Minutes: XDI TC Telecon Thursday 1-2PM PT 2009-04-02

Following are the minutes of the unofficial telecon of the XDI TC at:

Date: Thursday, 26 March 2009 USA

Time: 1:00PM - 2:00PM Pacific Time

ATTENDING

Giovanni Bartolomeo

Markus Sabadello

Nick Nicholas

Drummond Reed

John Bradley

1) CONTINUE $HAS AND $HAS$A DISCUSSIONS

Our main topic was discussion the open issues dialog at the bottom of:

http://wiki.oasis-open.org/xdi/XdiMetamodel

Giovanni started with the statement inferred by =abraham/$has/+son, which is =abraham+son/=isaac/=abraham+son=isaac. His first concern is that from an RDF standpoint, the semantics may not be clear. His second concern is that they may conflict with the semantics of the original RDF document.

RE the first concern, we discussed what it means that =isaac appears as a predicate in the statement =abraham+son/=isaac/=abraham+son=isaac. Drummond pointed out that this is a valid RDF statement, however it is not the type of statement for which RDF properties are typically used.

Giovanni said that the problem is that =isaac is an instance, and you can't use an instance as a property. Drummond pointed out that it is common in XDI RDF for a subject to also be a predicate, in fact that is the entire premise of XDI RDF dictionaries: every predicate appears in the dictionary as a subject. He used the example of +email as a subject, predicate and object:

+email/$is$a/$a$xsd$string

=drummond/+email

=drummond/+email/(=drummond+home/+email)

=drummond/+email/(=drummond+work/+email)

=drummond/$has$a/+email

Nick gave the example of putting a person in context, for example, placing =drummond in two different contexts: @cordance=drummond and @oasis=drummond. The $has statement for the first is

@cordance/$has/=drummond

This infers the statement

@cordance/=drummond/@cordance=drummond

Here is a good example of where =drummond (typically looked at as an “instance”) has a clear use as a predicate identify a specific context for =drummond.

Giovanni then gave two specific examples:

=abraham/+son/=isaac

=abraham/+son/=abraham+son

He said that in the first case, the range of +son would be instances of the class +son. Whereas in the second case, the range is another class – a specialization of +son.

Giovanni pointed out that in the first case the object is an instance of the range of the predicate, but not in the second case. So the second case is an exception to the normal behaviour in RDF.

Drummond pointed out that this is the fundamental “extension point” for all of XDI RDF. $has statements are the way of creating all the classes and instances that exist, and the addresses for those classes and instances.

He said this is similar to the concept of defining new names in any language, such as English. For example, “A plane made by Boeing is called a Boeing plane.” That’s how you create the new concept, “Boeing plane”.

RDF does not have such a concept because, to our knowledge, there are not RDF statements that create new URIs. URIs are opaque in RDF. So when i XDI RDF the statement =abraham/$has/+son asserts a relationship of two identifiers that results in a third identifier, we have stepped outside of the semantics that can conventionally be expressed in RDF.

Drummond wondered whether the way to make this compatible with RDF to propose that there is an algorithmic way to extend RDF ranges in XDI RDF, and it is built into all XDI RDF range types. This “range extension algorithm” would be defined as:

“Any XRI that is a concatenation of XRI A with XRI B is an extension of the range of XRI B.”

This would meet the goal Giovanni suggests of keeping XDI RDF as close to original RDF as possible, without taking away the power of XDI RDF addressing.

We closed by Giovanni saying he would like to propose a different solution for how XDI RDF addressing could work, which will will make the subject of next week's call (and discussion on the list this week if there is time).

# ACTION: GIOVANNI to write up his proposed alternative for XDI RDF addressing.

xdi message