... itself¹

It could then lead to more than one fragment identifier in a URI... This would not be incompatible with [] but I've never seen such URI used.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... identifier²

In an XML syntax, XPointer ([3]) could be used.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... properties³

Actually, [1] defines a fourth property for statements, rdf:type, which must be valued by rdf:Statement. Since this is generalized by the concept of Class in [2], we don't mention it here.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

... RDF⁴

Furthermore, this constraint may seem too strong: cycles could be used to express equivalence between properties.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

...rdf:Property⁵

a common super-property, in my opinion, would have been more useful than a class, to federate those membership properties.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.