This is a bit of a grey area that we deliberately avoided with PDFTribute.net; it seems perfectly acceptable to aggregate and index metadata, just as long as you're not actually storing any data.