W3C home > Mailing lists > Public > public-sws-ig@w3.org > November 2005

RE: Semantics in WSDL-S ?!

From: Shi, Xuan <xshi@GEO.WVU.edu>
Date: Mon, 21 Nov 2005 10:52:45 -0500
Message-ID: <D81F456794C18B4DA3E2ABC47DBBEEF2094E3C@onyx.geo.wvu.edu>
To: "'Battle, Steven Andrew '" <steve.battle@hp.com>, "Shi, Xuan" <xshi@GEO.WVU.edu>
Cc: "'public-sws-ig@w3.org '" <public-sws-ig@w3.org>

My real question is described in this section, as for WSDL-S:

"Semantics in this context refers to the meaning of objects or information."
Semantic annotations in WSDL-S "define the meaning of elements in WSDL
document by referring to a part of a semantic model". 

This means "Semantics" in WSDL-S is the meaning of objects used in WSDL 
elements (generated by OOP) rather than the meaning of service (may have 
nothing with OOP or WSDL). Thus WSDL-S handles the content within WSDL, but
unfortunately, the meaning of service is beyond the WSDL document itself. 

Do you think the names of objects, functions, etc. inside WSDL is the
meaning of themselves? When you deals with those simple business transaction
models, like credit card transactions, buy tickets, buy whatever, the name
of your object is self-describing, thus is meaningful. However, when you get
into more complicated models like the geospatial functions and services, the
same interface DOES have different meaning.

For example, WSDL interface:

Function intersect (String Polygon1, String Polyon2): String Polygon3.

The meaning of this function depends on the purpose of your action. That's
to say, you can use this same interface to perform 2 functions: 1. process
geospatial dataset to generate a new feature dataset, or, 2. proferm spatial
query to select a subset feature from one of the two input polygons. In this
case, you need to specify the order of two input string data types.
Otherwise, you get wrong result.

Again, the meaning and specification are ALL defined INSIDE the string data
type "Polygon1" or "Polygon2" while you can name such data object in any way
you want, then how can you add semantic annotations OUTSIDE such objects?
When I communicate with people in this group, it seems everyone understands
that the name of the objects/functions inside WSDL is not the meaning of
service. Then how can we add semantic annotations onto such objects inside
WSDL, considering most of the content of WSDL document are meaningless and
useless to requesters to retrieve the answer they want?
 

-----Original Message-----
From: Battle, Steven Andrew
To: Shi, Xuan
Cc: public-sws-ig@w3.org
Sent: 11/21/05 6:28 AM
Subject: RE: Semantics in WSDL-S ?!


> -----Original Message-----
> From: public-sws-ig-request@w3.org 
> [mailto:public-sws-ig-request@w3.org] On Behalf Of Shi, Xuan
> Sent: 21 November 2005 04:54
> To: 'jeff@inf.ed.ac.uk '; 'David Martin '
> Cc: 'Amit Sheth @ LSDIS '; 'public-sws-ig@w3.org '
> Subject: Semantics in WSDL-S ?!
> Importance: High
> 
> BTW, why is it said that "the current WSDL standard operates 
> at the syntactic level"?  What is any more semantic about the 
> things that are labelled "semantic"?
>
> 
> WSDL operates at the syntactic level because it only contains 
> the names of the object data types and functions, etc. that 
> are used developing the service, as well as the object 
> hierarchy, relations, etc. which are ALL of the outcomes 
> generated in Object Oriented Programming (OOP) approach. This 
> means, WSDL contains the details of coding process generated 
> by OOP, however, the meaning (semantics) of the service is 
> not described and defined within WSDL. 
> 
...
> 
> the meaning (semantics) of this Web service is:
> 
> if the requester can invoke ESRI's authentication Web service 
> to obtain a validated token string, then requester can use 
> this token string with the input "address" object data type 
> as well as the specified data source to invoke Address Finder 
> Web Service, which will then return the location(s) 
> (longitude and latitude) that can match the input address or 
> will then return an error message to the requester. 
>

Clearly, there's a lot you can't say in WSDL. What it does well is to
describe the service _signature_ in terms of the operations available
through one or more service interfaces. Like you say it doesn't capture
any of the 'meaning' as you describe above, which includes composition,
inputs and outputs (and their types). But this is a different question
as to whether it is 'semantic' or not. WSDL 2.0 now has an RDF mapping
that furnishes it with a model theory. You can think of a WSDL
description as a set of logical assertions (about a service interface).
This in itself is an _amazing_ thing and is a qualitative step-up from
the XML world view, but is about as 'semantic' as its ever going to get.


The real action is how you can extend this model to incorporate the more
dynamic aspects of the service that you're interested in. This is as
much about how to establish shared ontologies (of address, location,
errors) as figuring out how to add the right markup to the WSDL.

Steve.
Received on Monday, 21 November 2005 15:52:19 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Sunday, 16 March 2008 00:11:02 GMT