Fortran must die
Every once in a while a statistician says something that I wish I had said. Because it is so obvious, but so commonly ignored in the literature I have seen.
Just like normality is a useful statistical property that is rarely found in real world data according to various authors.
The notion of extrapolation proper implies that future development of the process under study will be the same as in the past and the present. In other words, it concerns stationarity of the process. Stationary processes are very attractive from the forecasting standpoint but they unfortunately do not exist in nature, as any real process is subject to change in the course of its development.