XSLT 2.0 and XPath 2.0 Programmer's Reference, 4th Edition (152 page)

Copying Type Annotations

The
validation
attribute of the

instruction gives you four options on how to handle type annotations:

• The
  strip
  option removes all type annotations from the element and attribute nodes being copied, leaving them only with the generic type annotations
  xs:untyped
  and
  xs:untypedAtomic
  , respectively. The main advantage of doing this is that you know exactly where you stand. This is the default, unless overridden using the
  default-validation
  attribute of the
  
  element, and it means that the behavior will be the same whether or not the processor is schema-aware.
  
• Sometimes the
  strip
  option is needed because the existing type annotations do not make sense in a new context. For example, you might be copying a
  price-range
  attribute from an element in which the name is constrained to be a sequence of two numbers, to an element in which any string is allowed. Retaining the type annotation would cause spurious effects, or errors, if the attribute in its new context is then used for comparisons or sorting.
  
• The
  preserve
  option leaves all the type annotations intact. One might expect that this would be the default, but the value
  strip
  was chosen for consistency with other instructions. Generally speaking, if the source nodes have been validated and annotated, this will often be the option that makes most sense on
  
  .
  

  This option works because the validity of an element or attribute depends only on its content (that is, on its attributes, its descendant elements, and their attributes). Cross-validation constraints such as
  ID/IDREF
  constraints are not taken into account by the XPath type system. This means that if the source data has a particular type annotation, you can take it on trust that the data is valid against that type, and if you copy the whole subtree, then the new nodes will still be valid against these types.
  

• The
  strict
  and
  lax
  options discard all existing type annotations, and then invoke the schema processor to validate the copied nodes. The same happens when you use the
  type
  attribute instead of the
  validation
  attribute. The way that validation works is exactly as described for the
  
  instruction in the case of element nodes (see page 312), or the
  
  instruction in the case of attribute nodes (see page 258), or the
  
  instruction in the case of document nodes (see page 304). For other nodes and atomic values the
  validation
  and
  type
  attributes are ignored.