Maybe completely ignoring a domain is different than a implementing subroutine that ignores indexing a domain's images but not its other (text-based) content?