Re: Request from the ATF team to the authors of the WSDL Part II Spec

Jean-Jacques:
Thanks for your thoughts.  Some discussion:

>I note there is no support for attributes
>as such in SOAP (or HTTP). I assume attributes would be returned as a
>(fragment) XML document, possibly in the body of a SOAP (or HTTP)
>message.
Indeed.  Per the document we posted, the notion is that the get/set
functionality is indeed represented
by the values of the attributes being represented as an XML document in the
body of the SOAP or HTTP
message

>can attributes (+values) be seen as
>an XML document, just like an SNMP MIB can (after conversion)
Indeed, they can be viewed as such.  In fact the attributes ARE
associations between XML elements and
the interface(s).  We could have modelled a simple association between an
interface and a single root element.
However, this would be difficult to reason about in the light of interface
extends hierarchies.  Furthermore, many
applications don't view a single document as being sufficient for all
attributes.  Some attributes can be set
by a "bulk load" of a single document, setting the values for a subset of
the listed attributes.  Some subset of the
attributes will provide a mechanism for value change notification. Some
attributes can be queried in combination,
in a style such as XPath and XQuery.

>If so, shouldn't we
>  leverage technologies such as XQuery and XPath, rather than inventing
>our own extension?
Indeed, it is quite possible for OGSI or WSDM in OASIS to define interfaces
that suggest operations that support
XPAth or XQuery support on top of the attributes associated with a service
(ie those attributes associated with the
interface(s) implemented by that service). What the ATF is attempting to do
is to formalize the association between
interfaces and XML elements that represent the attributes, and formalize a
very simple access mechanism to get/set
values of single attributes. More sophisticated attribute value access
mechanisms can then be built using domain-specific
interfaces defined by GGF, OASIS/WSDM, etc.

>Now down to your specific question: how does this fit into a WSDL
>binding? I am under the impression that this is not covered (yet) by
>your document.
Hmmm... the section "accessing attributes" contains our initial thoughts
on the wire format.  However, it is our intuition that there would not
need to be any changes to the component model described in the bindings.
We would need to state in part III that attributes are to be interpreted
in the way suggested by "accessing attributes", and I don't think we need
to do much more. I don't think we need to describe how the wire syntax is
described in WSDL.  If the service uses a SOAP binding with an interface
decorated with attribute declarations, the wire syntax is the way the
client
can expect to send messages to get/set values of the attributes.

>(BTW, if we
>decide to publish your work in its current form, I think we would need
>to prepare a companion SOAP spec, that describes your SOAP encoding for
>attributes.)
If we do agree on attributes in the WG, then perhaps this should not be
a companion spec, but rather a section within Part III?

>To cover the SOAP/HTTP serialization you are suggesting, wouldn't we
>have to introduce a new mode, in addition to the existing literal and
>rpc modes?
I don't think it is a new mode.  Rather, what we are saying is: since
attributes are "siblings" of operations in the interface declaration, there
is a parallel first class interpretation in the binding for attributes.
The wire syntax suggested is what the client can expect.

>Also, I am wondering why you are not envisaging setting/getting several
>attributes at once.
Indeed we are.  However, such sophisticated functionality was considered
beyond the scope of what is appropriate for the WSDL language itself.  I
expect
that other groups such as OGSI in GGF and WSDM in OASIS will define
(hopefully one)
interface that would provide more sophisticated attribute value access such
as you suggested.

>I realize I have not followed the ATF's work too closely; you may have
>answered these sorts of questions already; as said earlier, I apologize
>in advance if I am completely off topic.
Good questions.  I appreciate the dialog.

sgg

++++++++
Steve Graham
sggraham@us.ibm.com
(919)254-0615 (T/L 444)
STSM, On Demand Architecture
++++++++



                                                                                                                                             
                      "Jean-Jacques Moreau"                                                                                                  
                      <jean-jacques.moreau@cr        To:       Steve Graham/Raleigh/IBM@IBMUS                                                
                      f.canon.fr>                    cc:       jeffsch@windows.microsoft.com, public-ws-desc-state@w3.org                    
                                                     Subject:  Re: Request from the ATF team to the authors of the WSDL Part II Spec         
                      09/19/2003 09:58 AM                                                                                                    
                                                                                                                                             
                                                                                                                                             




Steve et al,

I said I would try to provide comments on your document[1]. I have not
followed the ATF's work too closely, and apologize in advance if I am
completely off the track.

First, a high level comment. I note there is no support for attributes
as such in SOAP (or HTTP). I assume attributes would be returned as a
(fragment) XML document, possibly in the body of a SOAP (or HTTP)
message. This raises the question: can attributes (+values) be seen as
an XML document, just like an SNMP MIB can (after conversion)? If it
can, doesn't the problem become: a) being able to describe an XML
document; and b) being able to query that document? If so, shouldn't we
  leverage technologies such as XQuery and XPath, rather than inventing
our own extension?

Now down to your specific question: how does this fit into a WSDL
binding? I am under the impression that this is not covered (yet) by
your document. Section "Accessing Attributes" describes the wire syntax
for SOAP and HTTP "attribute getting and setting" messages, but it does
not actually indicate how this wire syntax is
modelled/described/represented in WSDL. I think it should. (BTW, if we
decide to publish your work in its current form, I think we would need
to prepare a companion SOAP spec, that describes your SOAP encoding for
attributes.)

To cover the SOAP/HTTP serialization you are suggesting, wouldn't we
have to introduce a new mode, in addition to the existing literal and
rpc modes?

Also, I am wondering why you are not envisaging setting/getting several
attributes at once.

I realize I have not followed the ATF's work too closely; you may have
answered these sorts of questions already; as said earlier, I apologize
in advance if I am completely off topic.

Jean-Jacques.

[1]
<http://lists.w3.org/Archives/Public/public-ws-desc-state/2003Sep/0000.html
>

Steve Graham wrote:

> Jean-Jacques/Jeffrey:
> In today's teleconference of the Attributes Task Force, we agreed to
> request preliminary review of the updated ATF report [.1] from the
authors
> of the WSDL Part II spec. Since the major changes in this draft relates
to
> the way we propose that interface/attribute elements are interpreted in
the
> binding, your thoughts would be very valuable in our work.
>
> If possible, we would also request that you attend our ATF telecon on
> Monday, Sept 15 (12-1 EST) to discuss your thoughts. [.2]. Please let us
> know if you are willing to take the time to review this document and/or
> attend our next telecon.
>
> Thanks in advance.
> sgg
>
> [.1]http://lists.w3.org/Archives/Member/w3c-ws-desc/2003Sep/0011.html
>
[.2]http://lists.w3.org/Archives/Public/public-ws-desc-state/2003Sep/0000.html

>
> ++++++++
> Steve Graham
> sggraham@us.ibm.com
> (919)254-0615 (T/L 444)
> STSM, On Demand Architecture
> ++++++++
>

Received on Friday, 19 September 2003 15:22:40 UTC