2002-02-25#19, Fragment identifiers, words for the primer

My action from the F2F:
2002-02-25#19 Graham
Draft text for the primer on use of frag id's with appropriate warnings re 
mime types

In drafting this, I've taken account of some off-list exchanges with Pat 
Hayes (I hope I've adequately reflected the essence of what you said, 
Pat:  some details are a bit different).

Of interest, I think, is that as presented here the presumed MIME-type 
dependency becomes something of a small side-issue rather than central to 
the debate (which I had originally assumed).

#g
--

<<<START>>>

Fragment identifiers, when used with RDF, are treated as a simple extension 
of the URI to which they apply, whose interpretation is not dependent on 
the context in which they appear.  This reflects the fact that there is no 
special treatment of the fragment identifier part of URIrefs in the model 
theory for RDF -- that is, they are simply a syntactic part of a name that 
denotes some resource.

There is sometimes an unwarranted expectation that the thing identified by 
a URI with fragment identifier bears some particular relationship to the 
thing identified by the URI alone.  For example, the RDF statement:

    urn:isbn:0-520-02356-0#page10 ex:contains "metatheory" .

might be regarded as having a particular relationship to the statement:

    urn:isbn:0-520-02356-0 dc:title "Metalogic" .

but this would be an error.  As far as RDF is concerned, 
'urn:isbn:0-520-02356-0#page10' and 'urn:isbn:0-520-02356-0' are two 
different names with no defined relationship.  This is different from the 
normal use of fragment identifiers when retrieving web documents, where the 
URI with fragment identifier is taken to represent some view of the 
document referenced by the URI alone.

This is not to say that a URI and that URI with fragment identifier may 
never be related, just that no such relationship is presumed by 
RDF.  Returning to the example above, it is quite possible that some RDF 
document defines a relationship between these terms:

     urn:isbn:0-520-02356-0 rdf:type ex:Book .
     urn:isbn:0-520-02356-0 dc:title "Metalogic" .
      :
     urn:isbn:0-520-02356-0 ex:consistsOf _:a .
       _:a rdf:type rdf:Seq .
       _:a rdf:_1 urn:isbn:0-520-02356-0#page1 .
       _:a rdf:_2 urn:isbn:0-520-02356-0#page2 .
        :
       _:a rdf:_10 urn:isbn:0-520-02356-0#page10 .
        :
       (etc.)

This RDF graph makes specific assertions about relationships between things 
denoted by the URI and URI-with-fragment-identifier.

Finally, note that in the special case of a document containing RDF/XML 
statements (MIME type application/RDF+XML???), the syntax presumes a 
convention for relating the document name to the resource names whose 
definitions it contains.  Specifically, resources described using an 
rdf:ID='...' attribute have an identifier that consists of the RDF document 
URI plus a fragment identifier of the given rdf:ID attribute value.  But 
observe that this is a purely syntactic convention, and does not of itself 
presume any semantic relationship between the defining document and the 
thing defined.

<<<FINISH>>>



------------
Graham Klyne
(GK@ACM.ORG)

Received on Thursday, 7 March 2002 12:13:44 UTC