Re: additional draft notes on the QT data model

"C. M. Sperberg-McQueen" <cmsmcq@acm.org> writes:

> At 2003-09-03 14:43, noah_mendelsohn@us.ibm.com wrote:
>
>>Michael:
>>
>>I generally agree with what you've written, but this one surprises me a
>>bit:
>>
>> > 3.6 p -1 (last p before 3.6.1) suggests that attributes
>> > from the XSI namespace appear as ordinary attributes in
>> > the data model.  It may be worth pointing out to the
>> > reader that in the normal course of events, such
>> > attributes will not appear in the PSVI even if they
>> > appeared in the original XML data stream.  This also
>> > has implications for re-serialization.
>>
>>I thought we had a rule that schema assessment never eliminated from the
>>PSVI an info item or property that had been present in the input Infoset.
>>Are these attributes not carried over from the input Infoset?
>
> Hmm.
>
> I think you are probably right; I was probably thinking about the
> fact that if a validator calls a function to check all the attributes
> on an element, the function must ignore them.  I over-generalized
> this 'ignore them' into a 'suppress them', which is indeed in conflict
> with our general rule.
>
> I find I don't know for sure what form they show up in the PSVI in, though
> (whether they have PSVI-only properties or not).  The DM spec seems
> to expect them to have been validated; that I don't think likely.

Well, it's not clean, but piecemeal the information that validation is
expected and/or required is there.

In Element Locally Valid we have

  "If . . . there is [an xsi:nil] attribute information item and its
  *actual value* is true"

Actual values require validation, so there's a strong implication.

Also,

  "The *normalized value* of [an xsi:type] attribute information item
  must be *valid* with respect to the built-in QName simple type"

That's clear enough.

Section 2.6 says "All schema processors have appropriate attribute
declarations for [the xsi:] attributes built in".  And section 3.2.7
makes this explicit in detail.

Section 5.1 refers to the *actual value* of xsi:schemaLoc and
xsi:noNsSchemaLoc.

So I think a an issue is in order, to clarify that these four must be
validated using the built-in declarations whenever they appear, and
must be valid, and get PSVI-only properties.  Needs an extra clause in
Element Locally Valid (Complex Type), and possibly in Schema
Information Set Contribution: Assessment Outcome (Attribute).

ht
-- 
  Henry S. Thompson, HCRC Language Technology Group, University of Edinburgh
                      Half-time member of W3C Team
     2 Buccleuch Place, Edinburgh EH8 9LW, SCOTLAND -- (44) 131 650-4440
	    Fax: (44) 131 650-4587, e-mail: ht@cogsci.ed.ac.uk
		     URL: http://www.ltg.ed.ac.uk/~ht/
 [mail really from me _always_ has this .sig -- mail without it is forged spam]

Received on Thursday, 4 September 2003 04:48:30 UTC