ABSTRACT
The MAD project was started in 1981 at CERN to provide a flexible framework for particle optical computations. From the beginning the following features were considered important:
Open-ended and modular design of the program to allow easy addition of new features.
Format-free input language to facilitate data preparation. Later the requirement was added that the language should be understood by a variety of other programs. When moving data to other programs this would avoid extensive translations of the data with the corresponding danger of making errors.
The internal data structure should describe the machine in a flexible manner and access to this structure should be simple.
The program should be easy to maintain.
Canonical variables should be used throughout the computations.
