Devonshire

KA raised the issue of what precisely is meant by the "URI" content of the @corresp attribute. According to the guidelines, the value is one or more of data.pointer, which is defined by RFC 2396. This is identical to the data type of ref/@target. The RFC does not mention XPath (it dates from 1998). According to the RFC, these are the components of a URI:

URI-reference = [ absoluteURI | relativeURI ] [ "#" fragment ]

However, the guidelines go on to specify a range of different options for the fragment component, including the use of XPath. According to the current incarnation of the guidelines (or at least, my reading of it), because the first component (absolute or relative URI) is optional, and its absence means that the current document is the target, we should be able to use XPath as in the following example:

#xpath1(//note[@xml:id='fn6']/p[1])

I think it's the case that my original formulations for the Devonshire look like this:

xpath1(//note[@xml:id='fn6']/p[1])

So we need to prepend the hash.