, typically having large 
virtual memory systems, the data base can very often be 
implemented by using data-arrays, each corresponding to a 
relation (when relational data base model is used). This 
approach is often suitable, because the volume of the data 
we are dealing with is usually limited. 
Because of the separate definition and implementation 
parts, the data structures can be modified later in the 
system development phase without affecting the 
application’s view of the data. In the case of using the 
prototype approach in system design, this is of great 
advantage. 
6.1.3 Abstract data bases as programming paradigm 
Above the concept of data bases has been used when speaking 
of data that has some meaning also outside the application 
program. However, the concept can also be used as a 
programming paradigm for storeing any internal data. The 
package approach from Ada is still valid; we define the 
data and the access methods that are visible for the 
outside world. A great deal of flexibility is offered on 
all levels of program hierarchies. On the top level, this 
has been pointed out by Elassal (1985) and Hatzopoulos 
(1985). 
6.2 Abstract processes 
Abstract data bases can be classified to be abstract state 
machines. Data is stored in the state vector of the machine 
and the outside world sees the machine trough the access 
functions. Other kind of abstract machines are abstract 
processes. The difference (even small) is in the way how 
the machine functions with respect to time. In an abstract 
data base, we can usually assume that the state vector is 
static between two access calls. In an abstract process, on 
the other hand, the process can be active also between the 
time points of interaction, the process carrying out 
independent work that changes its state. 
Abstract processes can be used for several computational 
tasks. For example, bundle block adjustment for aerial 
triangulation can be described conveniently by using 
abstract processes. Assuming the usage of the conventional 
method applying the pointwise treatment of observation 
equation and the reduction of normal equation system such 
77597.