Requesting top-down analysis of the XRD/XRI metadata discovery frameworks.

["Steven Churchill" <steven.churchill@xdi.org>]

Hi all,

 

FWIW, I will blurt out my (albeit unsolicited) thoughts on XRD/XRI metadata discovery. I think that it's important at this "early" stage to try to evaluate things top down.

 

N.B.: the terms I use here, such as "metadata discovery framework", are just placeholders. Feel free to substitute your own terms while reading.

 

From our TC's perspective, a "metadata discovery framework" has these characteristics:

 

1)       An entity space.

2)       One or more identifier formats.

3)       One or more "discovery functions".

4)       An entity metadata descriptor format returned by 3.

5)       An entity metadata descriptor format used by 3.

 

From this perspective, I think that it's safe to say that the XRI TC currently is defining two distinct metadata discovery frameworks. I will give them names as place holders.

 

·         The “XRD Metadata Discovery framework".
This is being defined in the XRD spec.

·         The "XRI Metadata Discovery framework".
This is being defined in the XRI spec.

 

It's important at this stage in the TC process to be able to compare and contrast these two discovery frameworks. However, that is beyond the scope of this email. What I'll do is discuss the five characteristics above.

 

1)       An entity space.
The entity space is largely defined by characteristics 2-5. For example:

Is it hierarchical?
Do characteristics 2-5 support the navigation of parent-child entity relationships?

Is it polyarchical?
Do characteristics 2-5 support the navigation to an entity via multiple parent relationships?

Do entities have "distinguishable identity"?
Do characteristics 2-5 support formally distinguishing one entity from the others in the entity space? (E.g., CanonicalID.)

2)       One or more identifier formats.
A single identifier is passed in to the "discovery function" in 3 which maps to a single entity in 1 (and then returns the entity’s metadata.) A discovery function could (conceivably) allow two or more formats (syntaxes) for the identifier.

3)       One or more "discovery functions".
A given discovery function takes as input a single identifier; it then negotiates the entity space to a single entity; and it returns metadata in the format defined in 4.

A given discovery function must take as input a single identifier, but it can also take as input other stuff--such as the Service Type and Media type in XRI 2.0. Note that in XRI 2.0 the mapping is said to be "context sensitive" in that varying the context (e.g., the service type) can result in navigating to a different entity for the same identifier.

A given discovery function might support “authenticated navigation” of the entity space and might produce an “audit trail” through the entity space as optional output.

4)       An entity metadata descriptor format returned by 3.
This is the output of the discovery function. It is what the client who invokes the discovery function is looking for. For 3.0, it appears that the same XRD format will be returned by discovery functions in both the "XRD Metadata Discovery Framework" and the "XRI Metadata Discovery Framework". I.e., the client gets back the same format when using either framework.

5)       An entity metadata descriptor format used by 3.
This is a metadata descriptor format used by the discovery function to negotiate the entity space. In XRI 2.0, it is returned by an invocation to an entity's "authority server" protocol. In XRI 2.0 this happens to be the same format as 4, but this is not necessary. (Unlike XRI 2.0 the TC needs to become more prudent with regard to gratuitous overloading.)

 

So we can run down how XRI Resolution 2.0 stacks up w.r.t these characteristics: it supports a polyarchical entity space with distinguishable identity; its identifier format is XRI Syntax 2.0; it's discovery function is context sensitive; it supports authenticated navigation of the entity space; it optionally outputs an audit trail; and so on.

It is not that clear to me how the above characteristics are defined for the new "XRD Metadata Discovery framework" or the new “XRI Metadata Discovery Framework”.

I think the most important thing at this point is to first understand that there are indeed two distinct discovery frameworks now being defined (both with different entity spaces; different identifier formats; different discovery functions; etc.)  It would also be really good for this TC to describe the characteristics of both frameworks so that they can be compared and contrasted.

 

Those be my thoughts. Back into the woodwork. J

 

~ Steve

 

PS: Note that under 4 above, is said that: it appears that the same XRD format will be returned by discovery functions in both the "XRD Metadata Discovery Framework" and the "XRI Metadata Discovery Framework". Again, I suggest that the TC exercises caution with regard to overloading here.