The paper describes a generalized numerical database translator. Numerical databases refer here to large, dense, and sparse matrices. A proposal of a data model for numerical databases is attempted. The use of data model concepts is confusing; therefore, we feel the need to make the following clarification:
(1) Logical schema would describe the structure of a matrix, i.e., its type and dimension.
(2) Logical data representation is the usual matrix.
(3) Physical schema would describe the physical representation type, i.e., the compaction type and the access type.
(4) Physical data representation is encoded matrix, stored on some secondary storage.
Points (1) and (2) above form the Stored Data Description Language (SDDL), while (3) and (4) form the Stored Data Mapping Language (SDML). I note the lack of a manipulation part, meaning the definition of operations between matrices of same or different types. The main concern of the paper is the problem database translation. It is not clear whether the author defines the translation as being done within a same data model, or whether he views the different types of matrices as defining different data models. I contend that the first view is the correct one; therefore, the described process refers to logical schema restructuring and constructive database mapping, i.e., database reorganization.
Although its formal precision shows some flaws and the paper is lacking in examples, it is worth reading. I see it as an interesting application of various data modeling aspects to the peculiarities of an unusual environment.
