Today we say:

> Rational numbers that have a decimal representation. See below about the precision of the number
> Regex: -?(0|[1-9][0-9]*)(\.[0-9])?([eE][+-]?[0-9])?

> XML Representation: union of xs:decimal and xs:double (see below for limitations)

> JSON Representation: A JSON number (see below for limitations)

So in the XML + JSON representations, we constrian these to being xs:decimal (at least 18 digits, on my reading) or xs:double (64 bit floating point) – but at the level of FHIR (e.g., for other serializations, or for internal models) they're jsut... arbitrarly large, arbitrarily precise values? Does a FHIR implementation need to cope with unlimited size and precision? Is there a reason we only state expectations on serialization formats? For the other data types, we explicit list details under "Value Domain", like:

> Note that strings SHALL NOT exceed 1MB (1024*1024 characters) in size

> A signed integer in the range ?2,147,483,648..2,147,483,647 (32-bit; for larger values, use decimal)