Having used OWL-S for some time, both in the form of the 'Virtual Machine' [12] and our own experimental implementation [7], the Cashew project has drawn a number of conclusions about its existing process model which we assert as follow: 1. OWL-S is not service composition; 2. OWL-S defines orchestration but not choreography; 3. OWL-S is incompatible with choreography; 4. OWL-S is insufficiently composable; 5. OWL-S needs compositional semantics. This position paper will details, and attempt to justify, these assertions and sketch the Cashew approach to their solution.
The first major result of the Cashew project has been to define a
compositional semantics for OWL-S2 [
The second intermediate in our semantic translation is Cashew-Nuts, a true
process language. This is a process calculus defined by extension of CCS [
Having discussed the last two assertions, and our existing work, we now move
on to make our position statement as embodied in the first three assertions. The
WSMO group have proposed that the behavioural models for each semantic web
service should consist of both an ‘orchestration’ and a ‘choreography’, that is to
say an internal and external view of its composition [
What seems not to have been documented is how a lack of fit with the
general view of services, asserted as 1, affects the ability to adopt choreography
into the OWL-S world. In the ‘Web Services Architecture’, services in general
offer multiple operations [
The problem here is that (one of) the aim(s) of (WSMO’s) choreography is to establish a ‘protocol’ by which a client’s session with a service can invoke its operations. In the commonly used bookshop example, it might be necessary to log in before a purchase can be made. Even though an OWL-S orchestration could involve both the ‘log in’ and ‘purchase’ operation, it would not be made clear that these belonged to the same session (one orchestration could operate across more than one session, for instance in mediating between two accounts). 2 Note that it makes no sense to ask whether OWL-S is compositional or not — this is a semantic property that applies to a given semantics for OWL-S. 3 In this, the use of the term choreography, which is not novel or original to WSMO, is slightly different to other definitions, such as W3C’s, where a multi-party conversation, to achieve some particular task, is implied.
IRS-III [
On the other hand, client choreographies, formalised in [
In this way the coordination between low-level operations, and the dependencies between them (as expressed in the deployed service’s choreography), is isolated from the workflow. IRS-III can be viewed as a broker allowing an abstract orchestration of goals to be built, to achieve some high-level task. At run-time IRS locates client choreographies by which the goals can be realised.
In the current implementation, client choreographies are directly expressed
in abstract state machines [
Most importantly, we shall show how our compositional OWL-S semantics can be built upon to give an operational semantics to all these models, and how these semantics can be grounded in abstract state machines for compatibility with WSMO. In this way we shall create a full implementation for orchestration within the IRS and, furthermore, have a formal model on which conformance properties can be checked.
Log−in
Input username
password Ready
/ready Waiting
execute
Executing [!validated]
[validated] Output
Output /nack
/ack
Input username
password Ready
/ready Waiting
execute
Executing [!validated] [validated]
Amazon
Input
Search author
title title author
Ready Log−out
Input
/ready