"Type-Tagged XML"

In today's telecon, it was again mentioned that some folks just don't
like RDF.  It was said that the problem with "Rigid RDF" is that it
has the term "RDF" in its name.  The point was made that there is a
significant market/perception issue here.

So let us instead call this proposed style of XML: "type-tagged XML".
The point of type-tagged XML is this: given an XML document in this
form, you can deserialized it into RIF frames (or RDF triples, or
objects in a dynamically-typed language (like Perl or Python), or
relational tables), without any out-of-band information.

A simple example might be:

    ...
       <Person>
           <age>42</age>
       </Person>
    ...

vs
    ...
       <Person>
           <age rdf:datatype="&xs;int">42</age>
       </Person>
    ...

In the first case, it's not clear whether "42" is to be understood as
a string or an integer.   In the frame/object/database form, you have
to know, but you don't without consulting some specification or
schema.  In the second case, we have type-tagging -- so you know that
"42" is to be converted to an integer.

To do this, in general, requires knowing the XML schema datatypes of
everything and knowing when things are lists.  And knowing that things
are fully-striped (for the object-types, aka "classes").  So, I
suggest the fairly-simply syntax changes I detailed earlier (as "rigid
RDF" [1]) are a good way to do this -- it's a way to do type-tagging
that happens to be aligned with RDF/XML.

There may be other negatives to this proposal.  It still has bits from
the rdf namespace, but the point was made very strongly in the telecon
that XML people are used to random stuff from random namespaces.  It
still is a little more verbose.  ...  Other problems?

     -- Sandro

[1] http://lists.w3.org/Archives/Public/public-rif-wg/2008May/0099

Received on Tuesday, 20 May 2008 18:06:12 UTC