FpML 5.0

1
FpML 5.0

• Experiences and Status

2
Current Model Summary

• FpML 1.0 thru 4.2 are all single namespace
• Keeps instance documents simple
• Products largely modelled from post trade point
  of view
• Strict type model ensures whole product must be
  specified
• A range of generic names given specific post
  trade interpretations.

3
Approaches to Modelling State

• Quotable products
• Limited number of namespaces
• Too many namespaces
• Schema generation

4
Quotable Products

• Introduced in FpML 4.0 to model RFQ process
• Only implemented for FX product
• Creates a new family of products
• Prefixed to avoid collision with existing
  products
• No direct relationship between pre-trade and
  post-trade types
• Disjoint inheritance structures
• Consistency by review only

5
Limited Number of Namespaces

• Small number of view schemas
• Constructed from chameleon schemas containing
  components
• Pros Cons
• Complex internal schema include structure
• Difficult to factor out view specific/generic
  types
• Difficult to create standalone valid subschema
  files
• High redundancy of types between views

6
Too Many Namespaces

• What if every FpML type had its own namespace?
• Most XML tools cant cope!
• Pros Cons
• All the complexity can be hidden in the schema
• But editing them is really hard

7
What have we learnt?

• XML Schema is not very good at representing
  different points of view
• Some of the schema features to create views have
  issues with tools
• Multiple namespaces in a schema or document
  should be avoided
• But small extension schemas work well
• Code generation is a much better option

8
Code Generation

• Software Engineering approach-
• Model everything UML
• Ontological approach-
• Model everything in OWL/RDF
• Pragmatic approach-
• Add annotations to XML Schema

9
Modelling in UML

• UML tools are commonplace
• But targeted at software rather than message
  design
• XML Schema is almost but not quite object
  oriented like UML
• XML Schema is grammar oriented
• A,B?,(C?(D, E)),F becomes flattened in UML
  field dependencies are lost.
• Order is difficult to express in UML
• Especially for associations
• Current ISO 20022 approach

10
Modelling in OWL/RDF

• W3C recommendation since 2004
• Part of the semantic web project
• Tools currently immature and only partial process
  coverage
• Properties are used to define and relate
  instances
• Classes identify instances having common
  properties (e.g. swaps vs. FX options).
• Classes can have relationships with other classes
  
• Incorporates both structural definition and
  business rules within a single set of definitions.

11
Extended Schema

• Use annotations to provide alternative values for
  XML facets in different views
• Type, Min Cardinality, Max Cardinality, etc.
• ltxsdelement name"relativeEffectiveDate"
  type"AdjustedRelativeDateOffset"gt
• ltxsdannotationgt
• ltxsdappinfo reportingminOccurs"0"
  confirmationskip"true"/gt
• ltxsddocumentation xmllang"en"gtDefines the
  effective date relative to the trade date,
  typically in months.lt/xsddocumentationgt
• lt/xsdannotationgt
• lt/xsdelementgt

12
Extended Schema

• Generate different views by processing
• An XSLT script takes the schema with annotations
  (extended) and generates the different view
  schemas (standard schema files)
• Each is a complete and disjoint schema no
  interlinking

Reporting Schema
Extended Schema
.xsl
Confirm Schema
13
Extended Schema

• Pros
• Instance documents remain simple
• Current design tools continue to work
• Each subschema file is valid
• Only types affected by views need annotations
• No redundancy on the source schema file
• ACCORD standard uses a similar technique

14
Extended Schema

• Pros Cons
• Annotations will not be drawn in design tool
• Increases the size of distributed FpML Schema
• Some work needs to be done on documentation
• Should we publish the source schema? Or only the
  view schemas?

15
Recommendation

• Generation from XSD schema is the most promising
  approach to date
• Some small XML problems to resolve
• Longer term OWL/RDF is the W3Cs strategic
  solution for information description
• Tools are currently immature
• But we should keep an eye on it