When the <relation>
element is being used in an RDF-like way to express a relationship between two arbitrary elements by means of URIs for each element and the relationship itself, the predicate URI is often recorded in the @ref
attribute. This makes the @name
attribute (currently required) a bit redundant, since it contains only xsd:name(s) and not pointers, so at best might contain a more human-readable version of the URI in @ref
.
Given this usage, I suggest that @name
be made optional on this element. (If we are nervous about allowing relationships to be expressed with no predicate defined, we could add a schematron rule to require one or other of @name
and @ref
.)
An alternative might be to make @name
a pointer, but that would break backward compatibility, and seems unnecessary.
I'd support this, now that we have @ref and @key. enforcing @name to be present seems to hark to an earlier simpler world
the Schematron test would have to say that one of @key or @name or @ref is needed.
Agreed. Unless Council objects, marking as Green and making Sebastian the Owner.
duly completed, with Schematron check