This paper proposes a definition of 'memory-safe language' as one containing a syntactically explicit, compiler-enforced subset that is systematically free of all undefined behavior, and argues that heuristic reduction of UB through profiles, hardening, or contracts alone is insufficient to meet this definition. It contends that C++ should pursue a subset-of-superset strategy analogous to Rust and Circle, whereby new language features enable a useful, UB-free subset while existing unsafe code remains accessible outside it. The paper seeks EWG consensus on whether C++ should commit to becoming a memory-safe language by this definition, with lifetime safety identified as the primary technical challenge requiring significant new language features.