Re: LOD datasets "schema-triples"

>
>   The schemas are de-coupled from the datasets: a standard linked data
> schema can be used by any number of datasets, and a given dataset can use a
> number of schemas.  You can see what properties and classes are in use in
> the dataset, and the prefixes on those should be mapped to URLs at the top
> of the file.
>

I think this is one of the places where current RDF practice, and to an
extent even the spec, seems not fully thought through from the
data-audience's perspective. Most major RDF datasets declare a boatload of
prefixes, but only use small subsets of each of the reference schemas. As a
data consumer, the first thing I want and need to know about a dataset is
how it's structured. With RDF datasets I end up having to figure this out
myself by a bunch of fiddly exploratory queries. It would be much better if
there was an actual schema described.

This is bad enough for people, but it's *really* bad if you're trying to
write a generic data-browser. With no schema you have no easy way of knowing
what nav/display/facet/filter/analysis options to show people at any given
point. And it's even worse if you're trying to do (or write a tool to
support) collaborative editing of a dataset, or even single-author editing
over time, as there's no easy way to validate that things are being done
consistently. Knowing that the "foaf" prefix was declared doesn't tell me,
for example, that in this particular dataset foaf:name is the only foaf
property that's supposed to be used...