> That part is certainly true. To process the fragment (client-side > specialiser) one must first retrieve and dispatch the entire document, > according to it's media type as a whole, before then being > able to point to or position oneself at the particular fragment in question. Right. This is fundamentally different to sub-resource addressing.