CommentResponse:DB-14a

David,

Your request is that the str function be changed so that, when applied to a blank node, instead of an error, a string is returned that is store unique.

A store may not have such a string - it supposes certain design features such as persistence. By requiring one for SPARQL use, this would be a implication on RDF abstract syntax. Therefore the working group feel that the current design is the best balance.

For example, a lightweight SPARQL implementation may store all its data in an RDF serialization and read the data from time-to-time, maybe keeping an in-memory copy, but re-reading from disk when the process starts up or only regarding memory as a cache. RDF serializations do not have a persistent, stable identification of a blank node across different parsing of the file.

Inferred results may also not provide a notion of a stable identification of a blank node.

As previously noted, because STR of a blank node is an error, store implementations may, and some do, provide the facility as an extension.

The RDF Working Group is pursuing advice on skolemization of blank nodes and that would provide a RDF mechanism for what you aks. It would not be a required feature of toolkit supporting RDF.

We would be grateful if you would acknowledge that your comment has been answered by sending a reply to this mailing list.

Andy (on behalf of the SPARQL Working Group)