All clients need to agree on some standard, yes. I suppose as first mover you get quite a lot of say in what happens...
Don't think I have the experience or gravitas for that but will use RIPEMD160(SHA2(x)) for now.
I feel we need a BOLT for cyclic superhubs or at least a bit more discussion on the MLs.