This doesn't seem to be a new thing generally. Variable length arrays (if you like) are just what you have in Fortran. If you support scientific processing, you'll often have seen mysterious SEGVs due to people insisting on using Intel ifort, which allocates arrays on the stack by default (or used to?), and doesn't fare well with typical default stack limits. This is the equivalent of Gfortran's -fstack-arrays, which is only turned on by -Ofast. (Recursive procedures use the stack anyway.)