The aim of this section is not to determine which implementation of the 3-D grid object is the best, but rather to make a proposal of a new definition of a grid object more convenient to use for geological and/or geophysical purposes.
A quick look at the parameters file given as an example in the previous section readily shows that I have deviated a little from the original definition of the GRID3_t object as it is defined in GOCAD version 8.1. The reason for this is that Martin and I wanted to use a kind of SEP input format, since we had decided to use the SEP library to produce the output.
The first modification concerns the definition of the grid itself. Because in
most of cases, the grids have a vertical axis corresponding to the depth or
the time variable, it is not necessary any more to specify a vector for this
direction. On the other hand, it is useful to allow the horizontal directions
to be a linear combination of the usual vectors 
and 
,respectively unit vectors for the x and y axes. Instead of computing the
length of the generator vectors of the grid to obtain the sample on each axis,
it seems more convenient to give directly these values in the input.
Another important feature is that it is often necessary to keep more than one value at the nodes of a grid, for example, densities and porosities. It is convenient for the efficiency of the finite-difference algorithms and for problems of memory size, to have one grid with several variables attached to each node rather than several grids, each containing the values of only one variable.