The paper is not arguing for safe 'conventions'.

It describes a (sketch of) a statically provable memory safe dialect of c++. A more detailed description is in one of the referenced papers by Herb Sutter.

Unfortunately the promised static checker is, as far as I know, yet to be released, and we will have to see how well it will work in practice and whether it is feasible to piecemeal add the safety annotations to an existing codebase.