From: - Fabien Gandon - ([email protected])
Date: 05/30/03
Remarks on the white paper "DAML-S: Semantic Markup for Web Services"
-- Fabien L. Gandon, Norman M. Sadeh
Mobile Commerce Laboratory
Carnegie Mellon University
The following are a few informal comments on the new version of the
white paper "DAML-S: Semantic Markup for Web Services".
1, As the article tries to systematically make links with other domains
and terminologies, one could insert a sentence stating that the
matchmaking capabilities and the knowledge bases containing the
profiles/service advertisement are sometime respectively called 'yellow
pages services' and 'yellow pages'.
2, There is not enough emphasis on the necessity and use of an ontology
of the types of services (e.g. booking, renting, ordering, etc.) and the
domain of the services (e.g., cinema tickets, car, pizza) with its
specific concepts and properties (e.g. half price for students, 2 years
warranty, deliver at home) that specialize the upper class Service and
allow the matchmaking.
3, Why is it the case that the Service instance is not simply annotated
with domain-dependent ontologies? Why a taxonomy slot and not simply a
use of namespaces? Why a categoryName slot instead of using a subtype of
service in an ontology and multiple instantiation? This really seems
counter-intuitive since all these mechanisms already exist in RDF/S.
While we understand the rationale of section 4, we don’t see the point
of sub-sections 4.2.6, 4.2.8, 4.2.9 ; these are typically
domain-dependent and should be addressed using domain ontologies
(identified by their namespaces) and multiple instantiation. Possible
links between these ontologies and/or between these ontologies and the
Service ontology would be done using subsumption, class equivalence,
etc. In our opinion these domain-dependent aspects should be handled
independently, thus only part of the profile should be specified and the
rest should be left to extensibility and semantic search capabilities.
4, There is no reference to the RDF/S version of WSDL not even as
"related work" while it seems extremely relevant be it only for the
grounding. Advances on WSDL 2.0 should be acknowledged too.
5, Section 4.2.4: Failure during the activation of a Web service could
have effects too e.g. report misuse of credit card and ban access if
wrong PIN. It seems that the effects should really be linked to some
stages of the process. Are we missing something here?
6, The white paper should also make some links with existing ontologies
(time, space, persons, services, etc.) and acknowledge that some content
/ ontologies already exist.
7, In section 5.1, first bullet, when declaring the property 'input',
there is no use to overwrite the range since it is the same as the one
inherited from 'parameter'.
8, Concerning the SimpleProcess, it seems to us that a better labeling
would be "AbstractProcess" since this really is what its definition
means: we understand it as an abstract building block of the process
model (i.e. it cannot be invoked or grounded) and it has nothing to do
with being simple vs. complex or composite.
9, Section 7.1: what if a resource can be used exactly X times (e.g. a
10-trip bus pass, right to access a location tracking device 2 times per
day), should it be modeled as the capacity of a non 'replenishable'
resource ? Also these issues are closely related to security and access
rights this would eventually need further details.
Some typos / formatting details:
Section 3: "The class service provides aN organizational point of
reference (...)"
Section 6.1 : "XSD" acronym should be introduced when talking about XML
Schema data type, and may be it would be good to insert a line insisting
on what semantic web frameworks buy us compared to XSD
Section 7.2 : the acronym "DAML-L" is used without being defined.
Best regards,
Fabien and Norman.
--
____________
|__ _ |_ http://www-2.cs.cmu.edu/People/fgandon/
| (_||_) Carnegie Mellon University - School of Computer Science
This archive was generated by hypermail 2.1.4 : 05/30/03 EST