Master for slides. (C) TimBL1997,8 Matches (more or less, with some incremental editing) talks given at

• WWW7 Brisbane Australia 15 April 1998

Title

XML: Enabling the Next Generation of Publishing on the Web

Abstract:

The World Wide Web Consortium has adopted XML as the common format for expressing structure in data.

This talk

• What is evolvability?
  • The myth of the "futureproof"
  • Designing to be part of something else
• The past, present and future of HTML
• XML and general language for structure
  • MathML  (Rec 1998/4/7)
• RDF and a general model for semantics
• Meta languages: schemas and reasoning
• Evolution of the technology

Evolvability

Tim Berners-Lee
Director, World Wide Web Consortium
Warning: Includes personal points of view.

http://w3.org/Talks

This talk

• Evolution of markup languages
• Evolution of data on the web

What is evolvability?

• The ability to evolve easily
• Technology and human organization
• Do you believe "futureproof"?
• Designing to be part of something else

HTML part 1

• Wide development follows from open standard
• Fragmentation threatens interoperability
• Meetings agree on common specifications
• and so on.

HTML offloads

• W3C Moves from fire fighting to consolidation
• Orthogonal issues moved to other specs
  • CSS2 (Proposed Rec. closes May 5)
  • Document Object Model
  • XSL
  • WAI guidelines...
• W3C's free Amaya client demos and tests

See User Interface Tomorrow  in W3C Track

Free extension

• Ignore what you don't understand
• Fast development
• Ambiguity
• Retrospective definition
• Catch-up standardization

HTML (not SGML), HTTP

Well defined interfaces

• Remote Procedure Call, Dist. OO systems
• All  (works) or nothing (error)
• Have to find common interface
• Example: SGML, PICS self-describing documents
  • Compatibility check
  • Dynamic class download possible
• Web can be language registry

HTML needs

• Example: MathML (Rec. 7 April)
• Mixing new features with standard
• Well defined extensions

HTML future

• Transition to XML
• XML namespaces allow mixing of vocabularies
• New extensions do not impact HTML spec
• New extensions do leverage style and DOM

(Future of HTML Workshop in May)

Language mixing

• New and old features both well defined
• New: does not compare with RPC interface
• Old: like program calling different modules
• More powerful model than single interface
• Reality: HTML + Javascript
• Reality example: Invoice.

Optional features and Partial Understanding

• Goal: V1 software partially understands V2 document
• Optional features visible as such
• Example: "Mandatory" Internet Draft
• Example: SMIL (P.Rec. 1998/4/9)
• Conversion from unknown language to known language.

See Architecture today in W3C Track

Test of Independent Invention

A Thought experiment

• Same system invented in parallel
• Philosophy the same
• Arbitrary decisions meet
• Systems meet

what happens?

(See Design Issues article)

Philosophy to requirement

• We will be smarter in the future
  • Moving Version 1 to Version 2
• We are not smarter than everyone else
  • Decentralized evolution
  • Moving between parallel Version A and Version B

Story so far

• Free extension but well defined extensions
• Vocabularies "registered" in Web space
• Documents with mixed vocabularies
• Optional parts identifiable
• Operation: Convert V2 document to V1

Evolution of data

Data characterized by

• Well defined structure
• Defined but often invisible semantics
• Huge legacy of relational databases etc
• Schemas defined by users

First data on web: Metadata

• PICS labels - endorsement
• Intellectual Property Rights
• Privacy policy - P3P
• Structure of information
• Cataloging...Dublin Core etc

See Technology & Society in W3C Track on Friday

Publishing data on the web

Example: A stock price database

1. Export data in XML for machine consumption
2. Possible dual XML/HTML view
3. Namespace points to schema
4. Schema gives processing information

Levels of schema language

1. Human readable
2. Define structure
3. Define structure and optional parts
4. Turing complete conversions
5. Create logical model of document

You can upgrade with time!

0: URI only

• No supporting documentation
• Allows compatibility yes/no test

1: Human readable

(e.g. HTML)

• Allows compatibility yes/no test
• Settles arguments

2: Define structure

(compare: DTD)

• Allows compatibility test
• Allows structure verification
• Allows use of common parser
• Allows graphic UI generation (?)

3: Structure + Optional flags

• Allows compatibility test
• Allows structure verification
• Allows use of common parser
• Allows Version 2 to Version 1 conversion

4: Turning complete language

(e.g. XSL with Javascript)

• Allows compatibility test
• Allows black box conversion of Version 2 to Version 1
• Allows black box conversion of Version A to Version B

5: Expose logic of document

(Resource Description Framework: RDF)

• Allows compatibility test
• Allows structure verification
• Allows use of common parser
• Allows Version 2 to Version 1 conversion
• Allows Verson A to Version B conversion

RDF engines should

as the technology evolves be able to

• Use a simple idea of what an assertion (property) is;
• See a document is seen as a combination of logical assertions;
• Deduce Version A document from Version B;
• Include notions of digital signature;
• Draw conclusions by combining many documents
• Prove a statement is valid from a given proof;
• Attempt arbitrary queries using a search engine

PS: Need RDB experts in XML and RDF groups!

RDF maps the real world

• Existing databases with partial relationships
  • Example: Personnel, computer user, rolodex
• Documents with mixed vocabularies
  • Example: Invoice, MARC records...
• Retrospective documentation of relationships

Engines of the future

• Logical systems provide good answers but don't scale
• Search engines scale (?!) but produce wrong answers
• Combination Global Resoning Engines could be devastating...

Examples

• Can Joe access the party photos?
• Who are all the people who can?
• Is there a green car for sale for around $15000 in Queensland?
• Did someone driving a blue car send us an invoice for over $10000?
• What was the average temperature in 1997 in Brisbane?
• Please fill in my tax form!

What to do

• Get involved with XML, RDF, schema and conversion languages
• Model your intranet information
• Design global fact indexes and global proof finders
• Design for evolvability

Thank you

For details of World Wide Web Consortium:

For slides on the web: