The TEI is dead

Long live the TEI

C. M. Sperberg-McQueen

Pisa, 16 November 2001

1. Overview

2. Some things we did right

3. What to do better now

4. Political / organizational

(Why are you listening to me?)

5. Hopes

  • retain globalist / universalist aims
  • retain and expand international base
  • wider adoption and awareness

6. Technical program

Always and foremost: maintaining and improving the tag sets.
Some overarching challenges:
  • extensibility
  • modularity
  • integration with practice

7. Extensibility

  • The key: partial understanding.
  • The closed-world hypothesis: just say No.
  • Knowing how to extend in advance.
(How will we know what we think until we see what we do?)

8. Extensibility

Consider a precedent:
  • HTML's extensibility rule.
  • Two kinds of extension
    • ignore the tags
    • ignore the elements
    but only one kind of support.
  • This cup is half full.
Bottom line: err boldly.

9. Modularity

TEI pioneered modularity in DTDs: Pizza model.
We can do even better:
  • make the pizza model generically usable (Chinese menu!)
  • integrate better the work done elsewhere
    • MathML, SVG, etc.
    • CALS or HTML tables
    • Docbook
  • easier extension of TEI

10. Modularity how?

  • cleaner definition of extension
  • easier for developers to recognize and support standard TEI
  • easier to support single tag sets

11. Namespaces: half empty

The W3C Namespaces Recommendation does a lot less than meets the eye.
  • no magic (or non-magic) information about meaning
  • no universally unique names
  • no information about markup vocabularies or document types
This glass is mostly empty.

12. Namespaces: half full

The W3C Namespaces Recommendation does just one thing:
  • It defines one* way to tell your stuff from everybody else's stuff.
Little things can go a long way.

13. Namespaces: some issues

Namespace names are uniform resource identifiers (URIs). E.g.
What goes at that URL?
Namespace mechanism supports direct reuse.

14. Namespaces: some issues

Namespace names are uniform resource identifiers (URIs). E.g.
What goes at that URL?
Namespace mechanism supports direct reuse.
It does not support naturalization of predefined semantics.
Revive architectural forms?

15. Integration of theory and practice I

Integrate and cohabit with work done elsewhere:
  • Unicode
  • XPath
  • XML Schema, Relax NG, ...

16. Unicode

17. XPath and EPN

The TEI Extended Pointer Notation
  • implemented, but not really widely
  • numerous newer specs, interrelated:
    • XPointer
    • XSLT
    • XPath
    • various subsets of XPath (XSLT match, XML Schema, XForms, ...)
  • What is to be done?

18. XPointer -- issues

  • simplicity
  • expressive power
  • interoperability
  • semantic consistency (intertranslatability)
  • syntactic consistency (interoperability)
Wait and see?
Alternate EPN and XPointer tag sets?

19. Schema languages

There are lots and lots of XML schema languages:
  • Sox, Biztalk (XML Data Reduced), DDML, DSD, ...
  • Relax, Trex, Relax NG
  • XML Schema (W3C)

20. Other markup languages

Integrate and cohabit with work done elsewhere:
  • MathML
  • parts of XHTML
  • SVG
  • SMIL?
  • CALS/SGML Open/Oasis tables, HTML tables, ...

21. Theory and practice II

The community needs continuing work on `markup theory':
  • Thesis: markup represents an interpretation of the text
  • perfect markup is thus markup which perfectly captures our theory of the text;
  • a perfect markup language captures our theory of what is possible in texts.

22. Long live the TEI

The value of the TEI lies not any particular organizational form,
nor even in a particular SGML or XML DTD,
but in its role as an expression of our will to ensure ...

23. Long live the TEI

  • that the resources we create are reusable, because not tied exclusively to a particular application
  • that we work cooperatively to produce shared semantics for as broad and deep a universe of texts as possible
  • ... for behold, if they speak a single language, nothing they wish to do will be impossible for them.