Documents for review on today's XDI TC call

["Drummond Reed" <drummond.reed@cordance.net>]

XDI TC Members and Observers:

Per the agenda I just distributed, we will be reviewing the following
documents on today's call:

The v7 XDI metaschema proposal:

http://www.oasis-open.org/committees/download.php/9649/draft-xdi-metaschema-
v7.xsd

The v7 XDI addressing rules proposal:

http://www.oasis-open.org/committees/download.php/9650/xdi-addressing-propos
al-v7.doc

An extensively-commented example XDI document that illustrates the above:

http://www.oasis-open.org/committees/download.php/9651/xdi-examples-7-XSP-v1
.xml

The major advancement in this version of the metaschema is that it is now
fully harmonized with the RDF 3-tuple model (it consists of only 3 elements
- Resource, XRI, and Data - and no attributes). This means we now have a
much better answer to the question, "How does XDI differ from RDF?", which
is: "It's doesn't - at the level of the 3-tuple graph model. It's a
different serialization format optimized for graphs of assertions about data
identification, exchange, and control that XRIs as the identifier for each
RDF node that otherwise would have a URI or a bnode identifier."

The following table makes this easier to see:

3-TUPLE MODEL	SUBJECT		PREDICATE		OBJECT

In RDF can be	URI or bnode	URI			URI, bnode, or
literal

In XDI can be	Resource		Resource		Resource or
Data


In XDI a Resource can model either a subject, predicate, or object because a
Resource MUST have at least one XRI. The means if the Resource is:

* A Subject, then the XRI is either the URI or the bnode ID value.
* A Predicate, then the XRI is the URI value. 
* An Object, then the XRI is either the URI or the bnode ID value.

And if the Object is a literal, then it is serialized as an XDI Data
element.

As a serialization format, XDI is relatively compact compared to RDF
because:

* Statements about statements (nested to any depth) from the same authority
can be made just by building trees of XDI Resources.
* Statements about statements from other authorities can be modeled using
XDI references and link contracts.
* When there are multiple arcs (Predicate URIs) that can describe the
relation between one Subject and one Object, all of them can be serialized
using one Resource modeling the Predicate and assigning it multiple XRIs.

But most of all, using XDI serialization, every node at every point of the
XDI graph (meaning every "XDI statement") is addressable - and
cross-references can be made across between any two nodes using the "$ref"
resource.

This should also allow us to go from seamlessly from RDF/XML serialization
to RDF/XDI serialization (and vice versa, except for the challenge of when
an XRI permits expressing something that's not validas an XML QName,
although there's probably a normalization algorithm that can be worked out
to overcome that issue.)

I look forward to discussing this further on the call today.

=Drummond