Identifying Resources on the Web and Beyond

This presentation

http://www.w3.org/2008/Talks/0327-pid-fs/

Disclaimer

I am responsible for the materials and opinions expressed in this talk. I do not claim that there is consensus about them within W3C.

Key background documents

Published by the W3C Technical Architecture Group (TAG)

Topics

Web architecture: basic concepts
Metadata (in http: URIs)
Persistence of [http:] URIs (identifiers?) - a social challenge
Current and future issues about identifiers
Conclusion: Recommendations for persistent identifiers from a Web perspective

Topics

Web architecture: basic concepts
Metadata (in http: URIs)
Persistence of [http:] URIs (identifiers?) - a social challenge
Current and future issues about identifiers
Conclusion: Recommendations for persistent identifiers from a Web perspective

Key concepts: URIs, Resources, Representations

Three architectural bases of the Web

Identification of a resource (a weather report) via an URI
Interaction via a protocol, e.g. HTTP, Mail, Atom, ... (to retrieve current weather report)
Formats used by protocols (media type metadata, for retrieval or submission of representations)

Can http: URIs be used for things which are not web pages?

People, cars, concepts, ...
How does a user agent on the Web know what an http: URI identifies: an information resource on the Web or "real world objects"?
What do HTTP URIs Identify? says: successful retrieval (HTTP response code 200) means "identification of a resource"

Example: An http: URI for Alice

The 303 HTTP response code helps to identify Alice as a person, and to provide RDF data about her (for machines) or a Web page:

(From Cool URIs for the Semantic Web)

Dereferencing http: URIs? Not always

HTTP is an URI scheme which provides mechanisms for dereferencing HTTP: URIs.
However, there is no need to dereference all HTTP: URIs!
Some HTTP: URIs are better not dereferenced regularly:

http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd

Some URI schemes like Uniform Resource Names have no inherent mechanism for dereferencing

Do I need to invent a new scheme?

Like:

magnet:?xt=urn:sha1:YNCKHTQCWBTRNJIV4WNAE52SJUQCZO5C
info:lccn/2002022641
tag:hawke.org,2001-06-05:Taiko
@Jones.and.Company/(+phone.number)
xri://northgate.library.example.com/(urn:isbn:0-395-36341-1)

If at all possible, please don't:

"A specification SHOULD reuse an existing URI scheme (rather than create a new one) when it provides the desired properties of identifiers and their relation to resources."

Topics

Web architecture: basic concepts
Metadata (in http: URIs)
Persistence of [http:] URIs (identifiers?) - a social challenge
Current and future issues about identifiers
Conclusion: Recommendations for persistent identifiers from a Web perspective

Metadata used during HTTP-based interaction to decide about representations

Media type information in the HTTP content-type header tells a user agents what type of content to expect:

text/html
image/png
audio/mp3
...

Metadata - who has precedence?

Example: HTTP content-type is text/plain
http: URI is http://example.com/example.xml - but this must not be processed as XML, not if it's text/plain!
"Web software MUST NOT depend on the correctness of metadata inferred from a URI ..."

(From The use of Metadata in URIs)

Metadata in URIs: Document your Assignment Policies

"For the best weather information for your city, visit http://example.org/weather/your-city-name-here."

http://example.org/weather/your-city-name-here

(From The use of Metadata in URIs)

Identification of related resources with the same http: URI

Example: I want to GET the weather report in German or in English

GET /weather/
Host: example.org
Accept: text/html
Accept-Language: de, en

Content negotiation will get me the German, English or the default weather report

Identification of related resources with various http: URIs

Canonical http: URIs identify a generic resource
Enable discoverability of alternative representations by leveraging the hyperlink structure of the Web

http://example.com/mycorpus.html
http://example.com/mycorpus.html.en
http://example.com/mycorpus.html.en.v1
...

