RDF Schemas are purely descriptive, they don’t enforce anything (unlike XML Schema).

rdfs:domain and rdfs:range allow the inference of another additional triple, but don’t restrict the value of the subject or object of statements. In itself there’s nothing semantically wrong with:

pets:barkVolume rdfs:domain pets:Dog .
<#Basil> pets:barkVolume “10″ .
<#Basil> rdf:type pets:Cat .

though through an RDFS interpretation of the first two statements you can *also* infer that:

<#Basil> rdf:type pets:Dog .

i.e. given the above, Basil is both a cat and a dog.

(via http://www.w3.org/TR/rdf-mt/#RDFSINTERP)

If however you also stated:

pets:Dog owl:disjointWith pets:Cat .

that would make the three statements above inconsistent because no cat is a dog and vice versa (under an RDFS+OWL interpretation).

(via http://www.w3.org/TR/2004/REC-owl-guide-20040210/#DisjointClasses)