303 redirect to a fragment – what should a linked data client do?

Hi all,

  in our setup we are still somehow fighting with ill-conceived legacy URIs
from the pre-LOD age.  We heavily make use of hash URIs there, so it could
happen that a client, requesting http://example.org/foo#bar (thus actually
requesting http://example.org/foo) gets redirected to
http://example.org/baz#grr (note that I don't mean
http://example.org/baz%23grr here, but really the un-escaped hash).  I
observed that when serving such a result as XHTML, the browser (at least
Firefox) scrolls to the #grr fragment of the resulting page.

But what should an RDF-aware client do?  I guess it should still look out for
triples with the originally requested subject http://example.org/foo#bar, e.g. 
<rdf:Description rdf:about="http://example.org/foo#bar">, or (assuming
xml:base="http://example.org/foo") for <rdf:Description rdf:ID="bar">.  Is my
assumption right?

Thanks in advance for any help,

Christoph

-- 
Christoph Lange, Jacobs Univ. Bremen, http://kwarc.info/clange, Skype duke4701