The 80-20% rule

• XML 1.0 validation PM is clear.
• Most users are fine with it, it is useful
• Some users need more specialized processing
• XSLT/XInclude shows similar pattern

A canonical Processing Model will help, wont fit 20% of the users

Defining the pipeline

No Application Programming Interface, please !

• expensive to define
• not flexible
• portability problems anyway

Data Centric Interface:

• Infoset is our data mode
• Schemas XInclude
• Not sufficient

Infoset Processing Module

• Requirements based on Infoset Core
• How is Core Infoset processed
• What extensions are recognized how they are processed
• How unrecognized elements are processed

=> Infoset Processing Module pipeline

Conclusions

• try to define a canonical pipeline
• define modules based on Infoset transformation
• Must stay a guideline, not a conformance requirement

KISS and do it quickly !!!