From the definition of the multibase datatype:
The canonical mapping consists of using the lexical-to-value mapping.
This is wrong, because the canonical mapping is expected to be in the other direction (value-to-lexical). Furthermore, as the lexical-to-value mapping is not one-to-one (the same value has several lexical forms, in different bases), there is no trivial extrapolation (such as inferring that we are talking about the "reverse" mapping).
If a canonical mapping was to be defined, the spec would have to pick a default base (I assume that, for a given base, the mapping is one-to-one, so that would be enough). But an alternative is to not define a canonical mapping at all -- this is not required in the definition of a datatype.
From the definition of the
multibasedatatype:This is wrong, because the canonical mapping is expected to be in the other direction (value-to-lexical). Furthermore, as the lexical-to-value mapping is not one-to-one (the same value has several lexical forms, in different bases), there is no trivial extrapolation (such as inferring that we are talking about the "reverse" mapping).
If a canonical mapping was to be defined, the spec would have to pick a default base (I assume that, for a given base, the mapping is one-to-one, so that would be enough). But an alternative is to not define a canonical mapping at all -- this is not required in the definition of a datatype.