RE: Semantics

isn't this "web service discovery" which is already captured by the
goal/requirements AC019 within D-AG002 and isn't the semantics portion of
your request already covered by AC009 within D-AG003?
joel
 
-----Original Message-----
From: Francis McCabe [mailto:fgm@fla.fujitsu.com]
Sent: Friday, July 12, 2002 3:23 PM
To: www-ws-arch@w3.org
Subject: Semantics



This goal addresses many problems in large scale systems as well as a number
of business needs. 


D-AG009 Semantics 

The web services architecture must support the capability for an entity to
automatically discover a service and to automatically evaluate whether it is
appropriate for its requirements. 


Rationale: 


Enabling an entity to automatically locate and choose a service does not
mean that people are not involved, nor does it mean that this must always
involve some kind of late binding. However, discovering and evaluating are
key aspects of the marketplace: without the ability to discover new
suppliers and customers it cannot be said that there is a market. 


This also touches on web services management and configuration: the same (or
similar) technology that a client entity uses to determine whether or not to
invoke a given web service can also be used to automate the management and
deployment of web services. 


Finally, looking forward to success, sorting out the semantics properly now
may help with spamming of web services in the future. Having a close
correspondence between a description of a service and its actuality will be
a powerful weapon against spam. 


Comment: 

The modern approach to semantics can be characterized as being based on
ontologies. The definition of the word Ontology is `the study of what is
real' (my definition). More typically, in the computer industry, ontologies
are short hand for dictionaries; where the definitional aspect is replaced
by a graph of concepts. 


What this means is that rather than trying to get at true meaning, which is
harder than solving the halting problem, its enough to find an appropriate
link in the graph of concepts to something that the program designer has
hard-wired into the code. Using that trail it should either be possible for
a program to `understand' a concept, or for it to have a reasonable basis to
reject it. For example, looking for a Jaguar? It might help to know that the
XJ6 is a Jaguar, and its a car; however, BigCat is a resident of San
Francisco zoo (resident might be a hard-wired concept, animal is likely
also, and SF is a city which is a place which is also hard-wired). 


The graph technology used is subject to debate, (W3C appears to prefer RDF,
others have more sophisticated choices), but that is not of the essence
really. The ontology graph is separate from any description of a service,
but an entity uses a combination of both to determine questions of the
service (is it for me). 


For web services, having the ability to read and digest a description of a
service, is a big part of the technology needed to establish a fair
automatic market. 


Critical Success Factors for this goal: 


D-AC024, D-AC025, D-AC026 


D-AC026 ensures that a web service is properly characterized so that its
semantics is clear to an automatic agent. 


D-AC026.1 The Web Services Architecture should be aligned, where appropriate
and possible with the Semantic Web. This may require some modification of
current technology choices. 


D-AC026.2 It must be possible to characterize the semantics of a web
service, including elements within a choreographed service. 

D-AC026.2.1 It must be possible to publish references to an ontology in a
web service description 

D-AC026.2.2 It must be possible to publish a description of the service
using elements of one or more ontologies. 

D-AC026.2.3 It must be possible to characterize a service using purely
publicly observable semantics. I.e., the semantic description of a web
service should not rely on private agreements or on unobservable
characteristics of services and agents.