Tim Berners-Lee
W3C and AC Meeting, 2008 April 21

Saving the Web #27

Tim Berners-Lee, <timbl@w3.org>

Saving the Web

Threats:

• Fragmentation of standards
• Lack of modularity and extensibility

• Challenge from proprietary stacks

(See many recent blogs)

Fragmentation

avoiding it involves working together

Story: Cross-domain access

(Problem was: A malicious script running in a web page on your browser could access pages from inside your firewall and leak the data back to the bad guys, so access to other domains than the source of the script was blocked, but that was a pain)

2005-06 W3C starts work on access-control

2007-05 Microsoft reviews W3C spec as good

2007-12 Microsoft mentions work on alternative XHR2

2008-03 Microsoft Reveals XDR as a competitive alternative

Fragmentation: The XML-HTML tension

• HTML WG: no convergance with XML
  • But: There is an XML serialization
  • Divergence at the DOM level
  • Complex technical issues, e.g. document.write()
• Rejection of Namespaces
• XML issues...

Cost: Two stacks. More code. Slower startup. More bugs.

XML Issues

• Quotes around simple attribute values c="a" r="2"
• Repeating the tag name at close </TransferLicenceRecordName>
• Parsers stop at first error
• DTD legacy - obsolete
• Schema - complexity
• The bulk of namespace URIs
• The pain of having to close things you have opened
• Looking for URI-based extensibility -- from NVDL and RelaxNG?

  How to say Oh, an SVG circle can go anywhere an HTML IMG can go?

Why namespaces anyway?

• Introduced as a requirement by RDF
  • Never really fitted with non-RDF use?
• Goal: Decentralized extension
  • Not a goal for HTML??
• Modular design allows independent groups working on independent parts

Requirement: Follow your nose

Fundamental rule and property of the web

• The meaning of a document follows from the URI alone
• Based on:

  Transmitter and receiver having read the same set of base specs

  Bootstrapping using the Web.

• Eg: Mime type -> spec, XML spec -> Namespace

(see earlier talks, note about the stack of specs)

Assumption: Scale-free

• There is no "typical size of community"
• The global community is one size of community

  Lot of effort, short names agreed, no need fro URIs

  E.g. HTML (SVG?)

• Very large communities

  E.g. MathML, SVG?, FaceBook Markup Language, etc

  Need identifiers, URIs work, too long?

• For RDF, many smaller communities

  E.g. Protein scientists, History museums, Today's experiment

  Need URIs - anyone/anything can invent one at any time

  Needs also a very powerful extensibility model

(See Design issues note)

Designing for the scale-free world

Optimize each level appropriately

• HTML should not have to deal with URIs

  Make the HTML namespace implicit, set by the Content-Type

• SVG and MathML should be lighter

  Make shortcuts also bootstrapped from HTML

• Other communities are also really important

  Keep the ability for a random person to extend HTML for local use

Story: ARIA

Accessibility for Rich Web Applications

• Design need: add ~30 new attributes to xHTML & similar

  In XML: add namespace aria:foo

  In HTML: add attributes aria-foo

• Issues:
  1. Users confused
  2. Have to get HTML WG (etc?) to accept the attributes
  3. Scripts see this differently in XHTML and HTML
  4. General divergence of code

We have two (+?) working groups chartered in the HTML namespace

Story: RDFa

• Design need: Add semantics to *HTML

  Previous design: Overload class attribute with complex syntax

  In HTML: Add new attributes

  Issues:

  1. How to follow your nose? GRDDL? Change HTML spec?
  2. How to describe ARIA?
  3. Does HTML group need to be in the loop?

Story: Microformats

• Design Need: Add application-specific semantics to *HTML
• Design: Overload class attribute
• Issues:
  1. How to follow your nose?
  2. Every client has to learn each new microformat
  3. (GRDDL would work but GRDDL not microformats deployment path)

Motivator? Getting past the the W3C validator

Postel's maxim

Be conservative in what you produce, liberal in what you accept

A very important maxim in Internet protocol design

How has this left the web?

The conservative validator

The liberal browser

The motivating slope

The motivating slope

Extensabiliy and the Online page checkers

We need online tools which will:

• Point out simple syntax errors but continue
• Point out serious errors and continue
• Encourage (good) XML but not require it
• Accept and warn unrecognized elements and attributes, and continue
• Encourage the use of namespaces, looking up new namespaces found
• Rate a page with points according to severity

Is it possible to make a unversal parser for HTML5 and XML?

Q: More resources for online tools?

Fixing web pages: Browser to-do

1. View Source: show cleaned up source
2. Save As: save cleaned up source
3. This is my website -> show errors

HTML Mime type issues

• An excercise in deployment strategy
• You can't serve nice XHTML up as text/html :-(
  1. Character encoding issues: Different defaults
  2. etc
• IE does not recognize application/xhtml+xml Content-Type

Threats to XML

• People start embedding HTML within XML (e.g. in RDF), forcing the HTML parser to be deployed much more widely
• HTML adds all kinds of things
• HTML WG defines different extensability mechanism from NS
• Systes whcih are all-xml now, including SOAP and B2B, end up having to include HTML syntax browsers.

Should XML meet HTML halfway? XML2.0

• Allow attribute quotes to be omitted for simple values.
• Allow namespace to be default, given Content Type
• Short-hand for switching from one namespace to another? (grounded in Namespace document)
• Short form of close tag </> ?
• Remove DTDs.
  • Only buit-in entities. - or declared using elements?
  • No fixed or default atributes
• URIs for grammar, cross-schema links for mixed NS
• Remove PIs (Have a xml:pi element if you like?)
• ";" as alternative to "><" to simplify whitespace??
• Multiple root elements or mixed content as document?
• etc ... ?

(See Tim Bray 2002, Norm Walsh on XML2.0 in 2004, 2008),...

Threats to modularization

• Include all ARIA in HTML
• Include all RDFa in HTML
• Include all of SVG in HTML (in HTML syntax)
• Include all MathML in HTML

What it will become is what all 350 page specs become -- a 650 page spec! - J. Klensin, 19?? about something else.

Costs of lack of modularization

1. Testing all combinations
2. Product development time
3. Spec development time
4. Short supply of people who understand the spec
5. Users learning time?

Social modularization

The first 13+ years of W3C involved a constant theme of designing a modular organization, which would design a modular system in a decentralized way. Working groups coordinate with others inside AND outside W3C, technologies fit together like a quilt and build on each other like Lego. The lack of modularization in HTML5 follows from the lack of modularization in the HTML5 group. *

Each group, whether or not in W3C, has a duty to act as a responsible peer to other groups, recognize it is part of a larger communities, and to spawn independent subgroups to do cleanly defined parts of the work when the task is big.

It takes time. There is no free consensus.

* (Fred Brooks, Mythical Man Month, structure of system follows structure of organization)

Browser profile?

Modularization though leads to fragmentation in the selection implemented.

• HTML
• CSS
• HTTP
• DOM
• Ecmascript
• Graphics - JEPG, PNG, SVG
• Video codecs
• ...

Should we work more on user environment profiles?

The W3C Web Applications platform

• HTML+SVG+DOM+WebAPIs+SPARQL is actually a very powerful platform
• SVG is a great technology, implemented by every browser but one.
• The basis of the 'web 2.0' wave
• Competition from proprietary stacks:
  • Flash always been a non-standard part
  • Now silverlight and AIR compete

Questions for Advisory Committee

1. Priorities?
2. Press HTML WG to be modular?
3. Make new simpler, extensible, XML?
4. Design for the future vs. accomodate the past?
5. Resources for new page-checker tools?
6. Push Web Applications hard as the future platform?