This is an archived snapshot of W3C's public bugzilla bug tracker, decommissioned in April 2019. Please see the home page for more details.
The current draft of the precisionDecimal note does not say what namespace the datatype is in (at least not that anyone in the meeting could see). This should be required as part of the documentation for implementation-defined datatypes, so perhaps this will automatically be taken care of as part of the resolution of bug 13421. But if that turns out not to be the case, then that's a gap in the requirements for implementation-defined datatypes. (The rules for implementation-defined types say a simple type definition is required; that seems to entail specifying a namespace, but it might be helpful to be more explicit.) (There is also some question about how to make a conforming simple type definition for a primitive, given that the base type has to be xs:anySimpleType or xs:anyAtomicType, which aren't usually legal. Further investigation needed.)
Discussing this on the phone with Dave Peterson today, it seemed to me that the right thing to do, both with respect to the namespace issue and with respect to the question in the final parenthetical remark, is to include in the definition of any implementation-defined primitive a tableau analogous to those in section 4.1.6 (http://www.w3.org/TR/xmlschema11-2/#builtin-stds) of the Datatypes spec. Such a tableau will (among other things) specify explicitly the target namespace and base type of the new primitive. Dave Peterson points out that it would not be a bad idea to say something like this explicitly in the Datatypes spec. In appendix H, change item 3.c from c. The values of the ·fundamental facets·. to c. The values of the ·fundamental facets· and of all applicable component properties.