Would the following changes resolve this issue satisfactorily?

1 In section 2.4.2 Special vs. Primitive vs. Ordinary Datatypes,
expand the existing note from 

    Note: As normatively specified elsewhere, conforming processors
    must support all the primitive datatypes defined in this
    specification; it is ·implementation-defined· whether other
    primitive datatypes are supported.

to

    Note: As normatively specified elsewhere, conforming processors
    must support all the primitive datatypes defined in this
    specification; it is ·implementation-defined· whether other
    primitive datatypes are supported.

    Processors MAY, for example, support the floating-point decimal
    datatype specified in [Precision Decimal].

2 In section 2.4.3 Definition, Derivation, Restriction, and
Construction, expand the first existing note from 

    Note: The properties of any ·implementation-defined· ·primitive·
    datatypes are given not here but in the documentation for the
    implementation in question.

to

    Note: The properties of any ·implementation-defined· ·primitive·
    datatypes are given not here but in the documentation for the
    implementation in question.  Alternatively, a primitive datatype
    not specified in this document can be specified in a document
    of its own not tied to a particular implementation; [Precision
    Decimal] is an example of such a document.

3 In section 3.3 Primitive Datatypes, immediately before section
3.3.1 string add a new note immediately following the paragraph that
now reads

    Conforming processors MUST support the ·primitive· datatypes
    defined in this specification; it is ·implementation-defined·
    whether they support others. ·Primitive· datatypes may be added
    by revisions to this specification.

        Note: Processors MAY, for example, support the
        floating-point decimal datatype specified in [Precision
        Decimal].

4 In the non-normative references, add

    Precision decimal

    World Wide Web Consortium. An XSD datatype for IEEE
    floating-point decimal, ed. David Peterson and
    C. M. Sperberg-McQueen. W3C Working Group Note
    to-be-published-rsn. Available at:
    http://www.w3.org/XML/Group/2004/06/pD/precisionDecimal.html
    (member-only link; to be published; this citation will be
    updated silently).

Sometimes I've thought we might also wish to mention the
precision-decimal document at the places where we talk about
implementation-defined facets (specifically the bottom of 4.1.5),
but I don't see a good hook there this morning.

This proposal has not had editorial review, but I'm taking the 
liberty of marking it needsReview anyway; the other editors should
not be held responsible for any flaws in the proposal.

> MSM suggested such a pointer could be
> in the discussion about implementation-defined
> datatypes.

Yes, though it makes a lot of sense to me to also/instead have a brief mention and link from notes in at least the decimal type section, and conceivably float/double too (on the theory that pD is both a decimal type, and unified in its IEEE specification with the binary floating types).

I wouldn't make a big fuss if you decide to stick with just implementation-defined, but I do think at least some of the other mentions make at least as much sense. Thank you.

Noah

amendment:  After para 1 of 3.3.3 decimal, add "Note:  For a decimal datatype whose values do reflect precision, see [Precision Decimal].

Resolution: adopt the proposal as ammended.

The proposal adopted on 6 May (as amended) has now been integrated into the status-quo documents, so I'm marking this issue resolved.

Dave, if as the originator you would check that the fix has taken and then close or reopen this issue?  Thanks.

looks OK, pending publication of the separate Note which will entail changing the pointer thereto.