This application will convert arbitrary Google DataWiki data-sets to RDF/XML by using XSL Transformations. You can then convert the RDF/XML to your favorite serialization.
If you have a particular DataWiki set you wish to retrieve (e.g. http://datawiki.googlelabs.com/wiki/Guestbook, simply visit:
http://telegraphis.net/data/datawiki/rdfxml.php?uri=<datawiki-uri>
and the contents of the DataWiki will be transformed into RDF/XML.
As of right now, this conversion only works for DataWikis hosted on datawiki.googlelabs.com itself (and not separate hosts), although the XSL file should work with any arbitrary Atom feed that follows the same format as those produced by DataWiki.
I currently reference the data-item URIs in the <link> element, and not the UUID URNs in the <id> element, as the former are at least dereferenceable, which is a good thing. I'm not entirely sure why the latter are not made the same as the former, other than for relocatability reasons, since atom:id elements must not change. But since Linked Data assumes that all URIs are this way, there's no reason not to make the URI where the data item was initially created into the atom:id of the item.
Unfortunately, XML output does not currently work on individual data items, so they cannot be dereferenced via this service. Queries also do not appear to work with DataWiki at this time and are not proxied through.
As for namespaces, I currently use the format namespaces (e.g. http://datawiki.googlelabs.com/wiki/formats/Guestbook) as the RDF namespaces for all items. Unfortunately, these only dereference to HTML descriptions of the Web Service API rather than the schema themselves, and the namespaces lack the fragment identifier (so I have taken it upon myself to add one to keep it simple, even though this isn't necessarily correct).
This service does provide CORS headers for remote AJAX requests (e.g. for rdfQuery), but only supports the GET method at this time, and thus is not directly supportive of Read-Write Linked Data. The Wiki also does not currently allow for linking or data-types, so we interpret all values as untyped RDF literals.