Making RELAX NG schemas normative rather than non-normative

[Rex Jaeschke]

Re the root element problem: If this can be stated in XSD then why don't we?
If it can't then how does the (very large) XSD world deal with this supposed
limitation? Is this widely recognized to be such a big problem that W3C is
doing something about it for XSD?

Re constraints: I understand that some schema languages are better at
expressing certain things than are others. If XSD is deficient in this area,
presumably we have written text to deal with constraints that aren't
expressible in the schema. As such, making RELAX NG schema normative won't
reduce the need for that written text.

Re VML drawing parts cannot be validated against XSD schemas: Assuming that
is so, wasn't the whole point of the BRM's creating Part 4/Transitional to
help steer  implementations to supporting Strict (which excludes VML)? That
is, why spend time/effort doing something to further support the approach
that is being deprecated?

We also need to hear from implementers.

Rex



> -----Original Message-----
> From: eb2mmrt at gmail.com [mailto:eb2mmrt at gmail.com] On Behalf Of MURATA 
> Makoto
> Sent: Wednesday, March 16, 2011 5:08 PM
> To: SC34
> Subject: Re: Making RELAX NG schemas normative rather than non- 
> normative
> 
> Rex,
> 
> > That aside, more importantly, I'd like to know just what real-world 
> > problem does the making RELAX NG schemas normative actually solve?
> 
> The answer is shown in the log of DR-10:0030, which I referenced.
> 
> >The RELAX NG schemas specify root elements of each OPC part, while 
> >the XSD schemas do not. For example, DrawingML_Chart.rnc specifies 
> >that the root element of the OPC part DrawingML Chart is a chartSpace
element.
> 
> >The RELAX NG schemas use the & operator for allowing different 
> >elements without specifying constraints on their sequences, but it is 
> >not always possible to use the xsd:all construct in W3C XML Schema.
> >For example, wml.rnc uses the & operator in w_EG_RPrBase, but wml.xsd 
> >does not use xsd:all in EG_RPrBase.
> 
> >The RELAX NG schemas represent co-occurrence constraints between 
> >elements and attributes. For example, pml.rnc specifies that the pic 
> >element and the attribute spid in p_CT_OleObject are mutually 
> >exclusive. Meanwhile, pml.xsd simply allows both in CT_OleObject.
> 
> >VML drawing parts (Part 4, §8.1) can be validated against RELAX NG 
> >schemas, but cannot be validated against XSD schemas. This is because 
> >there are no XSD schemas for the unqualified xml element, which is 
> >the root element of VML drawing parts.
> 
> Regards,
> Makoto <EB2M-MRT at asahi-net.or.jp>