Re: [wsrp] xsi:type usage for Extensions

[Rich Thompson <richt2@us.ibm.com>]

This conformance statement was included
for the same reason that the xsi:type attribute was defined. The key is
that there is no requirement that there exist a definition of the element
in any schema and while the type declaration is redundant in the case such
a definition exists, it allows elements with random (undefined) names to
be validated because they reference a defined type.

Rich

Subbu Allamaraju <subbu@bea.com>

09/25/2006 01:40 PM

To	OASIS WSRP TC <wsrp@lists.oasis-open.org>
cc
Subject	[wsrp] xsi:type usage for Extensions

WRRP v1/v2 has a conformance statement that requires
all extensions to 
explicitly declare their type using the xsi:type attribute.

Can someone clarify why this is required? In general, the type 
information can be derived by using the element name and namespace 
information, as in

<v1:extensions>
   <ext:foo xmlns:ext="foo:ext:namespace">
     <ext:bar/>
   </ext:foo>
</v1:extension>

When added, an xsi:type would essentially carry the same information, e.g.

<v1:extensions>
   <ext:foo xmlns:ext="foo:ext:namespace" xsi:type="ext:FooType">
     <ext:bar/>
   </ext:foo>
</v1:extension>

The only time xsi:type is required when the XML is using polymorphism as
in

<v1:extensions>
   <ext:foo xmlns:ext="foo:ext:namespace" xsi:type="ext:FooSubType">
     <ext:bar/>
   </ext:foo>
</v1:extension>

where FooSubType is extending FooType. In other cases, xsi:type is not
required for processing the XML.

Moreover, the xsi:type attribute cannot be set when the type is 
anonymous (i.e. when the element declares the type inline).

IMO, this conformance statement is not required. Any comments?

Regards,
Subbu