(Based on On Linking Alternative Representations To Enable Discovery And Publishing)

Identification of related resources via RDDL

Description of a target resource
Linked resources with a relation to the target resource
"nature" of links are URIs, eg. http://www.w3.org/2001/XMLSchema
"purposes" of links are documented at http://www.rddl.org/purposes/

RDDL example

From the W3C namespace document for ITS 1.0

<p>This document describes the namespace of the
<rddl:resource
    id="Core_Spec" xlink:title="Internationalization Tag Set (ITS) Version 1.0"
  xlink:role="http://www.isi.edu/in-notes/iana/assignments/media-types/text/html"
xlink:arcrole="http://www.rddl.org/purposes#normative-reference"
xlink:href="http://www.w3.org/TR/2007/REC-its-20070404/"
xlink:type="simple" xlink:show="none" xlink:actuate="none">
<a href="http://www.w3.org/TR/2007/REC-its-20070404/">
Internationalization Tag Set (ITS) Version 1.0</a></rddl:resource>
specification (November 2006 version).[...]</p>

Topics

Web architecture: basic concepts
Metadata (in http: URIs)
Persistence of [http:] URIs (identifiers?) - a social challenge
Current and future issues about identifiers
Conclusion: Recommendations for persistent identifiers from a Web perspective

[http:] URI Persistence - a social challenge

From URNs, Namespaces and Registries:

[http:] URI persistence: a management issue
Protocol dependency of [http:] URIs is just one kind of protocol dependency
[http:] URIs are not locations of information resources
[http:] URIs provide access to metadata

"URIs don't change: people change them." (Tim Berners-Lee)

Examples of policies related to persistence in W3C

W3C Persistence Policy

Various policies and guidelines:

Referencing electronic documents from W3C specifications
URIs for W3C Namespaces
Version Management in W3C specifications
XPointer Scheme Name Registry Policy

Example tools for checking persistence of W3C resources

W3C link checker
Namespace checker
Both are being applied within the W3C publication rules checker (usage example)

Persistency policies for URIs in various organizations

IETF does not maintain draft versions - dead links are intended, current drafts expire!
Draft for IETF Namespace URIs, including a registry
OASIS Filenames, URIs, Namespaces and OASIS Metadata and Versioning guidelines
If you need an example identifier, rely on the RFC 2606 "Reserved Top Level DNS Names" specification

http: URIs are cheap! Too cheap?

Everybody can create an http: URI, but:
Domain names can go away, change their owners, ...
How to assure persistence of http: URIs?

Persistence, Delegation and URIs

How to assure persistent access of representations of resources? Proposals from Henry Thompson:

Delegation as centralisation ("Put all your eggs in one basket, and watch that basket!"): all participants in the community agree on one repository for representations and one domain name
Delegation as replication ("Split up, one of us is bound to survive!") via two lookups:
- naming-system-specific naming authority; implements the
- second lookup, which identifies the repository

After all:

If a persistence mechanism is successful or not, is not mainly a technical problem
Your target community needs to be able to use the mechanism
If you have a limited amount of money, spend it (also?) for outreach about your approach

Topics

Web architecture: basic concepts
Metadata (in http: URIs)
Persistence of [http:] URIs (identifiers?) - a social challenge
Current and future issues about identifiers
Conclusion: Recommendations for persistent identifiers from a Web perspective

How to identify various resource types and sub resources?

Targets: text, sound, images, video, ontologies, time series, ...
Draft about identification of text fragments
Image: survey of Image Annotation on the Semantic Web
New area in W3C: Video on the Web
Planned Working Group Media Fragments Working Group

Internationalization of identifiers

Internationalized Domain Names - is http://example.straße.com equal to http://example.strasse.com?
Internationalized Resource Identifiers (IRI)
http://example.com/慶應義塾大学
Tasks: Comparison of identifiers with non-ASCII characters? Normalization & case folding?

URIs (not only http: URIs) have found some answers to these questions over the years

Identifiers for objects versus identifiers for services

Topic extensively discussed at W3C Workshop on Web of Services for Enterprise Computing (position papers, report)

Web Services using End Point References.
Contrast: URI-based, RESTful Services on the Web

Issues:

There is no canonical mapping of an EPR to a URI = no general method to identify an EPR-based service or to compare EPRs
Web Services offer benefits currently hard to realize with REST Web Services (e.g. security, reliability)

Topics

Web architecture: basic concepts
Metadata (in http: URIs)
Persistence of [http:] URIs (identifiers?) - a social challenge
Current and future issues about identifiers
Conclusion: Recommendations for persistent identifiers from a Web perspective

Personal conclusion about persistent identifiers on the Web

Use URIs, or better IRIs
Establish a policy and potentially tools which help checking how URIs in your field should look like
Global naming leads to network effects
Communities can define a system which implements persistence guarantees relying on http: URIs, or use http: URIs to allow the possibility to use the deployed Web infrastructure to tell you something useful about the thing the URI identifies -- whether that URI identifies a document or identifies something very different from a document.
Think twice: Do you need a new (URI) scheme?