ISSUE-63: Metadata Architecture for the Web

Outline

1. Received Web metadata architecture
2. Looming interoperability issue
3. Metadata accountability (AWWSW)

Goal: determine next steps on interoperability issue

JAR 2011-02-09

Generic approach to TAG issues

• Get oriented to problem domain
• Review current theory (specs)
• Document current practice
• Document theory/practice discrepencies
• Document pain points
• Assess reform/revolt tradeoff
• Reform or revolt

Background

• Web (1989)
• HTTP (1990?)
• Metadata architecture (1997)
• URI (1998)
• Resource description framework (1999)
• httpRange-14 resolution (2005)
• 'Authoritative metadata' finding (2006)
• Use of redirects, ISSUE-57 (2007)
• AWWSW (formed 2007)
• POWDER (2009)
• Web linking (ISSUE-62; RFC 5988, 2010)
• Durability, P2PHTTP, etc.

Received wisdom (webarch)

• IR = metadata subject = document, image, ...
• Use URIs to refer to IRs
• When writing metadata, use a URI to specify (select, name, 'identify') its subject
• GET/200 is somehow binding on which IR is meant by a URI

<http://example/cheese> dc:creator "Jiang Yuhe".

Deployment

• Dublin Core - dc:title dc:creator ...
• FOAF - foaf:topic foaf:homePage ...
• CC REL - xhtml:license ...
• RDFS, SPARQL, OWL, ...
• ...

Consensus at risk - syntactic space threatened

Because Web URIs are reserved for their IRs, other URIs have to be used for other things. Those who find this difficult have their eyes on Web URIs.

HH: "It [303]'s not a standard. It's on no W3C Recs. It's a W3C TAG finding, and primarily something Tim wrote in a design note that got put forth by SWEO as Note. So there's no process to go through to change it to begin with."

"... people forget to put it [the #] there when writing and cut and pasting URIs, and it's applied inconsistently across the XML and RDF universes. It puts a magic redirect in where one isn't necessary to begin with."

The interoperability issue

1. Document R retrievable at URI U talks about a subject S (a document)
2. Tool #1 adds statement {U license L1} to R, meaning U=R (webarch)
3. Tool #2 adds statement {U license L2} to R, meaning U=S (alternative)
4. Tool #3 cannot correctly interpret both statements

The interoperability issue (2)

1. Document R retrievable at URI U talks about a screwdriver
2. Tool #1 adds statement {Bob wants U} to R, meaning he wants document R (webarch)
3. Tool #2 adds statement {Bob wants U} to R, meaning he wants the screwdriver (alternative)
4. Tool #3 cannot correctly interpret both statements

Notational approaches, webarch vs. alternative

where you see U, pretend that you see http://example/e

Can we maintain consensus?

• Start by documenting current practice
• Designate a place where conversation can take place
• Figure out compatible ways to address pain points

[Discuss process here. Echoes, TAG purview, important, fatigue, etc. More technical stuff follows.]

Deeper issues (AWWSW)

• Those expressing metadata are responsible for its correctness.
• How is the correctness of metadata assessed?
• How to reconcile 'authority' with accountability?
• Exactly how is GET/200 informative?
• If the architecture is not rigorous, why would anyone risk using it?

What does metadata express - intent or truth?

Robin Berjon: "There shouldn't be a way of specifying dependencies external to the content or it's authoritative metadata all over again."

Media type and dependencies express intent: You can't hold me responsible if my content is interpreted in some other way.

Other metadata is objective - e.g. authorship, size.

How is metadata true?

A: Check it using GET/200.

• Ordinary metadata properties are representation invariants.
• Example of extraordinary property: is a Oaxaca weather report
• Properties of http: resources are predicted, not known.
• (obligatory Popper reference)

Why would I let myself be held accountable for my referential use of someone else's http: URI, if GET/200 has authority over reference and the URI owner is in control of GET/200?

1. It doesn't matter very much if I'm wrong.
2. I can successfully predict the owner's behavior.
3. I trust the owner.
4. I can hold the owner accountable.
5. Maybe the authority is a 'law' instead of an 'owner'.

WRONG QUESTIONS:

• What is an information resource?
• When is X a representation of Y?
• What does a URI 'identify'?

Segue to persistence discussion

Actionable persistence = a way to hold services accountable, and a system for obtaining universal consensus on what they're accountable for

Persistent Reference Interventions