RE: [w3c sml] [ISSUE 4684] sml:keyref reference to xs:key/unique

This should also cover 4995 (which blocks 4684).

A couple of minor comments.

1. In the example, you mentioned: now consider a "new model". I think you 
didn't mean a new SML model, but ration a different portion 
(definitions+instances) of the same SML model.

2. About the issue at the end. To make sure the implication is clear. A 
QName means that the scope has to be specified by a direct child of the 
current element with that QName. I like the simplicity, but not sure 
whether this would be over-restrictive. I think we may want to use the 
same XPath allowed on <xs:field>. This doesn't add (much) burden to 
implementations, because they already need to evaluate selector/field 
XPaths.

Thanks,
Sandy Gao
XML Technologies, IBM Canada
Editor, W3C XML Schema WG
Member, W3C SML WG
(1-905) 413-3255 T/L 969-3255
 



"Wilson, Kirk D" <Kirk.Wilson@ca.com> 
Sent by: public-sml-request@w3.org
2007-09-07 04:37 PM

To
<public-sml@w3.org>
cc

Subject
RE: [w3c sml] [ISSUE 4684] sml:keyref reference to xs:key/unique






With assistance from Sandy, I think I?m now prepared to make a proposal on 
how to allow sml:keyref to refer to xs:key/xs:unique.  See the attached 
discussion.   Following discussion and if the proposal is accepted, I will 
submit a write up to the editors and provide the necessary change for the 
spec.
 
To briefly remind you of why we want to allow sml:keyref to refer to 
xs:key/xs:unique:  The possibility exists that existing XML documents can 
be used as parts of a new SML model.  That is, the SML model builder may 
have to opportunity to build an SML model from documents (schema and 
instance documents) that were built without any awareness that they might 
have a role to play in an SML model.  The elements in these document, 
which the SML model will want to reference in the new model, may already 
be defined with xs:keys or xs:uniques.  It seems to be a wasted exercise 
to require the SML model builder to essentially reconstruct the sml:key or 
sml:unique definitions in the SML model.  Therefore, we should provide the 
ability of an sml:keyref in the new model to reference the keys and 
uniques found in the existing XML documents that are being referenced by 
the new model. 
 
As will be obvious from the attached, I owe Sandy credit for the 
recommended solution to the problem?a new attribute for specifying the 
scope of sml:keyref in the presence of inter-document references.
 
Kirk Wilson, Ph.D.
Research Staff Member
CA Labs
603 823-7146
 

From: Wilson, Kirk D 
Sent: Friday, August 24, 2007 3:09 PM
To: public-sml@w3.org
Subject: [w3c sml] [4684] sml:keyref reference to xs:key/unique
 
All,
 
This email will serve to start the discussion of 4684 regarding whether 
sml:keyref should be able to reference xs:key/unique elements.  I must 
admit that I took this bugzilla because the basic idea presented in the 
following paragraph rather interested me.  However, assessing this idea 
turned out to be a tremendous undertaking because I don?t understand 
enough about the background structures being maintained in the PSVI.
 
This is an issue undoubtedly because the current SML spec is silent on 
whether sml:keyref can refer to xs:key/unique elements.   I would assert 
that there is a powerful reason for allowing this: namely, to allow SML 
models to be built from existing models that already have xs:key/unique 
structures defined for the elements that need to referenced in the SML 
model.  Inter-document references could utilize identity constraints over 
existing keys in the referenced documents based on existing keys. 
Therefore, I believe that issue here is whether there is anything in XML 
Schema preventing this kind of identity constraint.
 
The issue, I believe focuses upon the XML Schema rule that the refer 
attribute of xs:keyref should refer to a xs:key/unique element defined 
under the same element or under one of that element?s children. Presumably 
this same rule holds for sml:keyref and sml:key/unique?at least that is 
the way I understood an answer I was given when I raised the question of 
whether sml ICs must comply with the same structural restrictions as XSD 
ICs during the first F2F.  How these tables can interrelated should be 
discussed within the WG.  On the other hand, the refer attribute of 
sml:keyref is a QName and could presumably identify a xs:key/unique in 
another namespace.
 
So, I hope this servers to instigate discussion!
 
Kirk Wilson, Ph.D.
CA Inc.
Research Staff Member, CA Labs
Intellectual Property and Standards
Council of Technical Excellence
W3C Advisory Committee Representative
Tele: + 1 603 823-7146
Fax:   + 1 603 823-7148
<mailto:kirk.wilson@ca.com>