NXDL: use xs:decimal instead of xs:float or xs:double?

[woutdenolf]

We are using

• Single-precision decimal number: https://www.w3.org/TR/2004/REC-xmlschema-2-20041028/datatypes.html#float
• Double-precision decimal number: https://www.w3.org/TR/2004/REC-xmlschema-2-20041028/datatypes.html#double

Did we really intend to restrict the precision? If not we could use

Arbitrary-precision decimal number: https://www.w3.org/TR/2004/REC-xmlschema-2-20041028/datatypes.html#decimal

[woutdenolf]

It seems that xs:decimal does not support Nan and Inf which might be why we don't use it.

[woutdenolf]

As far as I understand

System / Concept	What it is	Example values
xs:decimal (XML Schema)	Arbitrary-precision base-10 decimal type	10.5, 0.1, 123456789.123456789
xs:float (XML Schema)	32-bit IEEE 754 floating-point	0.1, 3.14, 1.2345e3
xs:double (XML Schema)	64-bit IEEE 754 floating-point	0.1, 3.141592653589793, 1e308
Python float	64-bit IEEE 754 floating-point (same as double)	0.1, 2.718281828, 1e-9
Rational numbers (ℚ)	Fractions of integers a/b	1/2, -7/3, 22/7
Real numbers (ℝ)	All rational + irrational numbers	π, √2, -5, 0.333...

I guess xs:decimal only applies to string representations.