W3C home > Mailing lists > Public > www-rdf-comments@w3.org > April to June 1998

XML serialization of RDF

From: A. G. McDowell <mcdowella@mcdowella.demon.co.uk>
Date: Thu, 18 Jun 1998 19:44:08 +0100
Message-ID: <k7jTBDA4$Vi1EwKx@mcdowella.demon.co.uk>
To: www-rdf-comments@w3.org
Cc: Andrew.McDowell@scisys.co.uk
Comment on the serialization of RDF as XML.

I would like to serialize RDF into stretches of XML containing only
elements and attributes whose names do not depend on RDF properties. I
wish to be able to validate this XML against a DTD. I do not foresee any
need to distribute these instances to users in other organisations, but
I would like to make use of COTS standards and software wherever

The motivation for this is that I would like to generate and process
instances, much of whose structure can be agreed upon in advance, but
with room for expansion at very short notice. The idea is to agree a DTD
that defines elements and attributes hand-tuned to the agreed structure,
and can be used to enforce it via validation, but which includes one or
more elements that may contain stretches of serialized RDF. We wish to
be able to include previously unforseen property types in this RDF, and
we can't do this quickly enough if we need to modify the DTD to define
new elements for them.

Is there already a defined serialiation format that would allow this?
Both the Serialization Syntax and the Abbreviated Syntax defined in
http://www.w3.org/TR/1998/WD-rdf-syntax-19980216 use elements whose name
depends on the property they encode.

Is there any reason why this requirement is not sensible or feasible? I
can see that deriving element names from property names allows you to
use the XML namespace specification to stop accidental collisions of
property names. We could, however encode property types as attribute
values while still expressing them as namespace-qualified names.

This follows the example of the XML-Data note, which abbreviates
<element href="http://books.org/date"/> as <element href="bk:date"/>.
Some such mechanism seems to be required after reification, in any case.
When you express the statement "Ralph believes that the document 'The
Origin of Species' was authored by Charles Darwin" as RDF, the 'Author'
property becomes changed from a property in its own right, which you
serialize using the element <BIB:Author>, to the value of the
'RDF:Propname' attribute, which you serialize as either PCDATA in the
full syntax, or as an attribute value in the abbreviated syntax.
Science Systems (http://www.scisys.co.uk)
Received on Thursday, 18 June 1998 14:46:29 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:15:12 UTC