This is the plan that's emerging for handling the keywords:

• Keywords will be supplied without content, simply with a pointer to elements in the names.xml file.
• For display on the page, the system will look up the item in names.xml, and if it finds a child element with an attribute value n="1", it will use that item for display in the keyword box on the page. If it doesn't find an @n attribute, it will use the first child element in the linked item.
• Clicking on the keyword in the page should submit a different kind of search parameter, kwId, to the search page; this can be used to search very efficiently in the keywords for any instances of the item.
• The question remains as to how the text search engine is supposed to work. If searches are based on keywords, and keywords are essentially just xml:id attributes, then how can a user conduct a search? This needs some feedback from PAB.