I won't be so strict. Even though a homebrew implementation won't be widely interoperable, an experience of its active development and use in a limited environment (e.g. within a company) would be valuable both to inform the RFC and to serve an example implementation.