Documents vs' Data, Schemas vs' Schemas

1
Documents vs. Data, Schemas vs. Schemas

• Bob DuCharme
• XML 2004
• November 16, 2004

2
(No Transcript)
3
Outline

• Data-oriented vs. Document-oriented XML
• meaning
• terminology
• relevance to choice of schema languages
• what about you?

4
How popular are these terms?

• 8/04 Google search
• xml data-oriented document-oriented
• over 700 hits (8/04)
• Google xml data-heads doc-heads over 140 hits

5
Data-head, Doc-head
6
Data-oriented XML

• XML as interchange format
• Data that began or ends up in an RDBMS
• A step in a transaction
• Transactions, XML, B2B

7
Document-oriented XML

• XML for eyeballs
• Bad fit for RDBMS
• More optional structures
• Roots in SGML world

8
McGrath's Power Law Distribution Theory
9
Element text node siblings

• ltparagtThis is a ltemphgtmocked-uplt/emphgt
  test.lt/paragt

10
Terminology

• Bob Glushko (xml-dev) transactional, narrative
• Rick Jelliffe (weblog posting) DBMS-oriented,
  publishing-oriented
• or... transaction-oriented, content-oriented

11
The Continuum

• Documents like RSS fall in between
• So why bother with the distinction?

12
W3C Schemas

• First Working Draft May 1999 Recommendation
  May 2001
• Some of the 31 member companies (DBMS vendors,
  aspiring B2B players) Microsoft (2), Intel,
  Informix, Xerox(2), HP, CommerceOne (2), IBM (2),
  Software AG, Oracle (2), Progress Software (2),
  webMethods, Sun(2).

13
W3C Schemas DBMS, OO Features

• Data typing
• nil values
• More sophisticated identity constraints
• Separation of type definitions from element
  declarations
• Type derivation, abstract types

14
RELAX NG Content-oriented XML

• Greater granularity in specifying content model
  flexibility.
• Attributes as part of content model
• Greater control over use of text nodes

15
Content Model Flexibility

• How to allow these
• ltnamegtJohn Smithlt/namegt
• ltnamegt
• ltfirstgtJohnlt/firstgtltlastgtSmithlt/lastgt
• lt/namegt
• but not this
• ltnamegtltfirstgtJohnlt/firstgt
• ltlastgtSmithlt/lastgt, Jr.lt/namegt
• which is what the DTD (and XSD) solution would
  allow
• lt!ELEMENT name (PCDATAfirstlast)gt

16
RELAX NG Content Model Flexibility

• ltgrammar xmlns"http//relaxng.org/ns/structure/1.
  0"gt
• ltdefine name"first.element"gt
• ltelement name"first"gt
• lttext/gt
• lt/elementgt
• lt/definegt
• ltdefine name"last.element"gt
• ltelement name"last"gt
• lttext/gt
• lt/elementgt
• lt/definegt
• ltdefine name"name"gt
• ltelement name"name"gt
• ltchoicegt
• ltgroupgt
• ltref name"first.element"/gt
• ltref name"last.element"/gt
• lt/groupgt
• lttext/gt

• first.element
• element first text
• last.element
• element last text
• name.element
• element name
• (first.element,
• last.element) text
• start name.element

17
Schema Language Traction

• W3C Schemas
• SOAP
• XForms
• UBL
• ebXML
• UDDI

• RELAX NG
• OpenOffice
• TEI
• DocBook
• XHTML 2

18
What About You?

• DTDs
• Existing, working system?
• W3C Schemas
• Choice may have already been made
• Or not trang converts to W3C Schema
• Transaction, DBMS oriented application?
• RELAX NG
• Publishing industry, or large-scale documentation
  effort?

19
Ask Yourself

• Requirements imposed by tools?
• Dependencies on tools?
• Standards?
• Complexity, flexibility of content structures?
• Business partner expectations?
• Forseeable reliance on language specs?
• Do you really need to move beyond DTDs?

20
Documents vs. Data, Schemas vs. Schemas

• Bob DuCharme
• XML 2004
• November 16, 2004