Currently on a SPARQL LOAD query Stardog sends a request to the server serving the graph to be loaded with
Accept: text/html, image/gif, image/jpeg, *; q=.2, */*; q=.2
This doesn’t seem very appropriate as an RDF representation should be preferred.
But what’s worse is that Stardog ignores the Content-Type header in the response but instead guesses the response type from how the IRI of the loaded resource ends. This is wrong as URIs are opaque.