"Other" is actually a fascinating concept in its own right, and to me reflects a fundamental limitation of the concept of enumerations in XML Schema. Controlled vocabularies are by definition "controlled" - adding, modifying or deleting terms from within a controlled vocabulary is done expressly by a curator. If you view such controlled vocabularies as unique text terms corresponding with associated keys, then the specific linear controlled vocabulary could be seen as the result of query against a larger ontology of resources. "Other" then implies a UX mechanism for notifying the curational system that their list in incomplete, and requires that there be a specific provision within the associated vocabulary or ontology for binding the label to a new entity and approving that entity for inclusion.
This means that in the model for the resource (something easier to express in OWL than in XSD unless you use annotations), a specific provision needs to be bound to the property (and possibly to the type of the subject) to indicate whether the list in question is closed (only a curator to introduce new terms), privileged (certain people can add new terms that have the given relationship), or open (anyone can add new terms, with the system itself assigning keys and approving or disapproving new entries (etc.)). This means that the rendering mechanism for the user interfaces (if dynamically generated) would be responsible for adding an "Other" field and exposing a means to put together a minimally sufficient entry for that field. In the simplest case, this may be a label, but it could in fact be a fairly exhaustive form in its own right.
Thus, "Other" should not be a part of any enumerated set per se, but the "Other" mechanism should provide the means to pass the constructor information to the curator or curation systems.
