I guess I am really curious as to why the PDF’s have to be embedded, bore me with the long story :}

Unfortunately I can't get really get into the technical details, for reasons that have to do with confidentiality.

I am not committed to putting PDFs into a database -- but it's a design choice we'd like to try out to see what we can see about performance and how it simplifies the code. If we end up with 10,000 documents each of which is 10K (say), then we're over the 2G limit, and we have a problem. The existing solution uses a directory structure and works well -- but we're exploring alternatives.