SAWSDL for SWS and Semantic Modeling in the Cloud
I'm writing to complain about two recent articles in IEEE Internet Computing
The authors might be misleading in their research findings because they might not have a correct idea about service semantics. Semantic annotation has been proposed for the Semantic Web, Semantic Web services (SWSs) such as SAWSDL, and REST services such as SA-REST. Obviously, no miracle has happened in the past decade, even after a variety of markups were placed on the Web, because the annotation approach isn't grounded in scientific theory and lacks scientific methodology. Because only simple use cases were deployed, this approach has little science behind it except stories about buying something via credit-card transactions.
SWSs have comprehensive goals to enable automatic and dynamic service discovery, matchmaking, composition, and invocation. Apparently, SAWSDL is useless for dynamic invocation. 3
Web service technology's main contribution stems from its interoperability capabilities. Because service invocation is determined by the syntactic definition, not semantics, of services, adding semantic annotation to syntactic terms doesn't help dynamic invocation.
A semantic definition's purpose is for service discovery, not invocation. However, service semantics are neutral and independent from syntactic API definitions. Moreover, not every concept in semantic definitions has a corresponding syntactic counterpart. This, again, renders adding semantic annotation to syntactic terms useless for service discovery because those syntactic terms might not have any relation to the service's semantics. 4
With regard to service-related research, the concept of service isn't scientifically or technologically consistent. Semantic modeling such as SAWSDL is questionable in cloud computing because services in SAWSDL are APIs and services in SA-REST are URLs. However, the concept of service in cloud computing has a much broader semantic scope. If scientific research doesn't start from a correct concept, how can there be a correct result? Thus, the annotation approach is pointless for marking up the infrastructure, platform, and other components — which are all services — in the cloud because even the definition of varied services is ambiguous.
Amit Sheth responds:
Xuan Shi seeks to inquire whether the current approach for semantic annotation of Web services and the respective W3C recommendation — SAWSDL — are well grounded in science, have good use cases, or have been widely adopted. Such an inquiry, suggestions for improvements, or even a criticism would generally be welcome, and have, in fact, been the subject of several commentaries and discussions. 1,2 Although much can be debated about SWS, Shi's commentary isn't constructive and repeats what he said during W3C's SAWSDL Working Group discussions and in his own previous work (for example, see "Semantic Web Services: An Unfulfilled Promise"; www.coa.gatech.edu/cgis/bios/shi_bio.htm).
Here's a brief response to some of his points:
• Theoretical grounding. An annotation itself is simply a way to provide metadata (for example, consider poshformats or microformats) or to link an object with a conceptual description. Carlos Pedrini and his colleagues describe various ways additional layers have been built, including OWL-S and WSMO's grounding to SAWSDL, which use richer, theoretically grounded descriptions. 3
• Adoption and use cases. A Google search for either "Sawsdl" or ".sawsdl" will give more than 15,000 documents and provide links to tools and enhancements (for example, for enhanced annotation or discovery of SAWSDL documents). Such a search will also give many use cases — including academic and commercial, simple and complex — in several domains such as life sciences (for example, caBIG project) and financial services (see www.securitiestechnologymonitor.com/issues/19_50/22047-1.html). SAWSDL isn't a solution for annotating all types of services or Web resources. Correspondingly, newer efforts, including two recent W3C submissions, SA-REST ( www.w3.org/submission/sa-rest) and WSMO-Lite ( www.w3.org/Submission/2010/SUBM-WSMO-Lite-20100823), either build on or provide complementary solutions to SAWSDL and related annotation work.
• Dynamic invocation. The basis of Shi's comment on this topic is unclear to me. A Google search on "dynamic service search, sawsdl" gives several examples, such as support for dynamic service binding, invocation, and other issues. 4
Having worked with several clouds, we do know what cloud interfaces look like. However, my recent column articles haven't discussed SAWSDL and the cloud, so it's premature to make connections between the two.
Selected CS articles and columns are also available for free at http://ComputingNow.computer.org.