Minutes: XDI TC Call Wed. 7 December 9PM-10AM Pacific

Minutes of the unofficial XDI TC telecon 2005-12-07

 

Date:  Wednesday, 7 December 2005 USA

Time:  09:00AM - 10:00AM
Pacific Time

 

Dial-in number: 218-936-6666

Access Code: 2647893

 

PRESENT

 

Jason

Drummond

Andy

Owen

 

 

AGENDA

 

1) DISCUSS XDI TRIPLE TRIPLE MODEL

 

We discussed a potential proposal, inspired both by work on the
"i-tag" initiative (see

http://www.seedwiki.com/wiki/itag/itag.cfm and in particular

http://www.seedwiki.com/wiki/itag/power_i-tags.cfm?wpid=220488) as well
as

work on simple link contracts (as requested by Victor).

 

The proposal is that the XDI graph can be optimized not by just
limiting it

to 3 levels deep, but also 3 levels "across". This means that
everything in

the graph can be represented by an RDF "triple triple" -- a
maximum of 3

linked RDF triples - a Subject triple, a Predicate triple, and an
Object

triple. See the additional
reference information appended below.

 

The group discussed the
potential advantages of this model, principally that it would produce
well-known "predicates" that enable parties to programmatically
construct the XRIs representing the well-known XDI address of a data object in
the context of a known contract in the context of a known relationship.

 

Before we could delve
into this deeply we discussed how XRI resolution would work using XDI documents.
This led into a long discussion about the type of XDI document that would be
returned from a request to an XRI registry. We agreed that there should be a
clear model for how XRI resolution is modeled using XDI documents, and that
this should be documented by the XDI TC.

 

We concluded with an
action item to create and post sample XRIs that would meet the "triple
triple" or "3D" model for XDI addresses, particularly for link
contracts. The goal is to do this before next week's call so we can continue the
discussion then.

 

# ACTION ITEM –
ALL: Compose and post sample XRIs that meet the constraints of the triple
triple model.

 

=Drummond 

 

***REFERENCE MATERIAL***

 

Following is an HTML excerpt from the early draft i-tag spec that explains the "triple triple"
model in more detail:

 

I-tag syntax is based on the RDF "subject-predicate-object"
statement model.XRIs are used to serializing these RDF statements because:

XRIs
support the use of global identifiers within other global identifiers (i.e.,
XRIs within XRIs, or URIs/IRIs within XRIs) in order to express structured,
cross-context relationships.

XRI
syntax enables the persistent identification of the resources (people,
organizations, dictionaries, media objects) participating in an i-tagging
relationship.

As applied to tagging,
the RDF + XRI model maps as follows:

 

  
  
Statement
  part

  
  
  
Usage

  
 

 

  
  
Subject XRI

  
  
  
XRI identifing the
  authoritative source of the tag

  
 

 

  
  
Predicate XRI

  
  
  
XRI identifing the tag
  used to describe the media object

  
 

 

  
  
Object XRI

  
  
  
XRI that identifing the
  media object being tagged

  
 

 

This RDF model is then
applied recursively one layer deep, so each of the three XRIs above can also be
expressed using its own RDF triple. This yields the following matrix:

 

  
  
 

  
  
  
Subject

  
  
  
Predicate

  
  
  
Object

  
 

 

  
  
Subject
  XRI

  
  
  
XRI identifying the
  author asserting the i-tag
  (e.g. 

  i-name or URL)

  
  
  
XRI identifying the type
  of tag source (e.g., a blog, website, album)

  
  
  
XRI identifying the
  instance of the tag source (e.g. a blog name, web page name, album name)

  
 

 

  
  
Predicate
  XRI

  
  
  
XRI identifying the
  author defining the i-tag (e.g.
  

  i-name or URL)

  
  
  
XRI identifying the type
  of tag (e.g., tag, license, rating)

  
  
  
XRI identifying the
  instance of the tag type (e.g., tag name, license name, rating no.)

  
 

 

  
  
Object
  XRI

  
  
  
XRI identifying the
  author of the media object (e.g. 

  i-name or URL)

  
  
  
XRI identifying the type
  of media object (e.g. blog post, video, song)

  
  
  
XRI identifying the
  instance of the media object (e.g. blog post number, video name, song name)