<p>It is implicit and ambient in the earliest scientific computing languages that the words of the machine are the box in which data fits, and so one needs “types” to class data when your computer gets new fancy boxes for doing e.g. fixed point arithmetic. On the business side, it takes a few revisions for FORTRAN to gain simple numeric types-as-classifiers and arrays. ALGOL 58 has type declarations after a few documented reports. These efforts feel - and are often described - as utterly ad hoc tho</p>