rdf or xml?

From: David R. Karger <karger_at_mit.edu>
Date: Mon, 6 Jun 2005 17:34:19 -0400

Stefano wrote this:

    I believe the biggest mistake we made was to consider the output of a
    lens an RDF model (a tree encoded in a graph with explicit names of
    relationships) instead of thinking of it as an XML model (a tree encoded
    in a tree-oriented structure with implicit nesting relationships). This
    creates unnecessary complexity to the 'style' part of fresnel and for no
    benefit.

In response, I want to think separately about the model and the
encoding/syntax. It is clear that our "grouping" layer wants to
create a tree. But I think it is important for that tree to hook back
into the original rdf, in case the display device wants to look
further into the original rdf to do its job. Simplest example is if
the user clicks on a displayed object---we need to know what
underlying model object the user is addressing. So from a _model_
perspective, it seems to me that we have something kind of like a daml
list---namely, a tree layered over the original rdf. The tree is
defined by a "child" relation, and each node in the tree has a "value"
pointer to something in the original rdf model (or a literal).

Of course, if we serialized such a model to rdf-xml, it would be a
mess. So, we should make up a canonical encoding of the tree into a
nicer XML model. It seems to me we should be able to encode into
something that looks like natural rdf when all the values being shown
are literals, but that stretches naturally to contain rdf pointers back
into the underlying model. As long as we formally define our syntax,
we will not be losing any representational power in the tree encoding.
Received on Mon Jun 06 2005 - 21:32:24 EDT