W3C home > Mailing lists > Public > www-ws-desc@w3.org > January 2003

Re: R120 URI-references, input for editors

From: <ryman@ca.ibm.com>
Date: Sun, 19 Jan 2003 18:35:11 -0500
To: www-ws-desc@w3.org
Message-ID: <OF137B05F4.B2D5C2E3-ON85256CB3.00808490@torolab.ibm.com>


The requirement to refer to conceptual elements comes from Semantic Web,
not WSDL. Within WSDL, most things are refered to by QName or NCName. A
<message> doesn't need to refer to its parts, but for Semantic Web, some
application might want to identify a message part.

Similarly, some application might want to refer to the type of an element
even though it's anonymous. I think it's reasonable to say that all XSD
have types, i.e. there's a mapping from XSD elements to XSD types.

R120 was a based on a general principle that we can't identify all future
application requirements, so we need to have a URI mechanism for
identifying the conceptual components of a model. Of course, it is up to us
to decide what is a conceptual component. I think any component of our
abstract model qualifies as an important conceptual component. After the
abstract model is finished, we'll have to update the URI-reference proposal
to match it.

I'd like to ask David Booth to further clarify the URI-reference
requirement since he represents W3C and therefore should be able to speak
for Semantic Web.

Arthur Ryman

                      Jacek Kopecky                                                                                                                 
                      <jacek@systinet.c        To:       Arthur Ryman/Toronto/IBM@IBMCA                                                             
                      om>                      cc:       WS Description WG <www-ws-desc@w3.org>                                                     
                      Sent by:                 Subject:  Re: R120 URI-references, input for editors                                                 
                      01/17/2003 06:20                                                                                                              


concerning non-top-level conceptual elements (those *not* named with an
XName - expanded name, what we used to understand under the term QName),
we in WSDL use an indirect approach to referencing them - message parts
are referenced by referencing the message with its XName and the part
with its name. Why do we think a single URI is going to be necessary, if
the main user of message parts - WSDL itself - doesn't need it?

I know we voted on the simpler approach at the f2f, but it seems we're
in fact going to mimic XML Schema's approach (the TAG was waiting to see
what XML Schema decides on, too).

In a different mail, you mentioned the need to be able to reference all
conceptual parts of our languages (WSDL, XML Schema).

Until I'm shown why it's needed, I dismiss the requirement for direct
URI referencing of non-top-level named components, like WSDL message

Finally, I completely disagree that e.g. XML Schema anonymous types need
to be referenced (other than by XPointer/XPath in a schema document)
because if they really need to be referenced, they should be named. XML
Schema itself cannot reference an anonymous type, can it?

To conclude: if something needs to be referenced, it should be named. If
something only makes sense in a context, it should be referenced
together with the context. XNames and namespaces should follow RDF/XML's
simplistic approach.

I hope to discuss this all with the XML Schema folks in context of
Component Designators or with TAG folks in context of TAG issues 6 and
maybe 28.


On Thu, 2003-01-16 at 22:56, ryman@ca.ibm.com wrote:
> Jacek,
> Concerning <part>, it is true that are not named with a QName, but they
> conceptual elements of WSDL. R120 says that all conceptual elements
> have a URI-reference.
> I looked at using raw XPointer/XPath and produced examples of the syntax.
> They do solve the problem, but they are very complex. I presented the
> alternatives at the last f2f. The WG voted on the simpler syntax even
> though it meant inventing something. The simpler syntax is compliant with
> the XPointer Framework.
> Consider the following fragment example which appears in

> The portType TicketAgent has an operation listFlights that contains an
> input listFlightsRequest
> The proposed XPointer Framework syntax is:
> #input(TicketAgent/listFlights/listFlightsRequest)
> The full XPath syntax is:
> #xmlns(w=http://schemas.xmlsoap.org/wsdl/) xpointer(//w:portType[@name
> ="TicketAgent"]/w:operation[@name="listFlights"]/w:input[@name
> ="listFlightsRequest"])
> Concerning your discussion of XNames, I don't understand the point you
> trying to make. Can you explain what an XName is?
> Arthur Ryman,
Received on Sunday, 19 January 2003 18:35:17 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 23:06:27 UTC