identity constraints

[James Clark <jjc@jclark.com>]

> Recently, I spoke with Jerome Simeon.  I learned that there are three
approaches.
>
> 1) ID/IDERF (as in XML DTD)
>
> 2) XPath (as in XML Schema)
>
> 3) Types and path expressions (as in UCM)
>
> The first approach is very weak.

Agreed here.

>The second one is complicated.  In particular,
> it is to difficult to ensure consistencies of key declarations.

The XML Schema PR  uses a totally trivial subset of XPath.

http://www.w3.org/TR/xmlschema-1/#coss-identity-constraint

This is a major change from previous changes, and makes the XML Schema
identity constraint mechanism much more reasonable.

> The third
> approach appears to be powerful, and is described in "A Unified Constraint
Model
> for XML (WWW 10)" (http://www.cis.temple.edu/~fan/papers/xml/ucm.ps.gz)

I am not very impressed by UCMThe paper claims it is radically different
from XML Schema.  I don't think it is.

- XML Schema is also using a totally trivial path language.

- In XML Schema, <key> and <keyref> elements occur as subelements of
<element> elements.  <element> elements in XML Schema are declaring types:
types consisting of a single element.   Thus in XML Schema constraints are
being associated with types.

I don't understand how UCM works when the type with which a constraint is
being associated is not of the form

type foo = element bar [...]

It seems like you need to do type assignment to subsequences of the content
of an element.

Also I don't see how in UCM you indicate the scope within which a key
uniquely identifies instances of the type to which it is applied.

I would recommend a subsequent paper (of which Wenfei Fan is also a
co-author):

http://www.cis.temple.edu/~fan/papers/xml/keys.ps

Also:

http://www.cis.temple.edu/~fan/papers/xml/relative.ps.gz

(Note that the path language used by XML Schema is now quite close to the
path language used in that paper.)

I think the current XML Schema approach is not bad and we should use that as
our starting point.

James