Restricting XML Schema Identity Constraints: Two Possibilities

1.   Background

CR feedback has identified Identity Constraints as an aspect of XML Schema which has caused some concern. Two particular classes of observations have been made:

• Full XPath is a problem for implementors without access to pre-existing complete XPath implementations
• Streaming processors have problems: in the worst case implementing Identity Constraints effectively converts them into full-image-in-memory processors

2.   Options for simplification short of removal

Three main possibilities can be identified:

1. Restrict the inventory of XPath axes, predicates and functions which can be employed to reduce implementation burden;
2. Further restrict the inventory to render streaming processing much simpler.
3. In addition remove multi-part keys to render streaming processing of identity constraints essentially the same as processing ID/IDREF.

3.   Concrete proposals

The first two options above are described in concrete proposals:

1. Simplification: http://lists.w3.org/Archives/Member/w3c-xml-schema-ig/2000Mar/0047
2. Further simplification: http://lists.w3.org/Archives/Public/www-xml-schema-comments/2000OctDec/0520

4.   Examples

Examples of paths allowed under different options

Currently

foo @bar foo/baz foo/baz/@bar foo[@baz] /foo foo[baz] //foo .//foo ancestor::state//foo preceding-sibling::foo .//foo[@bar/2=self::*/@baz+13]

Simplification

foo @bar foo/baz foo/baz/@bar foo[@baz] /foo foo[baz] //foo .//foo ancestor::state//foo

Further simplification

foo @bar foo/baz foo/baz/@bar foo[@baz] /foo