This paper, a shorter version of an earlier paper [1], proposes a FORTRAN 66 extension based on the computer arithmetic model in an earlier work by Kulisch [2].

The proposal adds real and complex intervals as intrinsic data types, and allows forming one, two, and three dimensional arrays of the intrinsic types. These arrays can be assigned values and used in expressions. The proposal includes many new intrinsic array and arithmetic rounding control operators.

The author calls his effort “a working for implementations,” but comparing it to the recent FORTRAN 8x proposal shows some flaws. While both stress array processing, the 8x proposal carefully considers the affect of FORTRAN’s notorious storage association features on the array extensions; this proposal does not. The 8x proposal adds powerful data abstraction and control flow extensions; this proposal adds nothing there to FORTRAN 66 (and pays no attention to FORTRAN 77). But the 8x proposal doesn’t have an underlying computer arithmetic model. This proposal is more elegant in that regard, but seems less practical overall.