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

Web services composition

From: Charlie Abela <abcharl@keyworld.net>
Date: Fri, 25 Jul 2003 10:07:03 +0200
Message-ID: <OBEOLJNEPKKHAJJIGKDKEEIHCFAA.abcharl@keyworld.net>
To: "Web Services List" <www-ws@w3.org>
Cc: "Sheila McIlraith" <sam@ksl.stanford.edu>


I came across the paper: "Planning with Complex actions" [1] which describes
the idea of treating web services as complex actions, representing them as
primitive operators and then be able to plan with them to generate
compositionThe author mentions two drawbacks for using automatic plan generation
techniques directly, one of them regards:
i. the fact that WSs can be composite and their primitive components are not
available to the user. He also states that "to achieve a goal, we cannot use
the primitive services, it is necessary to use composite services as
primitive operator. We must be able to use them as if they were atomic parts
of the plan. But we only know their definition in terms of primitive

This isn't clear to me: what's wrong with using a composite service as if it
were a black box? A composed service definition may not necessarily present
the service in terms of its primitives. What really matters is that the
preconditions, inputs, outputs and effects (IOPEs) are clearly defined. So I
might be using a composite service that someone has composed for a
particular use as if it were a primitive one. I think that this issue is
relaxed or constrained by the WS provider and if the IOPEs are defined then
this service could be incorporated within a composition by making use of
planning techniques and DAML-S, as defined in [2].

[1] http://www.ida.liu.se/~patdo/nmr2002/Final_Papers/mcIlraith.pdf
[2] http://www.cs.umbc.edu/~finin//papers/aamas03b.pdf

Am I missing something? Can someone comment these issues?


Charlie Abela
Research Student,
Dept. of Computer Science and AI
University of Malta,
MSD06. Malta
Web: http://alphatech.mainpage.net
Email: abcharl@keyworld.net
Received on Friday, 25 July 2003 04:02:42 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:37:09 UTC