XML Schema Validation: Not Exactly a Science

1
XML Schema Validation Not Exactly a Science?
Presentation to CIOC XML Working Group

• January 21, 2004
• Ken Sall (ksall_at_SiloSmashers.com)
• based on xml-dev discussion initiated by Betty
  Harvey

2
Mixed Results with ET Schema

• Our ET Pilot XML Schema validated with 2 parsers,
  but failed validation with 3 parsers.
• It passed another applications parsing but then
  the application 'blew up' when trying to validate
  or parse an instance document of the schema.
• We were migrating between MS Windows and Linux,
  as well as using online XSV validator from W3C.
• All parsers that failed generated different error
  messages!
• We never really determined the exact problem.

3
XML Schema Checking Service

• XML Schema Validator (W3C) XSV 2.5-2 of
  2003/07/09 produced these error messages
• http//www.eccnet.com/ET-Register/InformationTechn
  ologyComponent.xsd7913 Invalid per
  cvc-complex-type.1.3 undeclared attribute
  NoneminOccurs
• http//www.eccnet.com/ET-Register/InformationTechn
  ologyComponent.xsd7913 Invalid per
  cvc-complex-type.1.3 undeclared attribute
  NonemaxOccurs

4
XSV.exe Command Line

• C\KENgt xsv InformationTechnologyComponent.xsd
• lt?xml version'1.0'?gt
• ltxsv xmlns'http//www.w3.org/2000/05/xsv'
  docElt'http//www.w3.org/2001/XMLSc
• hemaschema' instanceAssessed'true'
  instanceErrors'2' rootType'Anonymous' s
• chemaErrors'0' target'file/C/KEN/InformationTe
  chnologyComponent.xsd' validat
• ion'strict' version'XSV 1.6 of 2002/09/23
  214752'gt
• ltinvalid char'3' code'cvc-complex-type.1.3'
  line'791' resource'file/C/KE
• N/InformationTechnologyComponent.xsd'gtundeclared
  attribute NonemaxOccurslt/inv
• alidgt
• ltinvalid char'3' code'cvc-complex-type.1.3'
  line'791' resource'file/C/KE
• N/InformationTechnologyComponent.xsd'gtundeclared
  attribute NoneminOccurslt/inv
• alidgt
• lt/xsvgt

5
Altova XML Spy (also Corel XMetaL)
6
Betty Harvey's xml-dev 11/25/03 Posting

• What is the "Mother of All Schema Parsers"? 
• What schema parser can we trust? Are there any
  Gold Standards?
• Are the differences in the parsers because of the
  complexity and/or ambiguities of the Schema spec?
• Most important How can XML implementers use XML
  Schema in a collaborative, multi-tool environment?

7
Henry Thompson's Replies

• "Without being specific in a way which would be
  inappropriate given my W3C position, I would say
  that my experience over the last six months is a
  significant decrease in interop problems among
  the three or four best schema processors.
  Unfortunately one of the ones _not_ in that
  category is also one of the most widely
  used/distributed, which skews people's
  perceptions."
• "I have seen very few examples in the last six
  months where XSV, SQC and Xerces W3C, IBM's XML
  Schema Quality Checker, and Apache, respectively
  don't agree, and most of those arise because of
  areas of the REC which XSV has never
  implemented."
• "I definitely would not want this comment to be
  understood as a positive or negative assessment
  of any other product."

Yes, but which products aren't mentioned?
8
U.S. Government Recommends XML Schema (and DTDs)

• "Only ISO 8879 Document Type Definitions and W3C
  Schema Part 1Structures and W3C Schema Part
  2Datatypes SHALL be used to define XML document
  structures. Developers of data-oriented schemas
  in DTD syntax SHOULD migrate to XML Schemas.
  Developers MAY elect to use DTDs for markup of
  data that is strictly document-oriented
  (sentence, paragraph, chapter, appendix, etc.).
  However, the XML Schema language is the preferred
  method."
• -- CIOC XML Developers Guide, April 2002

9
Joe Chiusano and G.Ken Holman

• JC "I would also like to emphasize that the
  Federal XML Developer's Guide is guidance, and
  not government policy."
• GKH "Regardless, it gets treated as policy
  (and they should have realized that, much of what
  the government calls guidance gets treated de
  facto as policy) and the ripple effect has been
  felt far and wide by such guidance not to
  consider alternatives."
• Claude (Len) Bullard "A procurement official can
  usually waiver this stuff, but when it shows up
  in some documents, it can become a mandatory."
• GKH "My understanding is that non-technical
  influences impacted on the decision to use W3C
  Schema and not even regard available technologies
  such as RELAX-NG as viable candidates."

10
Elliotte Rusty Harold

• "I do not consider DTDs at all deprecated or
  legacy. I know the opposite opinion is out there,
  but it's wrong. I wrote about this in Chapter 24
  of Effective XML (not online yet, sorry), and I
  have been careful in all my talks about schemas
  to make sure that everyone knows they are not a
  replacement for DTDs. "
• "I also see a lot of evidence that the W3C XML
  Schema Language is losing the schema wars to
  RELAX NG. Many high profile groups have chosen to
  adopt RELAX NG for their schema needs rather than
  the W3C XML Schema Language. The prime reason for
  the W3C XML Schema Language's current and, I
  think, transitory prominence is merely the W3C
  imprimatur. Among developers who realize they
  have a choice, the choice is increasingly likely
  to be RELAX NG."
• "The real innovation of XML was not making the
  DTDs simpler. It was making them optional.
  Documents that do not have document type
  declarations are incredibly interoperable, with
  almost no room for parser differences in
  interpretation."

11
Dare Obasanjo

• "Do you mean that it validates according to the
  rules at http//www.w3.org/TR/xmlschema-1/ or the
  rules at http//www.w3.org/TR/xmlschema-1/
  http//www.w3.org/2001/05/xmlschema-errata ?"
• Ken Sall The issue here is that some processors
  may go strictly by the XSD 1.0 W3C Recommendation
  while others may have already incorporated
  changes indicated by the errata. But since
  there's no version number differentiating the
  two, there's no way to tell what a parser
  supports, except by contacting the product's
  techies.
• Note xmlschema-errata is 110 pages, as of
  1/15/04!

12
Possible Conclusions

• Hard to recommend any one XML Schema tool. XML
  Spy, although the most popular by far, has flaws.
  Henry Thompsons implicit recommendation might be
  XSV, SQC or Xerces , but these all lack GUIs.
• XML Schema is not an exact science. Too complex,
  too ambiguous, and has lots of errata external to
  XML Schema 1.0 spec.
• DTDs and even no schema at all are viable
  alternatives for some applications that dont
  require strong datatyping.
• RELAX-NG is worth investigating. Why do so many
  developers prefer it to XSD? RELAX-NG became an
  ISO standard in 12/03.
• For the next version of the XML Developers Guide,
  the CIOC should consider re-wording the strict
  use of XSD (i.e., might permit RELAX-NG OASIS
  RELAX TC).