xml semantics

I've been thinking recently about the relationship between rdf and xml. It
would be cool if there could be some sort of "semantic reunification" of the
two down the road. Has there been any/much work to provide a formal
semantics for xml in terms of rdf (now that rdf is getting it's own that
is)?

In the past I've used rdf to describe the xml dom as a means of bringing xml
docs into rdf. This can be useful in conjunction with rules (to perform
"schema" conversion) to pull the intended meaning from the xml but has the
unwanted effect of mixing the machinery of conversion with the data being
converted. It would be far more useful if the xml could be taken directly to
be making statements about objects not just about dom elements.

For example:

<a>
    <b attrib="value"/>
    <c/>
</a>

Could be taken to mean (using p s o order):

rdf:type s1 a
rdf:type s2 b
rdf:type s3 c
?p1 s1 s2
?p2 s1 s3
attrib s2 "value"
?p3 s2 s3

where ?p1 and ?p2 represent the as-yet-unknown implied relationships that
are being represented by xml containment. Note the last statement captures
the fact that there may be some relationship (?p3) between s2 and s3 that
caused the xml author to put them in a specific order in the xml doc. This
approach of treating all tags as resources connected by anonymous properties
seems more intuitively correct than treating them as properties of anonymous
nodes and more workable in a general sense than trying to infer a correct
subj/prop "striping" a la rdfxml. In a perfect world there would be some way
to specify in the xml schema what the implied relationship was in each case.

Anybody have any thoughts along these lines?

-Geoff Chappell

Received on Wednesday, 17 July 2002 11:43:52 UTC