RE: XForms WD 20020821 3.3.1 Referencing Schemas - Catch 22?

Andrew,

Thank you for your comments.  I am answering for myself and not the WG at
the moment, and hence have dropped www-forms-editor from the recipients
list.

May I first draw your attention to an earlier message I sent on a related
topic:
   http://lists.w3.org/Archives/Public/www-forms/2002Aug/0102.html
A summary of that note: that there are two ways of adding a Schema to a
model: inline, by the xsd:schema child element of model, and by URI
reference on the schema attribute on model.  

The locating-schemas issue in 3.3.1 was intended to solicit feedback on the
"list of XML Schema documents" behavior of the model element schema
attribute. 

On the fragment issue, I personally do not think that the use cases for
having an inline Schema in the containing document but outside the XForms
model are strong, since a Schema can be included within the XForms model
xds:schema element already, but others may disagree.  I would propose simply
removing the fragment language from 3.3.1 and replacing it with a
defininition that schema list URIs each be processed according to "3.2.2
Linking Attributes" as described for the "src" attribute, include following
the XML Base recommendation from 3.2.2.

Furthermore, note that there is no way to assign Schemas on a per-instance
basis inside a model.  All Schemas are used for processing all instances in
the model.  Given this point, I believe that if we remove the URI reference
to a fragment from 3.3.1, yhere should then be no need to refer to
particular <xsd:schema> declarations by id or fragment URI and so the issues
you raise will be solved.

Leigh.


-----Original Message-----
From: AndrewWatt2001@aol.com [mailto:AndrewWatt2001@aol.com] 
Sent: Thursday, August 29, 2002 9:25 AM
To: www-forms@w3.org; www-forms-editor@w3.org; xforms@yahoogroups.com
Subject: XForms WD 20020821 3.3.1 Referencing Schemas - Catch 22?



I note that in 3.3.1 the WG seeks feedback on implementation of the schema 
attribute. I would suggest that there is another step necessary before 
worthwhile feedback can be provided - the WG needs to more clearly define 
what the schema attribute is intended to do.

3.3.1 mentions a URI fragment such as "#mySchema" without defining the 
semantics of the syntax. Is it a reference to an HTML/XHTML anchor? 
Presumably not - this is supposedly a cross-platform XML technology.

So is it a "bare names" XPointer? We are not told.

But a bare names XPointer is a shorthand for access to an XML element's id 
attribute and here we run into potential trouble.

Chapter 3.2.1 seems to imply that it is the host language, not XForms, which

adds an id attribute to the XForms elements.

So, unless I am misunderstanding all this (which is quite possible), the WD 
seems to expect a bare names XPointer to reference an id attribute which is 
(yet to be) provided by the host language.

Since the host language is not obliged to add an id attribute to the 
xforms:schema element which corresponds to the schema attribute of the 
xforms:model element there seems to be a Catch 22. The schema attribute will

likely be referencing a non-matching id attribute on an xforms:schema 
element.

If I have this all upside down the explanation in the WD needs to be 
improved. If I have it the right way up I would suggest that some design 
points need to be re-thought.

Andrew Watt

Received on Thursday, 29 August 2002 14:10:53 UTC