To answer the question about "datasets about datasets", William Waites has done some excellent work to produce RDF from CKAN at http://semantic.ckan.net/, which we are building on for the Dataset Inventory for the LATC project: http://dsi.lod-cloud.net/ (work in progress) which has a SPARQL endpoint at http://dsi.lod-cloud.net/sparql
The clever folks behind rkbexplorer have also created a repository of VoID data at http://void.rkbexplorer.com/ With regards to dataset URIs - with some datasets, the dataset publisher has chosen to use the base URI (eg: http://data.ordnancesurvey.co.uk ) as the URI of the dataset. In other cases the dataset publisher has chosen a different pattern. There isn't a right or wrong - it's up to the publisher to decide what their URIs mean. In some cases, for instance, it is more convenient for the publisher to use a hash URI in a hand-curated file, and that's fine. You are right though, that currently the patterns for discovering dataset descriptions (such as .well-known/void or "<doc> void:inDataset <dataset>" triples ) are not as widely implemented as they could be. HTH Keith
