2.1.4 [W3C-XSD] Section 3.2.5, double
V0003:
-
[Definition:] The double datatype is patterned after the IEEE double-precision 64-bit floating point type [IEEE 754-1985]. The basic ·value space· of double consists of the values m × 2^e, where m is an integer whose absolute value is less than 2^53, and e is an integer between -1075 and 970, inclusive. In addition to the basic ·value space· described above, the ·value space· of double also contains the following three special values: positive and negative infinity and not-a-number (NaN). The ·order-relation· on double is: x < y iff y - x is positive for x and y in the value space. Positive infinity is greater than all other non-NaN values. NaN equals itself but is ·incomparable· with (neither greater than nor less than) any other value in the ·value space·.
MSXML6
No error is reported if the NaN
value is compared with positive infinity or negative infinity.
V0004:
-
Any value ·incomparable· with the value used for the four bounding facets (·minInclusive·, ·maxInclusive·, ·minExclusive·, and ·maxExclusive·) will be excluded from the resulting restricted ·value space·. In particular, when "NaN" is used as a facet value for a bounding facet, since no other double values are ·comparable· with it, the result is a ·value space· either having NaN as its only member (the inclusive cases) or that is empty (the exclusive cases). If any other value is used for a bounding facet, NaN will be excluded from the resulting restricted ·value space·; to add NaN back in requires union with the NaN-only space.
MSXML6
If NaN is used as a facet
value for a bounding facet, when NaN is compared
to other non-NaN doubles, NaN
is less than any non-NaN number.