apparently SPARQL is not going to help me with my web-harvesting / page-scraping task

What a pity! Would have been far too elegant. But if everybody makes use of RDF, stealing is just too easy.