This is an archived snapshot of W3C's public bugzilla bug tracker, decommissioned in April 2019. Please see the home page for more details.
This issue was originally reported by Khaled Noaman/Henry S. Thompson. Resolve the issues associated with restricting types whose elements include identity constraints. Specifically, (1) the rule must changed to state that the restricted type must have a superset rather than a subset of identity constraints, (2) the term superset must be clearly defined, and (3) there must be a way to redefine identity constraints in the restricted type without causing duplicate name problems. See http://www.w3.org/2001/05/xmlschema-rec-comments#pfiIdConsRestrict: R-94. Input from Straw Poll O-10 Interacts with RQ-17 (#restrictn-rules). This item was discussed in the meeting of 2004-04-09 (http://lists.w3.org/Archives/Member/w3c-xml-schema-ig/2004Apr/0178.html). We agreed that we would like to see the proposal written down. The minutes say "The description in the minutes may suffice as written proposal; if not, HT will write it up. Others to say if the minutes are insufficient." This item was discussed, and phase-1 agreement was reached, in the meeting of 2004-05-13 (http://www.w3.org/XML/Group/2004/05/xml-schema-ftf-minutes.html).
In the context of bringing forward a proposal for part of 2820 to the WG, this assumes increased urgency. The required change comes as an addition to the second sub-case of the mapping rules for xs:element (file:///d:/work/1.1_xmlschema/structures/structures.html#declare-element), along the following lines: An element declaration as in the first case above, with the exception of its {target namespace}, {scope} and {identity-constraint definitions} properties, which are as below: {identity-constraint definitions} Let the *local ICDs* be A set consisting of the identity-constraint definitions corresponding to all the <key>, <unique> and <keyref> element information items in the [children], if any, otherwise the empty set. Then iff the EII has <complexType> as an ancestor and the corresponding CTD has {derivation method} *restriction*, the union of the *local ICDs* and the *inherited ICDs*, otherwise just the *local ICDs*, where the *inherited ICDs* are the *inherited ICDs* of the corresponding Element Declaration, as defined in [3.4.6] And we add after the definition of Particle Actually Restricts a somewhat tricky definition of the *inherited ICDs* of an ED as either the always-the-same set of ICDs found on _all_ the EDs in a base content model which ever 'map to' that ED in its (restriction) content model, or the empty set.
A wording proposal to resolve this issue was adopted by the WG at its call of 12 January 2007. Accordingly, I'm marking the issue resolved. Henry Thompson, one of the originators, has been made aware of the resolution of the issue through his participation in the WG; no separate notification is being sent.