I don't see what your aversion is to using the edge split modifier. Modifiers by definition are nondestructive. The Prism3D geometry format stores one set of normal, tangent, vertex colour, UV data, etc per vertex. For the exported model to have sharp edges, the exporter needs to split them. For the exporter to know it needs to split them, it relies on the edge split modifier.
If you are following good practices, you will be working from a master blend file (or several in a large project), and re-exporting models as necessary. So edge splits (using the modifier) shouldn't have any detrimental effect on your workflow, other than the minor inconvenience of marking sharp edges.
