WSI Submission

1
WS-I Submission

• W3C XML Schema User Experiences Workshop
• 21-22 June 2005
• Redwood Shores, CA, USA
• Erik Johnson, Epicor Software

2
Outline

• WS-I Schema Working Group Efforts
• Experience Report
• Schema Producers / Authoring
• Toolkits
• Testing Resources

3
WS-I Organization

• An open industry effort chartered to promote Web
  Services interoperability across platforms,
  applications and programming languages.
• A standards integrator to help Web services
  advance in a structured, coherent manner
• Approximately 130 member organizations
• 70 vendors, 30 end-user organizations
• Strong non-U.S. membership, including very
  influential Japan SIG

4
WS-I Schema Work Plan WG

• Work began in November 2004
• Discussions were high-level and constrained to
  using W3C XML Schema in web services
• Interoperability issues with the W3C XML Schema
  specification itself were not the problem
• Messages on the wire can be determined to conform
  to service descriptions
• Unsure that a profile (subset) is feasible
• Risks of unraveling the specification by
  disallowing features
• How else to describe an Infoset in a
  platform-independent way?

5
Schema Producers / Authoring

• Extensibility composition mechanisms
• Implementers need to create derivative schemas
  and compose schemas into WS descriptions
• Extensibility points are awkward to describe
• UPA is not known or correctly understood and some
  tools intentionally ignore the rule
• Modularity practices are not well-understood
• Some users attempt to abstract data from behavior
  in WSDL
• Namespace-based mechanisms are not agreed-to
• Need an element co-occurrence construct in
  message definitions
• Most attempts to create any modularity constructs
  are not likely to be recognized by human or
  machine consumers

6
Toolkit Support

• Inferring XML Schema documents to/from language
  types is difficult
• Special attribution or other extensions are used
  to help in serialization
• Some developers dont want to serialize objects
  others do
• Some want RPC semantics others dont
• Few implementers use XML Schema validation
• Some validation aspects get handled by
  combinations of type serializers, SOAP
  processors, and hand-rolled code

7
Toolkit Support (2)

• Many schema authors prefer designing schemas
  independent of any specific platform or toolkit
• Avoids bias
• Some want to leverage unique capabilities of XML
  in representing data
• This puts more pressure on downstream platforms
  and toolkits
• Broadens the set of XML Schema features and
  constructs generally used
• Modularity or other abstraction mechanisms well
  thought-out or not complicate the resulting
  schemas further

8
Toolkit Support Conclusions

• Interoperability issues seem to be witnessed
  more at design-time
• Ungraceful fallbacks when unsupported XML Schema
  constructs are encountered are a problem
• Little agreement what supported might mean
• What to do?
• Some feel that the WS-I should profile XML Schema
  and define a subset to make language mapping and
  programmability easier.
• Others feel that the XML Schema Specification
  itself is not the issue and that toolkits simply
  need to improve
• What is the fear?
• Web Service standards will split into camps of
  de-facto profiles
• Interoperability will suffer

9
Testing Resources

• Users want a way to test platforms and toolkits
  against XML schemas actually encountered
• Schema examples need to culled from real-world
  providers rather than some academic sample
• Tests might include running endpoints
• Different users have different expectations and
  requirements
• Ability to test locally if desired is ideal
• In the privacy of your own home
• Independent of individual (and possibly dated)
  toolkit claims

10
Conclusions

• Little support for creating an XML Schema profile
• Composition, versioning, and modularity are pain
  points for schema authors
• Toolkit support for some schema constructs is
  problematic
• But no agreement about the best course of action
• An improved set of test resources based on
  schemas in the wild would be appreciated
• With an ability to run them locally

11
WS-I Submission

• Thanks