I followed the W3C RDF Validation Workshop  over the last two days.
The web page has both written papers and slides from each presentation.
The short summary is that a number of users of RDF have found a need to
do traditional style validation (required, one or more, must be
numeric/from a list, etc.) on their RDF metadata. There is currently no
RDF-based standard for defining validation rules, so each of these is an
ad hoc solution which cannot be easily exchanged. 
The actual technology of validation in all cases is SPARQL. Whether or
not this really scales is one of the questions, but it seems pretty
clear that SPARQL will continue to be the solution for the near future.
I will try to write up a blog post that will give some more information.
 nota bene: Although OWL appears to provide validation rules, the OWL
rules only support inferencing. OWL cannot be used to constrain your
data to valid values.
[log in to unmask] http://kcoyle.net