Atom-based service registries

There is an interesting article in the September/October issue of IEEE Internet Computing (ISSN 1089-7801), pp 68-71: Active Web Service Registries by Martin Trieber and Schahram Dustdar of the Vienna University of Technology. They are investigating building service registries that disseminate their records using RSS (currently but intending to upgrade to Atom). Each service provider sets up their own RSS feed describing their services at a known location, ie their own local service registry, aka active web service registry. Then there are wider service registries that are built as aggregators of the distributed local registries by whatever selection. One could envisage building higher-level, even global registries by aggregating the distributed aggregators.

They are discussing registries of Web Services (ie. SOAP) only - partly it is seen as a solution for the demise of public UDDI registries. So they have a narrower remit than IESR's of supporting many service protocols and also descibing collections. Also I get the impression that they see users of this information being people, application developers, who look up details of Web Services to plug them in to applications. There is no mention of machine-to-machine use, apart from the aggregation.

But using Atom, and aggregating local services registries, seems a neat idea. It would certainly lend itself to investigation if new instantiations of service registries were to be designed. It may provide a paradigm for building global registries.

One particular point they make struck me as being interesting to consider. They recommend including examples of use with each registered Web Service, for example a concrete SOAP message that will invoke the Web Service, to help application developers using the information, and so to better advertise and assist take-up of the Web Service.

Another interesting article in the same journal issue is pp 17-25: Requirements and Services for Metadata Management by Paolo Missier, Pinar Alper, Oscar Corcho, Ian Dunlop, and Carole Goble of The University of Manchester. This talks about capturing annotations about data resources, here in a bioinformatics context. The mass of such annotations is in fact a valuable resource in its own right, which it would be useful to capture and associate with the data resource. Their solution is very semantic web / RDF based. But the ideas are interesting, especially in this age of social bookmarking. But maybe capturing annotations about the collections within a registry is still in the 'nice to have, but in the future when we have resource to implement it' category.