>>16842769
But can't we also say this exact thing about rejecting mappings from one to many? We can clearly recognize we're better off without them... I don't see why one-to-many is any preferable just because it's slightly nicer for our chosen inference rules.