base-uri of parsed documents should be the retrieval URL

[Chris Hillery]

Given a file /tmp/foo.xml and a query /tmp/foo.xq containing

  fn:base-uri(fn:doc("foo.xml"))

the output is "file:///tmp/foo.xq", ie, the URL of the *query*, not of the document.

The XQuery spec refers to the XML Base spec when describing the value of the base-uri property on node. XML Base defines how things behave when there is an xml:base attribute somewhere, and defers to RFC 3986 (URIs) for other cases. RFC 3986 says the base URI of content should be the first of the following things that exists:

  (a) The base URI is embedded in the document's content (eg., xml:base attribute).
  (b) The base URI is that of the encapsulating entity (message, document, etc).
  (c) The base URI is the URI used to retrieve the entity.
  (d) The base URI is defined by the context of the application.

This bug is referring to (c); we should set the base-uri of new nodes from external documents to be the URI used to load the document. So the query above should return "file:///tmp/foo.xml" (because the base URI must always be absolute).

[Chris Hillery (ceejatec)]

This may be part of the cause of bug 1105092, which is causing FOTS failures, hence I'm making it fairly high priority.