This paper describes the design and implementation of a database management system (DBMS) called MIDAS, which is aimed specifically at engineering design applications. MIDAS could in no way be described as a general-purpose DBMS. The authors claim that MIDAS/R, a subset of MIDAS, is a relational DBMS. However, database researchers would find little to substantiate this claim. It is, of course, fashionable to call every data set stored on a secondary storage device a database and every two-dimensional flat-file a relation] The authors also state that their Data Definition and Data Manipulation Languages (DDL and DML) are user friendly. However, even with the most liberal interpretation, embedding DDL and DML commands as calls to external procedures embedded in FORTRAN would not be regarded as user friendly.

The MIDAS system, therefore, must be judged strictly in the context of its relevance and usefulness to providing support for engineering design applications. MIDAS is essentially an extension to RIM, a system designed specifically for scientific data management. The extensions include support for multiple databases and large matrices. MIDAS itself has two subsystems: MIDAS/R, the “relational” database system, and MIDAS/N, a numerical data management system. Most engineering applications today are still written in FORTRAN and there is, therefore, no doubt that any data management system operating in this field must provide interfaces to FORTRAN. However, this does not necessarily mean that the system itself should be written in FORTRAN. It is difficult to imagine anyone writing a true relational DBMS in FORTRAN.

There can be no doubt that the requirements imposed by engineering systems in DBMSs are very different from conventional business systems. There is therefore a good case to be made for providing separate data management facilities for engineering applications. Just as it is difficult to imagine anyone writing a full, truly relational DBMS in FORTRAN, so is it difficult to imagine how anyone could implement efficient storage and retrieval of matrices using, for example, SQL]

In conclusion, MIDAS, judged in the engineering context, appears to offer a reasonable solution to the data management problems of engineering systems design. It addresses the areas where conventional business DBMSs are generally lacking--efficient manipulation of large amounts of numerical data, convenient interfaces to FORTRAN, and so on. The paper itself is somewhat terse in style and would have benefitted from more thorough editing. It was also not clear how the two subsystems (MIDAS/R and MIDAS/N) interact. The two appear to be quite separate, with databases in MIDAS/R and data sets in MIDAS/N. In the light of the current research interest in mixed-mode databases, it will be interesting to see whether specialized systems such as MIDAS will be made redundant.