Last modified: 2012-11-08 11:26:14 UTC
If i am requesting a page title that is not normalized like http://de.wikipedia.org/w/api.php?action=query&prop=langlinks&titles=New_York_City the response includes a normalized tag <normalized> <n from="new_York_City" to="New York City" /> </normalized> and the normalized page <page pageid="204956" ns="0" title="New York City"> For wbgetentities this should also be implemented. http://wikidata-test-repo.wikimedia.de/w/api.php?action=wbgetentities&sites=dewiki&titles=New_York_City currently only returns <entity site="dewiki" title="New_York_City" missing="" /> although http://wikidata-test-repo.wikimedia.de/w/api.php?action=wbgetentities&sites=dewiki&titles=New%20York%20City exists.
No, the normalization cannot be performed by Wikidata. It has to be done against the home wiki, e.g. first normalize against the Wikipedia and then use it on Wikidata. We do not know, e.g. that the namespaces "Benutzerin" and "Benutzer" on de.wp are normalized appropriately, so Wikidata would either need to ask the other wiki itself, or the caller does it beforehand. The first solution provides a potential DOS attack vector, I guess, so we go for the latter.
