Define support for fragment identifiers in URIs on p:import

[xml-project]

Currently the spec does not say anything about handling fragment identifiers in URIs on p:import. The same is for p:import-functions. The XSLT spec at least says that support is implementation defined.
I think we should say some more about this case. Here are my suggestions:

• Support for (types of ) fragment identifiers should be implementation defined.
• A processor not supporting (a given type of) fragment identifiers <RFC 2119>must</RFC 2119) raise a specific error (XS0052 or other).
• The document referenced by the URI (without fragment identifier) <RFC 2119>must</RFC 2119) be a well formed XML document. (I do not think we need to deal with xml islands in plain text documents, do we?)
• It is an error if the fragment identifier does not specify exactly one element node.
• All in scope namespaces on that element are respected for XProc processing.
• The evaluation of the fragment identifier is done before any evaluation of static options or [p:]use-when expressions.

I am not sure, I cover all relevant points, so please feel free to add comments.

[xml-project]

Addendum: We need to consider whether and how support fragment identifiers has consequences for duplicate, circular chain of imports, or a re-entrant import. E.g. consider import via id-fragmenter and xpointer-selector point to the same element. Is there a way for a processor to know, that this is a case of duplicate import?

[ndw]

That all sounds reasonable to me.